Multimodal monitoring setup for stroke patient measurements and development of an easy-to-use user interface for the data collection PDF Free Download

Name: Multimodal monitoring setup for stroke patient measurements and development of an easy-to-use user interface for the data collection PDF
Author: kara

1 / 68

0 views•68 pages

Multimodal monitoring setup for stroke patient measurements and development of an easy-to-use user interface for the data collection PDF Free Download

Multimodal monitoring setup for stroke patient measurements and development of an easy-to-use user interface for the data collection PDF free Download. Think more deeply and widely.

Multimodal monitoring setup for stroke patient

measurements and development of an easy-to-use

user interface for the data collection

University of Oulu

Information Processing Science

Master’s Thesis

Jari Paunonen

2025

Abstract

Stroke is a leading cause of disability in a world, and it causes huge costs to healthcare

systems. Diagnosing stroke early as possible is important, because in a brief time window,

treatment should have been started to lower the chances of brain damage and improve

rehabilitation. Early diagnostic methods that are quick and easy to use and also help

recognize stroke mimics can reduce healthcare system costs and improve recovery for

individuals who have had a cerebrovascular accident. Software integrated into these

methods, such as user-friendly graphical user interfaces (GUIs) designed for medical

devices, plays a critical role. For individuals with limited tech familiarity, GUIs should

feature large, high-contrast text, intuitive navigation, and clear visual cues to

accommodate potential visual or cognitive impairments. Voice-guided instructions and

touch-based inputs can enhance accessibility, ensuring rapid and accurate use in high-

pressure clinical settings. From a medical device regulation perspective, such software

must comply with standards like IEC 62304, ensuring traceability, usability validation

through human factors testing, and robust data security to protect patient information.

Stroke project was multidisciplinary project, with several collaborators from private

companies to universities. Data collection was done by recording data from several

different modalities in Kuopio and Oulu. My goal was to realise the multimodal

measurement setup and conduct data recordings from stroke patients for the stroke

studies. Furthermore thesis will cover the theory of designing and developing Graphical

User Interface (GUI) for a wearable medical device prototype, focusing on designing GUI

for elderly adults or nurses with a medical device regulation point of view considered

also.

Keywords

Stroke, Multimodal, NIRS, Near Infrared Spectroscopy, Usability, Accessibility, GUI

Supervisors

Minna Isomursu, Teemu Myllylä

Tiivistelmä

Aivohalvaus on johtava vammojen aiheuttaja maailmassa ja se aiheuttaa valtavia

kustannuksia terveydenhuoltojärjestelmille. Aivohalvauksen diagnosointi

mahdollisimman varhaisessa vaiheessa on tärkeää, koska aivohalvauksen

diagnosoinnissa tunnistaminen mahdollisimman aikaisessa vaiheessa on hyväksi

aivovaurion mahdollisuuksien pienentämiseksi ja kuntoutuksen parantamiseksi.

Varhaiset diagnostiset menetelmät, jotka ovat nopeita ja helppokäyttöisiä ja auttavat

myös tunnistamaan aivohalvauksen oireiden jäljittelijöitä, voivat vähentää

terveydenhuoltojärjestelmän kustannuksia ja parantaa toipumista henkilöillä, jotka ovat

kokeneet aivoverenkiertohäiriön. Näihin menetelmiin integroitu ohjelmisto, kuten

lääkinnällisiin laitteisiin suunnitellut käyttäjäystävälliset graafiset käyttöliittymät

(GUI:t), on keskeisessä roolissa. Henkilöille, joilla on rajallista teknologiaosaamista,

käyttöliittymien tulisi sisältää suurikokoinen, korkean kontrastin teksti, intuitiivinen

navigointi ja selkeät visuaaliset vihjeet, jotta ne huomioivat mahdolliset näkö- tai

kognitiiviset rajoitteet. Ääniopastetut ohjeet ja kosketuspohjaiset syötteet voivat parantaa

saavutettavuutta, varmistaen nopean ja tarkan käytön korkean paineen kliinisissä

ympäristöissä. Lääkinnällisten laitteiden sääntelyn näkökulmasta tällaisen ohjelmiston on

noudatettava standardeja, kuten IEC 62304, varmistaen jäljitettävyyden, käytettävyyden

validoinnin ihmistekijätestauksella sekä vahvan tietoturvan potilastietojen

suojelemiseksi.

Stroke projekti oli monitieteellinen tutkimusprojekti, jossa osallisia oli yksityisistä

yrityksistä eri yliopistoihin. Tutkimusdataa kerättiin usealla eri modaliteetilla kahdessa

eri kaupungissa, Kuopiossa ja Oulussa. Minun tavoitteeni oli toteuttaa usean modaliteetin

mittausjärjestelmä ja kerätä dataa aivohalvauspotilailta, aivoverenkiertohäiriöiden

tutkimukseen.Lisäksi opinnäytetyöni kattaa teorian graafisen käyttöliittymän (GUI)

suunnittelusta ja kehittämisestä puettavan lääkinnällisen laitteen prototyypille, keskittyen

GUI:n suunnitteluun iäkkäille ihmisille ja hoitajille, ottaen huomioon myös lääkinnällise n

laitteen säätelyn näkökulman.

Avainsanat

Aivoverenkiertohäiriö, Multimodaalinen, NIRS, Lähi-infrapunaspektroskop ia,

käytettävyys, Saavutettavuus, GUI

Ohjaajat

Minna Isomursu, Teemu Myllylä

Foreword

Working on Stroke project and NIRS research project was very interesting for me. I have

background on electronics, information technology and healthcare and I was able to use

all my expertise which I have gained from my previous jobs.

I have been working couple years in healthcare where I did help several customers who

were rehabilitating from stroke. This experience was very useful for me when I was

working on Stroke project with patients.

Electronics and information technology experience has also been very useful when I was

working with NIRS project. Participating in electronical development by building and

repairing NIRS devices when needed and also part of this work was finding optimized

sensors and electrodes to work with our NIRS headband. Working on Teemu’s wearable

NIRS device has provided me many opportunities to learn more about prototype product

development.

During the project, we have managed to develop NIRS device from large suitcase size to

wearable device which can be attached for example to arm if needed.

Headband development has been challenging work. There were many challenges to

overcome like for example, materials to use, how they are connected to skin to avoid skin

reactions and what kind of pressure on skin is needed to have a good signal quality.

Headband development was mostly done by developing new iteration of 3D models and

testing them to find out what is working and what to change for new iteration of headband.

Software development has included working with multidisciplinary environme nt,

planning and testing new features of our software and during this thesis, making a new

interface for our NIRS software.

I did participate also in electronical development by building devices and repairing our

NIRS devices when needed. Part of this work was also finding optimized sensors and

electrodes to work with our NIRS headband.

Working with NIRS development, has also provided good opportunity to get familiar with

current research on health technology with NIRS.

In summary, ongoing project had led to development of wearable biomedical device, and

we have successfully gathered substantial amount of data for development stroke

diagnostics based on multimodal data.

LIST OF ABBREVIATIONS AND SYMBOLS

A/D Converter Analog/Digital Converter

ABC Activities specific Balance Confidence Scale

ADHD Attention deficit/Hyperactivity disorder

AHI Apnea-Hypopnea Index

AI Artificial Intelligence

ALS Amyotropic Lateral Sclerosis

BBB Blood Brain Barrier

BCI Brain Computer Interface

BMI Brain Machine Interface

BMI Body Mass Index

BOLD Blood Oxygen Level Dependant

CBF Cerebral Blood Flow

CDSS Clinical Decision Support System

CLIS Completely Locked-In State

CNS Central Nervous System

CSF Cerebrospinal Fluid

CT Computer Tomography

DC-EEG Direct Current EEG

DSF Disease State Fingerprint

DSI Disease State Index

DSR Design Science Research

ECG Electrocardiography

EEG Electroencephalography

EMG Electromyography

EU European Union

FES Functional Electrical Stimulation

fMRI Functional Magnetic Resonance Imaging

fNIRS Functional Near-Infrared Spectroscopy

FUS Focused UltraSound

GUI Graphical User Interface

HbO Oxy-haemoglobin

HbR Deoxy-Haemoglobin

HbT Blood Volume

Hz Hertz

ICH Intracerebral Hemorrhage

ICU Intensive Care Unit

ID Identification Number

KYS Kuopion Yliopistollinen Sairaala

LAPSS Los Angeles Prehospital Stroke Scale

MDR Medical Device Regulation

MEG Magnetoencephalography

ML Machine Learning

MRI Magnetic Resonance Imaging

mRS Modified Ranking Scale

NIHSS National Institutes of Health Stroke Scale

NIR Near Infrared

NIRS Near-InfraRed Spectroscopy

OSA Obstructive Sleep Apnea

OYS Oulun Yliopistollinen Sairaala

PAT Positive Airway Pressure

PET Positron Emission Tomography

PREM Patient Recorded Experience Measures

PROM Patient Reported Outcome Measures

PSD Persuasive System Design

PTSD Post Traumatic Disorder

QMS Quality Management System

RF Radio Fregquency

rPPG Remote Photoplethysmographic

SAH Subarachnoid Haemorrhage

SDB Sleep Disorder Breathing

SPECT Single-photon Emission Computed Tomography

TBI Traumatic Brain Injury

THL Työ ja hyvinvointilaitos

TIA Transient Ischemic Attack

TUG Timed up and go

UT-NIRS Ultrasound-Tagged NIRS

VE Virtual Environments

VOT Vascular Occlusion Test

VR Virtual Reality

VTT Valtion Teknologian Tutkimuslaitos

Contents

Abstract ............................................................................................................................. 2

Tiivistelmä......................................................................................................................... 3

Foreword ........................................................................................................................... 4

LIST OF ABBREVIATIONS AND SYMBOLS ............................................................. 5

Contents............................................................................................................................. 8

1. Introduction ................................................................................................................ 10

2. Background................................................................................................................. 12

2.1 An overview of stroke........................................................................................ 12

2.2 Stroke recovery .................................................................................................. 14

2.2.1 Common practise in rehabilitation.......................................................... 14

2.2.2 Emerging rehabilitation methods............................................................ 15

3. Stroke diagnostics....................................................................................................... 20

3.1 Digital solutions in clinical practice in stroke diagnostics................................. 20

3.1.1 Differential diagnostic challenges – stroke vs. migraine symptoms ...... 21

3.1.2 Cerebral circulation in individuals with stroke and sleep apnea ............ 22

3.2 Neuroimaging methods to diagnose stroke ........................................................ 22

3.2.1 PET and CT in stroke diagnostics .......................................................... 22

3.2.2 MRI in stroke diagnostics ....................................................................... 24

3.3 Emerging techniques for stroke diagnostics ...................................................... 24

3.3.1 Functional near-infrared spectroscopy (fNIRS) for stroke detection ..... 24

3.3.2 Electroencephalography for stroke detection ......................................... 27

3.3.3 Accelerometer for stroke detection? ....................................................... 28

4. Aim of the work and research questions .................................................................... 29

4.1 Research context ................................................................................................ 29

4.2 Research questions ............................................................................................. 29

4.3 Research method ................................................................................................ 30

4.4 Development goal .............................................................................................. 30

5. Design challenges in realizing easy-to-use software user interfaces.......................... 31

5.1 Generation-Oriented approach vs Aging-oriented framework .......................... 31

5.2 Generation-oriented approach............................................................................ 33

5.3 User interface for individuals with limited tech familiarity............................... 34

5.4 Dialogue and design solution ............................................................................. 34

5.5 Persuasive system designing on software development .................................... 34

5.6 EU regulations on software in medical devices ................................................. 36

6. Measurement methods ................................................................................................ 37

6.1 Fundus photography........................................................................................... 37

6.2 EEG and ECG .................................................................................................... 38

6.3 NIRS ................................................................................................................ 38

6.4 Speech and video recording ............................................................................... 39

6.5 Accelerometer .................................................................................................... 40

7. Materials and measurement protocols ........................................................................ 42

7.1 Data collection and research methods................................................................ 42

7.2 Research subjects in Stroke project ................................................................... 43

7.2.1 Inclusion criteria for research subjects ................................................... 43

7.2.2 Exclusion criteria .................................................................................... 43

7.2.3 Consent of subjects ................................................................................. 44

7.3 Sampling ............................................................................................................ 44

7.4 Recruitment ........................................................................................................ 44

7.5 Data collection protocols ................................................................................... 45

7.6 Balance protocol ................................................................................................ 45

7.7 Video and speech protocol ................................................................................. 46

7.8 Video speech protocol including additional measurement devices ................... 47

7.8.1 Additional measurement devices used.................................................... 47

7.8.2 Test protocol with additional measurement devices............................... 47

7.9 Fundus photography with Optomed Aurora ...................................................... 49

8. Development of the user interface for immediate measurement by NIRS................. 50

8.1 Users feedback ................................................................................................... 50

8.2 Technical description ......................................................................................... 50

8.3 The primary concept GUI .................................................................................. 50

8.4 Functions provided............................................................................................. 51

9. Results ........................................................................................................................ 54

9.1 Feasibility of the multimodal setup for stroke data collection in terms of

easy to use and measurement time ..................................................................... 54

9.2 Software interface and device development results when using NIRS

interface.............................................................................................................. 55

10. Discussion .......................................................................................................... 59

11. Conclusion ......................................................................................................... 63

12. References .......................................................................................................... 64

1. Introduction

Stroke is a critical health event that occurs when the blood flow to the brain tissue is

disrupted. There are two main types of strokes, ischemic stroke and haemorrhagic stroke.

An Ischemic stroke takes place when a blood clot blocks a blood vessel in the brain,

disrupting the blood supply to a part of the brain tissue. Ischemic strokes make up the

majority of stroke cases (about 79% according to the Finnish Stroke Registry 2008)

(Maretoja et al., 2011)

A haemorrhagic stroke takes place when a blood vein in the brain bursts and bleeds into

the surrounding tissue. Haemorrhagic strokes are more uncommon than ischemic strokes,

but they are generally more severe and have a higher mortality rate (Maretoja et al., 2011).

The consequences of a stroke depend on the location and the level of severity of the brain

damage, as well as the patient’s age and general well-being. Stroke can result in variety

of symptoms, including one-sided body weakness or paralysis, difficulty with speech and

understanding, visual disturbances, and problems with balance or dizziness (Maretoja et

al., 2011).

Stroke is a leading cause of death and disability in Finland. According to data from THL

(2017 Link) there are approximately 25 000 new cases of stroke in Finland each year, and

stroke is also the country’s third most frequent cause of mortality. The incidence of stroke

in Finland is higher among men than women and increases with age (Maretoja et al.,

2011)

The costs of stroke in Finland are substantial. Average costs are 25 000€ in a year after

the first stroke and on average 85 000€ across the entire remainder of lifetime after the

stroke. Expected survivability of stroke patients is rising, which will increase lifetime

costs of stroke patients in the future. This includes the costs of hospital care, medications,

and rehabilitation. Secondary costs of stroke, such as loss of productivity and the value

of informal care provided by family members, are also significant. Overall share of the

stroke patients from national healthcare spending was 7% at 2008) (Maretoja et al., 2011)

Risk factors for stroke in Finland include high blood pressure, high cholesterol, smoking,

atrial fibrillation and diabetes. Stroke prevention strategies in Finland include promoting

lifestyle changes to reduce these risk factors, as well as providing access to effective

treatments for conditions such as high blood pressure and atrial fibrillation (Maretoja et

al., 2011).

Treatment for stroke in Finland Typically involves a combination of medications and

rehabilitation. Rehabilitation after stroke may cover occupational therapy, physical

therapy and speech therapy, as well as support from social workers and other

professionals. (Maretoja et al., 2011)

It is vital to correctly identify the type of stroke a patient has experienced to provide

appropriate acute care and to predict the patient’s prognosis and implement preventive

measures. There are several diagnostic tests that can be used to evaluate patients with

acute ischemic stroke, and the appropriate test or combination of tests should be chosen,

based on factors such as effectiveness, potential risks to the patient, and cost. In some

cases, the availability or cost of certain diagnostic methods may pose challenges for

practitioners. (Kumar et al., 2011)

The development of advanced diagnostic imaging techniques and treatment options has

necessitated the optimization of diagnostic protocols while maintaining high standards of

care that are evidence-based, cost-effective and accessible. This has led to more accurate

diagnoses but has also made the process of clinical decision-making more complex.

(Kumar et al., 2011)

Stroke diagnosis has evolved significantly through software and AI, with ongoing

advancements enhancing speed, accuracy, and accessibility. Traditional software

processed CT/MRI scans using rule-based algorithms, but AI now employs deep learning

to detect strokes and differentiate mimics like seizures, often outperforming human

experts. Modern GUIs display clear visual alerts, though they often lack accessibility for

users with technologically limited skills. Future AI-driven systems will integrate

predictive analytics, multimodal data (imaging, symptoms), and personalized treatment

guidance, presented via intuitive, high-contrast GUIs with voice and touch inputs,

compliant with IEC 62304 for traceability and safety. These innovations promise reduced

healthcare costs by minimizing misdiagnosis and faster interventions, improving recovery

for stroke patients, especially when designed for usability and accessibility.

2. Background

This chapter covers all the background information to understanding the context, and the

theory behind the information presented in this thesis. This starts with introduction of few

different technologies which are used in hospital to monitor or taking images of brain.

Furthermore, thesis will cover the theory of designing and developing Graphical User

Interface (GUI) for a wearable medical device prototype, focusing on designing GUI for

people with limited experience using technology , with a medical device regulation point

of view considered also.

Substantial part of this thesis focuses on project work, where we were collecting

multimodal data from stroke patients in Oulun Yliopistollinen Sairaala (OYS) And at

Kuopio’s Yliopistollinen Sairaala (KYS) and from healthy test subjects.

Another major aspect on this thesis is wearable multimodal near-infrared spectroscopy

(NIRS) device project, where software is developed for nurses and home use of elderly

adult’s target group.

2.1 An overview of stroke

Stroke refers to a neurological disorder caused by the blockage of blood vessels, resulting

in brain clots that obstruct normal blood flow. This blockage can cause arteries to clog

and burst, resulting in bleeding. When arteries supplying the brain rupture during a stroke,

brain cells rapidly die from oxygen deprivation. Beyond physical symptoms, stroke may

also lead to depression and cognitive decline such as dementia. Stroke is also one of the

primary causes of death and leading factor in long-term disability across the world.

Developing countries have the highest incidence of stroke, with ischemic stroke

representing the majority of cases. Our knowledge of stroke pathophysiology has grown

considerably, revealing increasingly complex insights into its mechanisms and the

essential processes that cause ischemic injury. The main objectives of stroke treatment

are to re-establish blood circulation within the brain and address neurological harm

caused by stroke. As a result of inadequate outcomes in latest clinical studies, animal

models have been implemented in stroke research. While there have been notable

improvements in managing stroke, post-stroke care remains major challenge, impacting

not only families but also healthcare infrastructure and the broader economy. Advancing

both clinical and pre-clinical approaches is likely key to enhancing treatment success and

long-term recovery (Kuriakose & Xiao., 2020)

When a stroke occurs, it is due to insufficient blood flow to the brain through the blood

vessels. A comprehensive understanding of neurovascular anatomy. Including the

anterior cerebral artery and the two posterior vertebral arteries, is crucial. Ischemic

strokes occur when blood and oxygen are unable to reach the brain tissue, while

haemorrhagic strokes are the result of blood vessel rupture or leakage (Kuriakose, 2020)

Ischemic occlusion can cause intracerebral bleeding, potentially leading to thrombotic

and embolic conditions in the brain. Thrombosis happens when atherosclerosis causes

blood vessels to narrow, obstructing the flow of blood. The build-up of plaque in the

vascular chamber can eventually cause clots, resulting in a thrombotic stroke. An embolic

stroke, however, occurs when an embolism reduces blood flow to a specific brain region,

causing substantial stress and premature necrosis (cell death). As a consequence, the

plasma membrane is disrupted, organelles swell, leakage of cellular material into the

extracellular space leads to the disruption of neuronal activity. Vital factors like

excitotoxicity, leukocyte infiltration, blood-brain barrier dysfunction, free radical

toxicity, inflammation, glial cell activation, complement activation, disrupted

homeostasis, energy deficiency, oxidative stress, cytokine-driven cell damage, elevated

intracellular calcium, and acidosis are central to stroke pathology. (Kuriakose, 2020)

Haemorrhagic stroke, comprising 10–15% of strokes, carries a high mortality rate. Brain

tissue stress or internal injuries cause blood vessels to rupture, leading to this condition,

which damages the vascular system and causes infarction. It includes intracerebra l

haemorrhage (ICH) and subarachnoid haemorrhage (SAH). ICH results from vessel

rupture, with blood buildup in the brain. Key causes include elevated blood pressure,

heavy anticoagulant use, vascular irregularities, and thrombolytic agents. Subarachnoid

haemorrhage, on the other hand, features blood pooling in the brain’s subarachnoid space,

usually from a cerebral aneurysm or head injury. (Kuriakose, 2020)

Recognizing the abrupt onset of focal neurological deficits of waking up with them is

crucial in diagnosing ischemic stroke. The commonly experienced symptoms that occur

with transient ischemic attack (TIA) are speech problems and loss of strength affecting

one side of the body. However, stroke can be mimicked by other conditions such as

seizures, migraine with aura and/or headaches, conversion disorder and hypoglycaemia.

Obtaining a patient’s medical history and conducting diagnostics tests can usually help to

rule out stroke mimics. To differentiate between TIA and intracerebral haemorrhage, in

addition to identify non-stroke conditions, neuroimaging is vital. Neuroimaging selection

is based on accessibility, thrombolysis eligibility, and contraindications. Subarachnoid

haemorrhage (SAH), typically indicated by a sudden, severe headache, is optimally

diagnosed with a noncontrast head computer tomography (CT) (Yew, K.S., Cheng, E.M.

2015)

The symptoms of acute stroke are often confusing and can be misunderstood by both

patients and clinicians. It is the responsibility of family physicians to identify and treat

acute cerebrovascular diseases promptly. Family physicians are not available in every

healthcare system, so then responsibility to identify stroke is on the physician treating the

patient. A prompt and precise evaluation of individuals exhibiting stroke symptoms can

minimize disability and reduce the chances of recurrences. Stroke can be categorized

based on the pathological mechanism and the affected vascular distribution.

Understanding the underlying pathological mechanism is crucial for determining

appropriate treatment decisions such as antithrombic therapy, thrombolysis, and

predicting prognosis. Mortality rate is greater for haemorrhagic stroke when compared to

ischemic stroke. According to Yew et al, of all strokes in the U.S., 87% are ischemic,

caused by large-artery atherosclerosis, small vessel occlusion, cardio embolism, or other

unspecified reasons. The other 13% are haemorrhagic, located in subarachnoid or

intracerebral regions. (Yew, 2015)

Clinicians typically diagnose stroke with moderate to good reliability, but those with less

experience or lower confidence show reduced diagnostic accuracy. Ischemic stroke

commonly presents with sudden onset of awakening with symptoms where unilate ral

weakness and speech disturbances are the most frequently observed physical findings.

While determining the time of onset does not aid in diagnosing stroke, but this

information is critical, as it establishes whether a patient qualifies for the 3–4.5-hour

thrombolysis eligibility window in ischemic stroke cases, guiding precise treatment

decisions. Physicians who are responsible for treating acute stroke patients must develop

knowledge of National Institutes of Health Stroke Scale (NIHSS). This scale

compromises 15 items and can be completed in approximately five minutes. While it can

aid in distinguishing stroke from other conditions that mimic stroke, its primary purpose

is to accurately assess the severity of a stroke, thereby enabling physicians to determine

whether administering tissue plasminogen activator is suitable. (Yew, 2015)

2.2 Stroke recovery

For many years, it was widely thought that the brain was rigid, unchanging structure and

that once an even such as a stroke occurred, brain structures and functions were

permanently lost. Recently our knowledge about motor learning, brain plasticity and

functional recovery following brain injury has greatly improved. Breakthroughs in basic

neuroscience have spurred further research in motor rehabilitation. The brain has a

remarkable ability to adapt and change its neural circuits, a phenomenon known as

plasticity, which can result in spontaneous recovery. Rehabilitative training can enhance

this process and enhance the brain’s plasticity. The restructuring of the remaining

components of the central nervous system helps with behavioural recovery. This can

occur through changes in interhemispheric lateralization, increased activity in the

association cortices connected to the injured area, and the reorganization of cortical

mapping. According to Hara, animal studies in animal models indicate that both motor

learning and cortical stimulation could modify intracortical inhibitory circuits, promoting

long-term potential and remodelling of the cortex. Specifically, the use of

Electromyography (EMG) controlled electrical muscle stimulation has been shown to

enhance the motor function of the hemiparetic hand and arm. According to Hara, NIRS

research indicates that, during non-invasive measurement of brain haemoglobin levels in

functional tasks, the greatest cerebral blood flow in the damaged sensory-motor cortex is

observed during an EMG-controlled functional electrical stimulation session. (Hara,

2015)

According to Hara, in 2015 there were no studies of brain activity during a therapeutic

Functional Electrical Stimulation (FES) intervention, due to the electrical stimulation

obstructing assessment methods such as fMRI, Single-photon emission computed

tomography (SPECT) and PET. These functional neuroimaging techniques cannot be

used during a typical FES intervention. NIRS has been recently developed as a

neuroimaging method that is not affected by electrical stimulation, making it possible to

assess brain activity during FES interventions in rehabilitation setting. NIRS offers

several advantages, including non-invasiveness, high sensitivity, a natural environment,

low operating costs and portability. It measures changes in concentration of deoxygenated

and oxygenated haemoglobin, which have been linked to regional cerebral blood flow

(CBF) and are considered a reflection of cortical activation. (Hara, 2015)

2.2.1 Common practise in rehabilitation

According to Finnish käypähoito (Duodecim, Käypä Hoito, 2024), the usual

rehabilitation path for stroke patient begins with a rehabilitation assessment made for him

with a week or as the overall condition allows. The assessment is based on the patient’s

need for physical and cognitive rehabilitation. The assessment requires an evaluatio n

made by a multidisciplinary expert team, about who benefits from rehabilitation.

The need for rehabilitation and individual goals are recorded in a rehabilitation plan,

which is prepared in cooperation with the unit responsible for care, the patient and his

relatives, and a multidisciplinary team. Healthcare professionals act as liaisons with

hospital districts and primary healthcare. They have a central role in the early stages of

the rehabilitation path and in monitoring the implementation of the rehabilitation plan. If

necessary, cooperation is also carried out with occupational healthcare, insurance

institutions, and Kela.

The mortality of those treated in a multidisciplinary rehabilitation unit and the risk of

remaining in permanent institutional care, are smaller than for rehabilitators treated in a

ward. The effectiveness of the treatment is reflected in shortened treatment times, reduced

disability, and improved quality of life. The multidisciplinary rehabilitation team includes

doctor, physiotherapist, nurse, social worker, occupational therapist, speech therapist,

neuropsychologist, liaison officer, and later a rehabilitation counsellor.

The most typical forms of therapy are, for example, physiotherapy, occupational therapy,

speech therapy, and neuropsychological rehabilitation. Physiotherapy can include various

walking exercises to promote walking ability and upper limb function enhancement with

different exercise methods. Increasing the intensity of physiotherapy improves mobility

rehabilitation, and the decisive factors are an early start and the practice of the desired

feature of skill.

In supportive therapy, the aim is to promote opportunities to act independently in various

everyday tasks or at work. The methods in the exercises can be various task-focused

repetition exercises, enhanced hand rehabilitation, mental image training, strength

training, or training in virtual reality (VR).

In speech therapy, communication skills and functional capacity as well as swallowing

functions are rehabilitated. The goals of speech therapy can be related to the

rehabilitator’s interaction ability. For example, communication guidance give to an

aphasic person can support the activity of the person being treated.

In neuropsychological rehabilitation, where cognitive disorder rehabilitation has been

deemed appropriate based on a neuropsychological examination, it is carried out using

neuropsychological methods. Rehabilitation is targeted at cognitive disorders and

behavioural changes as well as symptom awareness.

2.2.2 Emerging rehabilitation methods

Virtual Reality and Virtual Environments

Neuropsychological rehabilitation treatment options have undergone advancements in

recent times with the integration of technology like VR. Virtual environments (VE) offer

not only advanced diagnostic tools and additionally enable healthcare providers to create

highly targeted rehabilitation programs. Limited VR programs have been created to help

individuals undergoing neurological rehabilitation improve their ability to perform daily

activities. (Ansado et all., 2021)

Over the past 10 years, VR has been combined with both conventional and state-of-the-

art neuroimaging methods, resulting in two potential outcomes: 1. Charting neural

networks pre- and post-VR training, and 2. Live monitoring of brain activity while

performing VR tasks. Common neuroimaging methods utilized include

electroencephalography (EEG), Functional magnetic resonance imaging (fMRI) and

functional near-infrared spectroscopy (fNIRS) are critical for studying brain function in

stroke diagnostics. The integration of VR with these neuroimaging techniques is

increasingly researched, providing neuropsychologists with promising tools to enhance

stroke rehabilitation and differentiate stroke mimics. Participants in various clinical

studies have benefited from monitoring their cerebral activity during VR immersion. This

approach, often using fNIRS or fMRI, supports neuropsychologists in developing

targeted treatments, particularly for stroke rehabilitation and cognitive recovery. While

the emphasis is primarily on enhancing motor abilities and balance in conditions like

cerebral palsy and Parkinson’s, as well as utilizing exposure therapy to treat anxiety and

post traumatic disorder (PTSD), the current literature review focuses on exploring how

fNIRS, EEG, and fMRI integrated with VR can improve the rehabilitation of cognitive

functions, such as memory and attention, by allowing clinicians to identify biomarkers

that predict therapeutic success. This approach holds significant potential for stroke

recovery and personalized treatment planning. Particularly, the current examination

focuses on the rehabilitation of stroke, attention deficit disorder (ADHD), , traumatic

brain injury (TBI), and with or without hyperactivity as they are among the strongest

examples of recent advancements in VR applications and the potential benefits of

simultaneous neuroimaging. Ansado et al claims that by combining multimodal

neuroimaging with VR, a neurorehabilitation program can be tailored to an individual’s

unique brain activity pattern as affected by their specific traumatic event. Multimodal

neuroimaging can bring significant benefits to the developing field of VR-based

neurorehabilitation. (Ansado et al., 2021)

Neuroimaging is a technique used to determine which areas of the brain respond to

training or stimulation and evaluate the level of activation. Advanced brain imaging

techniques used in clinics help scientists understand how cognitive training influences

neuroplasticity. These methods track brain volume, activity, and connectivity by

observing blood flow in fine detail. Another tool, EEG, records brain waves using

electrodes on the scalp, offering a clear picture of brain activity over time. The data is

then analysed for unusual patterns or important signals. Lastly, fNIRS uses near-infrared

light to visualize changes in blood flow and haemoglobin levels, making real-time

imaging of brain activity possible. (Ansado et al., 2021)

In fMRI studies, participants are typically required to keep their head and body still while

lying in the scanner. To perform task, they often move a hand or a finger in a VR world

displayed through and head mounted device (HMD) or built into MRI scanner. An

alternative approach involves conducting neuroimaging scans on participants at multiple

stages during, after, and before treatment, to assess cerebral changes and enhance stroke

rehabilitation strategies. Unlike fMRI, EEG equipment is portable and can be used while

participants are sitting or walking within a restricted area, depending on the experiment’s

requirements. fNIRS offers balance between portability and temporal resolution like

fMRI, offering the best of both worlds. The choice between of which imaging technology

to use in VR training depends on the desired outcome, as each of the three techniques

provides unique information (Ansado et al., 2021).

EEG Typically provides sufficient temporal information for neurofeedback or simple

brain activity monitoring during performance. Advanced neuroimaging techniques, such

as fMRI and fNIRS, facilitate precise mapping of active brain regions and detailed

analysis of activation changes over time, particularly during, after, and before VR-based

training. In stroke rehabilitation, however, therapists encounter challenges with VR

applications, primarily due to their initial design for general audiences rather than

therapeutic purposes, necessitating specialized adaptations. However, developers have

made advancements in this area by incorporating features that are relevant to therapy,

such as adjustable intensity levels based on patient’s progress, tasks that can be controlled

by the therapists, and the ability to track and records movements to provide clear picture

of the patient’s progress and recovery. The features mentioned are also among the key

benefits of using VR in rehabilitation, particularly when combined with neuroimaging.

According to findings from a meta-analysis, this technology is generally associated with

beneficial results and is not linked to adverse outcomes. (Ansado et al., 2021)

Brain-Computer Interfaces

Brain-Computer Interfaces (BCIs) are technologies that enable control of computers or

other devices through brain activity, circumventing the peripheral nervous system. They

provide a vital communication pathway for those with significant motor impairments or

in vegetative states. (Naseer & Hong, 2015)

BCIs utilize brain activity to capture a user’s intended actions. The initial phase of

creating an fNIRS-BCI system involves gathering relevant brain signals, typically from

the primary motor cortex and prefrontal cortex. Signals from the motor cortex relate to

motor execution and imagery, while those from the prefrontal cortex stem from activities

like mental arithmetic, counting, musical imagery, or visualizing landscapes. Different

emitter-detector arrangements are used in these regions, but the distance between emitters

and detectors is kept within a specific range due to its significance in fNIRS

measurements. For example, a wider emitter-detector gap allows deeper brain imaging.

Naseer et al. suggested a 3 cm separation to effectively measure hemodynamic signals

from cortical regions. A gap less than 1 cm could mainly detect skin-layer signals, while

a separation exceeding 5 cm could result in weak, inconsistent signals. (Naseer et al.,

2015)

The prefrontal cortex activities are a favourable option for fNIRS-BCI as they generate

fewer motion artifacts and signals that are not as impacted by hair movement.

Additionally, they tend to be more effective for individuals with motor-related

disabilities. Given these benefits, most research has focused on prefrontal activities,

yielding encouraging outcomes. Some of the frequently employed prefrontal activities in

fNIRS-BCI include mental arithmetic, musical visualization, mental counting, and

landscape visualization. (Naseer et al., 2015)

Recent advances in fNIRS-BCI research have been notable, but applications remain

largely confined to training and demonstration. Two key limitations hinder real-world

use, slow information transfer rates and high error rates. Moreover, most evaluations

occur under controlled lab conditions, which do not reflect the difficulty of maintaining

focused mental effort in real-world environments. (Naseer et al., 2015)

BCI technology can aid in restoring lost motor and cognitive functions in individ uals

affected by stroke or spinal cord injury. This is made possible through the ability of BCI

feedback to activate self-regulation of brain activity. According to Naseer, previous

neurofeedback studies, EEG has been extensively used due to its high temporal

resolution. However, EEG has limitations such as imprecise localization and difficulty

accessing subcortical regions, making fMRI, which measures hemodynamic activity, a

more suitable option for neurofeedback studies to overcome these challenges. (Naseer et

al., 2015)

FNIRS is an appealing option compared to fMRI for accessing subcortical brain signals

due to its affordability, ease of use and portability. Because it can be worn on the head or

body, fNIRS is easy to use—even in an ambulance and it tracks brain activity faster than

many fMRI machines. Since it isn’t easily affected by movement, it holds strong promise

for use in neurofeedback studies. (Naseer et al., 2015)

Brain-Machine Interfaces

Brain-Machine interfaces (BMI) allow paralyzed patients to interact with their

environment by using their brain activity to control external devices. For example,

possible target groups can be patients with amyotrophic lateral sclerosis (ALS) or

restoration of motor impairment of people after severe stroke incident. (Chaudhary,

Birbaumer & Curado, 2015)

BMI offer promising solution for communicating with paralyzed ALS patients, as they

do not require muscle involment. Different methods have been explored to control BMI’s

for communication, including EEG and NIRS. According to Chaudhary et al, Previous

research indicates that EEG-BMI supports communication in ALS patients who still have

limited eye control, but it is ineffective for those with total paralysis. However, the recent

introduction of fNIRS-BMI has opened new avenues for communication with completely

locked-in state (ALS) patients who have completely lost the ability to move any muscle,

even in cases of complete locked-in syndrome. (Chaudhary et al., 2015)

Aside from assisted communication, BMIs are also being thoroughly studied for their

potential in promoting motor recovery after a stroke. This BMI method associates brain

signals related to a patient’s intent to move an impaired limb with sensory feedback from

that movement, aided by rehabilitation aids. Most studies use magnetoencephalography

(MEG) or EEG systems for their strong temporal resolution, supporting real-time

synchronization of movement intent and sensory feedback. A recent controlled study by

Chaudhary et al. found that EEG-BMI training significantly enhances motor function in

stroke patients with severe weakness. (Chaudhary et al., 2015)

While the conditions of locked-in state have motivated research in this area, there are still

only few systems which have been successfully used on this population. Three successful

technologies used are EEG-BMI, fMRI-BMI and fNIRS-BMI. In comparison to fMRI, a

NIRS based BMI’s can be conveniently used to assist bedridden and immobile patients

who desperately require communication. (Chaudhary et al., 2015)

According to Chaudhary et al., the first controlled assessment of a NIRS-BMI achieved

an 89% accuracy rate in differentiating right- and left-hand imagery with a 20-channel

NIRS system over the sensorimotor cortex, using a hidden Markov model classifier and

no prior training. Recently, NIRS was employed successfully to analyse functional

cortical responses in a complete locked-in syndrome (CLIS) patient to auditory stimuli,

such as true or false statements and open-ended questions. Hemodynamic changes in the

motor cortex were tracked over multiple sessions for more than a year to train a classifier

that accurately predicted “yes” or “no” responses at a rate of 71.76%. This was a

significant improvement, as earlier EEG-BMI systems had not yielded successful

outcomes. (Chaudhary et al., 2015)

Stroke is a major contributor to acquired disabilities among adults worldwide. Repetitive

motor task has been effective in restoring motor function in patients with incomplete hand

paralysis, but people with serious hand weakness often cannot benefit from regular

rehabilitation because they cannot do the exercises. But BMI training a system that uses

brain signals could help them regain movement. It works by helping the person repeatedly

try movements with their weak limb, while a device gives them visual and physical

feedback to support the process. (Chaudhary et al., 2015)

3. Stroke diagnostics

This chapter describes different challenges, methods and technologies on related to stroke

diagnoses.

3.1 Digital solutions in clinical practice in stroke diagnostics

Diagnosing posterior circulation strokes can be difficult. Dizziness is a common

complaint among patients but is rarely a sign of stroke. Research on adults aged 44 or

older who sought emergency care or were directly hospitalized with dizziness as their

primary symptom found that only 0.7% of those with isolated dizziness received a

diagnosis of stroke or TIA. However, in 44% of cases, the initial examiner failed to

identify their stroke, highlighting the need for precise diagnostic tools. (Yew & Cheng,

2015)

Among the patients who exhibit acute vestibular syndrome characterized by continuous

vertigo or dizziness lasting for at least one hour, combined with gaze-evoked or

spontaneous nystagmus, vomiting or nausea, intolerance to head motion and newly

developed gait unsteadiness, at least one-quarter may have a posterior circulation stroke.

Notably, up to two-thirds of patients with stroke-related acute vestibular syndrome may

lack obvious neurological symptoms. (Yew et al., 2015)

Neuroimaging is the only reliable method for accurately distinguishing between

haemorrhagic and ischemic stroke. While patients experiencing haemorrhagic stroke may

exhibit symptoms such as vomiting, headache, meningismus, diastolic blood pressure

exceeding 110 mm Hg or coma, these individual or combined findings are not conclusive

enough to make a definitive diagnosis. (Yew et al., 2015)

Subarachnoid haemorrhage (SAH) manifests differently compared to intracerebral

haemorrhage or ischemic stroke. In cases of aneurysmal SAH, approximately 80% of

patients experience a sudden, extremely severe headache. Another crucial historical

discovery is the presence of a prior sentinel headache occurring between two to eight

weeks before rupture of a cerebral aneurysm, which is found in up to 40% of SAH

patients. Accompanying the headache can be symptoms such as photophobia, vomiting,

seizures, neurological signs, meningismus and a reduced level of consciousness.

Funduscopy is essential, as intraocular haemorrhage is present in one in seven patients

with aneurysmal SAH. With bleeding occurring outside the brain, SAH patients may lack

focal neurological signs. (Yew et al., 2015)

When examining a patient with suspected stroke, healthcare providers should consider a

wide range of potential diagnoses. Conversion or somatoform disorder, seizures,

hypoglycaemia and migraine headaches are some of the most common conditions that

can mimic stroke. To determine if a patient is eligible for intravenous thrombolysis,

healthcare providers use checklists that explicitly require the detection of

hyperglycaemia, hypoglycaemia and recent seizures. The percentage of stroke

misdiagnosis in studies involving consecutive patients who did not undergo thrombolysis

ranges from 25% to 31%. On the other hand, among patients who received thrombolysis,

the proportion of individuals diagnoses with a stroke mimic ranges from 1.4% to 16.7%.

(Yew et al., 2015)

All patients with suspected stroke symptoms should undergo immediate neuroima ging

via MRI or noncontrast computed tomography (CT). The primary goal in suspected

ischemic stroke cases is to rule out non-ischemic central nervous system lesions and

distinguish ischemic from haemorrhagic strokes. Noncontrast CT effectively identifies

mass lesions, such as brain tumours or abscesses, and acute haemorrhage, but its

sensitivity for detecting strokes within three hours post-infarction is limited, with fewer

than two-thirds of cases identified. Furthermore, it exhibits reduced sensitivity for small

or posterior fossa strokes, necessitating advanced imaging for comprehensive diagnostics.

(Yew et al., 2015)

Multimodal MRI sequences, especially those involving diffusion weighted images,

provide superior resolution compared to noncontrast CT scans. Consequently, they

exhibit higher sensitivity in identifying acute ischemic stroke. The sensitivity of MRI

sequences in detecting intracerebral haemorrhagic stroke is comparable to that of

noncontrast CT scans. Although MRI offers superior resolution than noncontrast CT, the

latter is more widely available, faster and less expensive imaging modality. Additionally,

noncontrast CT can be safely performed in individuals with implanted devices or

claustrophobia. When a patient is eligible for intravenous thrombolytic therapy within a

specific time windows, guidelines recommend using noncontrast CT or MRI to exclude

intracerebral haemorrhage and assess for ischemic changes. Compared to noncontrast CT,

MRI has a lower misdiagnosis rate in individuals under 55 years of age who present with

stroke-like symptoms. This is attributed to the lower occurrence of vascular risk factors

and higher incidence of central nervous system stroke mimics in this age group. (Yew et

al., 2015)

3.1.1 Differential diagnostic challenges – stroke vs. migraine

symptoms

Most of neurological diagnostics is based on local neurological symptom identifying and

analysing. Most common symptom is migraine aura which is important to identify from

temporary transient ischemic attack (TIA). Migraine aura is mostly visual disorder which

includes positive (vibrations, stars) and negative (partial loss of sight) symptoms and it

increases and fades away in 5-60min. TIA is more suddenly happening and lasting shorter

time. TIA is usually having negative symptoms like partial loss of something like sight,

speech, able to swallow, eye movement and muscle strength without mentioned migraine

symptoms. Migraine aura also usually has headache while TIA does not. Clinically is

known cases where stroke has happened with in connection of long-lasting migraine aura.

Less than 1.5% stroke cases are counted as migraine strokes.

In practice it can be hard to decide when a local symptom can be monitored as migraine

aura and when patient should be sent to more tests because of TIA or stroke suspicion.

Early correct diagnose will affect on treatment and leads coming tests to correct way.

Migraine aura without complications does not need efficient treatments. TIA on other

hand is warning sign of threatening brain catastrophe, and it will need quick action for

starting treatments. Early differentials diagnose of TIA and migraine is mostly done from

discussing with patient because usually patient’s condition is normal when he comes for

doctor. Especially patients own precise description of symptoms is vital for diagnosing

TIA or migraine aura and it will speed up for treatment decisions (Kallela et al. 2012,

Duodecim). In Stroke project we are using fNIRS/DC-EEG device, which is developed

in University of Oulu and it should help on developing differential diagnose between TIA

and migraine aura.

3.1.2 Cerebral circulation in individuals with stroke and sleep apnea

Sleep-disorder breathing (SDB) is more common individuals who have experienced a

stroke compared to those without a stroke history. SDB is a separate risk factor for stroke

and can harm cerebral circulation through various mechanism, potentially contributing to

wake-up strokes. Ultrasound-tagged NIRS (UT-NIRS) is an innovative technology that

can non-invasively and in real-time detect cerebral blood flow and present it as the

cerebral flow index (CFI). The most effective treatment for SDB is positive airway

pressure (PAP). (Jurik M. et al., 2021)

In recent times, the prevalence of sleep-disordered breathing (SDB) has surged,

predominantly due to escalating obesity rates. The primary subtype, obstructive sleep

apnea (OSA), is identified when the apnea-hypopnea index (AHI) is 5 or more events per

hour, impacting roughly 24% of men and 9% of women globally. Strikingly, among

individuals with a prior stroke or TIA, SDB prevalence reaches as high as 72% after

adjusting for age. As a confirmed independent risk factor for ischemic stroke, OSA

exacerbates cerebrovascular pathology. Over time, OSA leads to elevated blood pressure,

imbalanced autonomic function with heightened sympathetic activity, impaired

endothelial function, metabolic changes, and a pro-inflammatory state. PAP is the top

choice for treatment and the most successful method for managing SDB. (Jurik et al.,

2021)

Alzheimer Disease (AD) has also been connected to SDB, with occurrence of over 50%.

Individuals who have SDB often encounter cognitive problems, such as deficiencies in

memory and attention, which are connected to changes in neuroanatomy and loss of grey

matter. It is worth noting that SDB can also exacerbate cognitive decline in the elderly,

and OSA is associated with neurophysiological indicators of multiple brain functions.

The frontal cortex, which is closely linked to memory, attention, and higher cognitive

abilities, as seen in AD patients, was the brain region where the hemodynamic response

was assessed. Findings of Ferdinando et al seems to support correlation between SDB

and AD. (Ferdinando et al., 2022)

According to Jurik et al., multiple studies have employed NIRS to assess cerebral blood

flow (CBF), demonstrating reduced cerebral oxygenation during OSA episodes,

particularly in severe cases. Despite NIRS’s advantages, including non-invasiveness,

continuous monitoring, and high temporal resolution, it is limited to measuring tissue

oxygenation rather than cerebral perfusion. A novel technology, ultrasound-tagged NIRS

(UT-NIRS), integrates NIRS’s strengths with ultrasound’s capability to track blood cell

movement, enhancing perfusion assessment. (Jurik M. et al, 2021)

3.2 Neuroimaging methods to diagnose stroke

3.2.1 PET and CT in stroke diagnostics

Positron Emission Tomography (PET) scans, in conjunction with Computed Tomography

(CT), offer a comprehensive approach to diagnosing strokes, leveraging the strengths of

both modalities for enhanced accuracy and precision. While PET imaging provides

functional information about tissue metabolism, CT imaging offers detailed anatomical

images of the brain’s structure.

PET scans represent a pivotal advancement in medical imaging technology,

revolutionizing the diagnosis and treatment of numerous diseases. At the heart of PET

scanning lies innovative fusion of nuclear medicine and computerized imaging

techniques. (Orhii et al., 2023 )

PET scans utilize radiopharmaceutical tracer, typical a molecule tagged with radioactive

isotope, administered into patient’s bloodstream. As the tracer circulates, it accumulates

in tissue with high metabolic activity, such as tumors or areas of inflammation. Crucially,

this process allows PET scans to visualize not just anatomical structures but also

functional processes within the body. (Orhii et al., 2023 )

In Hospitals, PET scans are indispensable tools in oncology for detecting, staging, and

monitoring cancer. By pinpointing areas of heightened metabolic activity, PET scans aid

in precisely locating tumours, assessing their aggressiveness, and evaluating treatment

response. Moreover, PET imaging extends beyond oncology, playing a vital role in

diagnosing neurological disorders like Alzheimer’s disease, cardiovascular conditions

such as coronary artery disease and various other ailments. (Thientunyakit et al., 2022)

In the context of stroke diagnosis, the combination of PET and CT, known as PET-CT

imaging, provides a powerful tool for clinicians. CT scans are often performed first to

identify any acute bleeding or structural abnormalities in the brain, such as haemorrhages

or ischemic damage. Once acute bleeding has been ruled out, PET scans can be used to

assess tissue viability and metabolic activity in the affected areas.

PET-CT scans in stroke diagnosis allow clinicians to precisely delineate regions of

ischemic injury and differentiate viable tissue from irreversible damaged tissue. This

information is crucial for determining the potential benefits of reperfusion therapies, such

as thrombolytic therapy or mechanical thrombectomy, which aim to restore blood flow to

the affected areas and salvage viable brain tissue.

By integrating the anatomical and functional information provided by CT and PET scans,

respectively, clinicians can make more informed decisions regarding stroke management,

including the selection of appropriate treatment strategies and the prediction of patient

outcomes.

In summary, the synergy between PET and CT imaging technologies in stroke diagnosis

exemplifies the power of multimodal imaging approaches in modern medicine. By

combining anatomical and functional imaging data, PET-CT scans enable comprehensive

assessment of stroke pathology and facilitate personalized treatment planning, ultimate ly

improving patient outcomes and quality of care. PET scans represent pivotal advancement

in medical imaging technology, revolutionizing the diagnosis and treatment of numerous

diseases.

The integration of PET scanning into hospital workflows has significantly enhanced

patient care by allowing earlier and more accurate diagnoses, custom treatment plans, and

ongoing therapeutic monitoring. As technology continues to advance, PET imaging

continues to evolve, promising even greater insights into inner workings of the human

body and further transforming the landscape of modern medicine. (Minoshima et al.,

2016)

3.2.2 MRI in stroke diagnostics

Magnetic Resonance Imaging (MRI) has evolved significantly over the past decades,

transforming from a promising technique with limited applications to a primary

diagnostics tool for various medical conditions. Initially MRI was mainly used to examine

the brain and spinal cord, but it is now used to examine all parts of the body. As our

understanding of MRI has grown, we have gained a better appreciation of how to use it

effectively, either on its own or in combination with other techniques, to achieve the

highest level of diagnostic accuracy. (Katti et al., 2011)

MRI is non-invasive way to visualize the internal structure and certain functions of the

body. It uses non-ionizing electromagnetic radiation and has no known harmful effects.

The procedure involves placing the patient inside a strong magnetic field and using radio

frequency (RF) radiation to produce detailed cross-sectional images of the body. This is

done by aligning the nuclei of certain atoms, such as hydrogen with the magnetic field

and then applying RF signal, which causes energy to be released from the body. The

resulting energy is detected and used by computer to create an MRI image. (Katti et al.,

2011)

Functional magnetic resonance imaging (fMRI) is widely used, non-invasive method for

detecting changes in brain activity during preclinical and clinical studies. The fMRI

leverages the magnetic properties of oxygenated and deoxygenated haemoglobin to

measure blood flow changes linked to neuronal activity, known as the blood oxygen-

level-dependent (BOLD) response. During periods of neuronal activity, blood flow to the

region increases to provide oxygen to the active neurons through a process called

neurovascular coupling. fMRI allows researchers to identify and quantify brain activity

in response to a task or stimulus by using blood oxygen level dependant (BOLD) response

as indicator of neuronal activity. It is safe and non-invasive method, making it particularly

useful for studying neurovascular conditions like stroke. However, fMRI has some

limitations in preclinical research, including the need to use anaesthesia with animals,

which can affect the results of the study. (Crofts et al., 2020)

The earliest MRI sign of an acute cerebral infarction is the loss of signal drop-out from

arterial flow, which may create asymmetry in large infarcts. In contrast-enhanced

imaging, reduced arterial flow enhances affected arteries. Within hours, swelling causes

cortical thickening, sulcal narrowing, and increased T2 signal, with contrast scans

showing intravascular enhancement. Even after 24 hours, ~20% of infarcts remain

undetectable on T2-weighted images. MRI is less sensitive for acute infarcts, making CT

equally reliable for diagnosis, except in posterior fossa infarcts and suspected basilar

artery thrombosis, where MRI is superior (Duodecim, n.d.)

3.3 Emerging techniques for stroke diagnostics

3.3.1 Functional near-infrared spectroscopy (fNIRS) for stroke

detection

Functional near-infrared spectroscopy (fNIRS) is a non-invasive optical neuroimaging

technique that uses near-infrared light to measure cortical hemodynamic responses

associated with brain activity. (Myllylä et al 2020) Near-infrared (NIR) light is emitted

into the skull, where it undergoes random scattering and absorption processes within

tissues, enabling fNIRS to measure cortical hemodynamic responses. Some fraction of

photons creates banana shaped path back to the skull surface and finally there, it is

detected by sensitive photodetector as shown in figure 1 (OpenNIRS, 2022)

Figure 1. Banana Shape and principle of NIRS. Adapted from OpenNIRS.org

Most parts of tissue consist of water and are relatively transparent in NIR range to the

light. NIR light can penetrate the cranium and reach sufficient depth on brain tissue. This

optical range is approximately 650-950nm and in this range, tissue is relatively

transparent to the light. This is often referred as Optical Window (figure 2). Absorption

and scattering on tissue do remain fairly constant, deoxy-haemoglobin (HbR) and oxy-

haemoglobin (HbO) strongly absorb the NIR light while their concentration is changing

with blood flow and metabolism (OpenNIRS, n.d.).

Figure 2. Optical Window of NIRS. Source OpenNIRS.org

To accurately measure the concentrations of different chromophores, it is necessary to

select at least one wavelength on each side of the isosbestic point of interest. Isosbestic

point is specific wavelength at which the absorption coefficients of two of more molecules

that absorb light are equal. This means that at this specific wavelength, the absorption of

light by the molecules is the same, regardless of their concentration. This property can be

used in analytical chemistry and medical imaging to monitor changes in concentration of

molecules in a sample or tissue.

Isosbestic point for HbO and Hb is approximately at 800 nm and for water and HbO at

approximately at 950 nm. Absorbance peak for water is located approximately at 980 nm.

The device used for stroke data collection employs a frequency-domain technique to

modulate the intensity of illuminating light and measure both the phase and attenuation

of backscattered light. When selecting specific wavelengths for measurement, both

absorption and scattering properties must be considered, as scattering affects the depth of

light penetration.

For non-invasive measurements, it is essential that a sufficient number of photons scatter

back to detector. However, at a source-detector distance of 30 mm, the detected signal is

highly attenuated. Variability at larger source-detector separations may result from

individual differences in brain and skull structure. Measurements and simulations beyond

30 mm are increasingly affected by noise, making result interpretation more challenging.

Conversely, at distances shorter that 10 mm, the detected optical radiation does not reach

the cerebrospinal fluid (CSF) layer. (Myllylä et al., 2020).

NIRS technology has been increasingly used in studies of cerebrovascular incidents. One

of the most serious conditions in this category is intracranial hemorrhage (ICH), which

can result in high mortality if not diagnosed and treated quickly. The potential of NIRS

for early detection of ICH in patients presenting with headaches in the emergency

department has been investigated in a recent study.

This study included 378 patients, categorized into four groups based on their final

diagnosis: migraine, tension-cluster headache, ICH, and a control group. NIRS

measurements of cerebral regional oxygen saturation (rSO2) were taken from both the

right and left sides of the forehead at the initial medical contact. The study found that

patients with ICH had significantly lower rSO2 values on both sides of the brain

compared to those in other groups. Additionally, there was a significantly greater

difference between the right and left rSO2 values in patients with ICH. (Çınaroğlu et al.,

2024)

Gato et al. (2023) conducted a study to measure cerebral oxygen saturation and the

relative concentrations of oxygen and deoxyhemoglobin in the right and left frontal lobes.

They developed an automated classification framework to distinguish between patients

with ischemic stroke, hemorrhagic stroke, and those who are healthy. The framework

shows promising accuracy, which could potentially reduce the time needed for diagnosing

stroke. (Gato et al., 2023)

Li Y. et al. studied the use of NIRS technology to assess systemic low-frequenc y

oscillations (sLFOs) as potential biomarkers for vascular integrity in ischemic stroke

patients. NIRS was utilized to measure oxyhemoglobin fluctuations at various peripheral

sites, such as the toes, fingertips, and earlobes. The findings indicated that NIRS could

detect asymmetries in sLFO transmission caused by embolization, suggesting more

extensive circulatory dysfunction. The study emphasizes NIRS as a valuable tool for

monitoring vascular health in stroke patients, with the earlobes identified as particularly

informative sites for these measurements (Li et al., 2020).

In the context of neuroimaging techniques for diagnosing and managing ICH, several

novel approaches have emerged for rapid pre- and in-hospital diagnosis. NIRS is

highlighted alongside other advanced technologies like dual-energy CT, rapid MRI

techniques, and automated ICH detection methods. NIRS technology shows promise in

enhancing the speed and accuracy of ICH diagnosis, which is crucial for effective patient

management (Rindler, 2020).

NIRS technology has also been utilized for post-stroke depression (PSD), the most

common mental health condition affecting mood, following a stroke and a major factor

limiting patient rehabilitation and recovery. Koyanagi et al. investigated the application

of NIRS in diagnosing and evaluating PSD. Their study emphasizes that measuring oxy-

HB with NIRS is an effective method for diagnosing PSD in post-stroke patients

(Koyanagi et al., 2021).

NIRS has also showed promising preliminary results on detecting patients who had

neurovascular surgery, where NIRS detected reduced perfusion by observing decreased

oxygenation from patients who had ischemic stroke during the operation. Similar

monitoring has been used in aortic and cardiovascular surgery. (Guo et al., 2023), NIRS

is also an emerging bedside tool with advantages of being portable, easy and non-invasive

(Atais et al., 2024) for monitoring cerebrovascular physiology from the patients with

acute brain injury or from those who have high risk of developing cerebrovascular injury.

(Thomas, 2023)

NIRS technology has also been demonstrated in individuals with mild problems with

memory or thinking, who are at high risk of developing dementia. The study found that

groups with lower Mini-Mental State Examination (MMSE) scores exhibited a greater

increase in deoxygenated hemoglobin. In the stroke group, the increase in oxygenated

hemoglobin concentration was smaller compared to the control group, likely due to

reduced cerebrovascular reactivity. (Huang, 2023)

3.3.2 Electroencephalography for stroke detection

Electroencephalography (EEG) is a technique for observing brain electrical signals as

they appear on the scalp. An EEG, the recorded output, reflects alternating electrical

activity captured through scalp-placed electrodes and conductive agent. (Teplan, 2002).

(Olejniczak, 2006) defines EEG as “graphical representation of the difference in voltage

between two different cerebral locations plotted over time”. The EEG when measured

with using depth probes is called Electrogram and when measured from directly from the

cortical surface, it is called Electrocorticogram. I will refer EEG to only measured from

the head surface in this paper.

EEG measurement is a non-invasive procedure that can be safely used on adults, children,

and patient’s multiple times with minimal risk or limitations.

When brain cells are activated by different processes, they create local current flows. EEG

measures most of the cases current flows during synaptic excitations in the cerebral

cortex. Electrical potential differences arise in pyramidal cells, forming electrical dipoles

between the neuron’s body and neural branch. Brain electrical currents are mainly

produced by Cl-, K+, Na+ and Ca++ ions moving in a direction set by the membrane

potential. This is rough description, and more detailed microscopic picture is lot more

sophisticated. (Teplan, 2002).

It takes only large population of neurons to generate enough recordable electrical activity

on surface of the head. Weak signals detected by electrodes on the scalp are amplified

massively and then stored to computer memory or hard drive or printed on paper. (Teplan,

2002).

Encephalographic measurement recording system consists of electrodes, amplifiers with

filters, recording device and A/D converter (Analog/Digital converter). Electrodes are

reading signals from the surface of the head, converters change signal from Analog to

Digital, amplifiers will bring microvolt signals in to range where they can be accurately

digitalized and computer or other device stores and displays the gathered data. Minimal

configuration for single channel EEG consists only one electrode and multichannel can

go up to 128 or 256 electrodes. There are many types of electrodes with different

characteristics, which can be listed as disposable electrodes, reusable disc electrodes,

needle electrodes, saline-based electrodes and headbands and electrode caps. For

multichannel measurements, electrode caps are preferred. (Teplan, 2002).

Skin surface is generally cleaned from sebum (oil) and scrubbing dried parts away is

recommended. With disc and disposable electrodes, abrasive paste is used. When used

needle connections, right hygiene and safety protocols are important. EEG measured by

mounted cap for same electrode points will have threat of pain and bleeding for a subject,

because needles can penetrate the skin. (Teplan, 2002).

Brain waves have been categorized into four mostly used frequency bands. These are

Delta (0.5-4Hz), Theta (4-8Hz), Alpha (8-13Hz) and Beta (>13Hz). (Teplan, 2002). There

are also Gamma band which is at 30 and up (HZ. Files, 2011). There is also inconsiste nc y

between different frequency bands between Teplan and HZ Files. Delta frequency starts

at 1Hz and Alpha ends at 12 HZ by Hz.Files.

While EEG is not primary diagnostic tool like CT or MRI, it offers valuable

supplementary information in specific clinical scenarios, particularly in assessing brain

activity changes, monitoring recovery, and detecting post-stroke complications.

3.3.3 Accelerometer for stroke detection?

Primary objective of Wilkinson’s study was to investigate the effectiveness of a low-

priced, transportable EEG system among individuals who recently suffered stroke in

differentiating between patients who had an ischemic stroke and a control group

compromising individuals who displayed suspected stroke symptoms, the stroke mimics.

Using the Muse device, participant movement was analysed by collecting data from the

gyroscope and accelerometer. The research determined that individuals with stroke

displayed greater accelerometer variability in the X-axis and lesser variability in the Z-

axis, suggesting more tilting head movements and less shaking head movements. These

differences may reflect impairments in the motor system or attentional deficits due to

contralateral neglect. It is worth noting that since the data was collected with eyes closed,

impaired motor function, including postural support, may have contributed to the

observed differences. The absence of significant differences in the gyroscope’s rotational

data highlights the accelerometer’s potential as a more sensitive instrument for detecting

movement disparities between controls and stroke patients. This finding suggests that

overall movement patterns may provide greater diagnostic value than rotational metrics.

(Wilkinson et al. 2020).

4. Aim of the work and research questions

This chapter’s main focus is on describing the goals of the Stroke Project, the research

questions and the methods used.

4.1 Research context

Stroke data project was a two-year Business Finland project which goal is to develop

solutions for early Cerebrovascular accident recognition, prevention and rehabilitatio n.

Solutions are developed for patients and healthcare professionals. Especially goal is to

develop data-based Decision support solutions for physicians and other healthcare

professionals and solutions based on Artificial Intelligence (AI) and machine learning

(ML) methods to support cerebrovascular accident diagnostics, also for example video

and speech analysis techniques. Also, project aims to develop new sensors for measuring

patients with near infrared spectroscopy and low frequency brainwave measuring devices.

Project aims to develop automatic evaluation system which uses audio visual and bio

signals to support stroke diagnosis. Project also is going to develop iteratively business

model ecosystem for helping developed solutions getting on domestic and foreign

markets.

4.2 Research questions

The following are the research questions (RQ) for this thesis work.

RQ 1: How feasible is a multimodal brain monitoring setup for stroke data collection in

terms of ease of use and measurement duration?

RQ 2: What are the possibilities to develop wearable NIRS brain monitoring technology

that could enable easily and immediately to start brain monitoring of patient with possible

stroke?

My work on the project began with learning to use various devices for data collection.

Next, we tested the measurement protocols to evaluate their feasibility. It quickly became

evident that the planned protocol was not suitable for actual patients, as stroke conditions

vary significantly among individuals. Consequently, we made adjustments to the protocol

and identified a suitable location within the hospital for conducting all measurements.

One of my tasks during the project included training research nurses at Kuopio University

Hospital. I travelled there twice to conduct training sessions and a few additional times to

deliver spare parts and participate in measurements. This allowed me to assist with data

collection and monitor how the measurements were conducted.

During the development of NIRS device, I did participate on several different task, for

example building a new electronic board for NIRS device, repairing device when needed,

developing, designing and testing new headband and NIRS device, software development

and designing GUI and also doing user testing on the NIRS software.

4.3 Research method

In my work, I have chosen the Design Science Research method. In Alan R Hevner’s A

Three Cycle View of Design Science Research presents a structured approach to

understanding Design Science Research (DSR) by introducing three interconnected

cycles, the Relevance Cycle, the Rigor Cycle, and the Design Cycle. The Relevance cycle

connects with real-world problems by gathering requirements from the application

domain, which consists of people, organizations, and technical systems. It ensures that

research remains applicable by introducing the developed artifacts into practical field

testing, allowing for validation and iterative improvement based on real-world feedback.

The Rigor cycle provides the theoretical foundation for research, drawing from scientific

theories, established methodologies and domain expertise to ensure that design efforts

grounded in prior knowledge. This cycle also contributes back to the knowledge base by

expanding existing theories, refining methodologies, and documenting research insights.

At the core of the process, the Design Cycle operates as an iterative loop where artifacts

are created, evaluated and refined. This cycle is essential for ensuring that the designed

solutions are both effective and scientifically rigorous.

Hevner emphasizes that these three cycles must be present in any high-quality SR project,

as they ensure a balance between practical relevance and scientific rigor. Unlike

traditional natural sciences, which focus on understanding phenomena, Design Science is

inherently problem solving oriented, aiming to create artifacts, such as models, systems,

or methodologies, that address real-world challenges. The research methodology

integrates systematic design, empirical evaluation, and theoretical contributions, making

it an essential paradigm in fields like Information Systems, Engineering and Architecture.

By recognizing DSR as a pragmatic science, its dual contribution is highlighted as

improving real-world systems while simultaneously advancing theoretical understanding

(Hevner, 2007).

In my work, the Relevance Cycle involved data collection from patients using one version

of an fNIRS device. We identified issues with conducting measurements and recognized

development goals that needed improvement. The Design Cycle involved developing

iterative prototypes, such as a new headband, based on insights from the Rigor Cycle.

Ultimately, this process led us to develop a smaller, digital wearable device that caused

less discomfort for test subjects.

4.4 Development goal

The development goals focus on advancing the NIRS device into a wearable form while

enhancing the head unit connection for improved usability. The objective is to facilitate

quicker and easier attachment to the forehead while ensuring accurate measurement

distances. Additionally, the device and its user interface is designed to be more user-

friendly, accommodating a diverse range of user groups. Another key goal in headband

and device development is the integration of additional EEG, ECG and accelerometer

measurement capabilities to make NIRS device wearable multimodal measureme nt

device.

5. Design challenges in realizing easy-to-use

software user interfaces

This chapter will go through some theoretical background on designing User Interfaces

and introduce a few methods which can be used to design easy-to-use interfaces for

different user groups typically with limited experience of using software interfaces for

medical monitoring purposes.

Nurses represent the primary user group and encompass a highly diverse population. This

diversity is reflected in significant age differences and varying levels of computer

proficiency. Younger nurses generally exhibit greater familiarity with computer

applications than their older counterparts. However, many nurses’ express apprehension

about adopting new applications or using technical devices. Among the numerous

challenges, this thesis focuses solely on developing intuitive and user-friendly

applications to address specific concerns and facilitate adoption.

5.1 Generation-Oriented approach vs Aging-oriented framework

According to Liu.S, Generation-oriented approach and the aging-oriented approach vary

in the following ways.

Overview of variations between age groups. Elderly people typically have more

challenges on using today’s interfaces of present-day technology compared to younger

people. The Generation-oriented approach examined the relationship between the

generation-based information and experience of elderly people and the demands of the

current user interface to understand their difficulties with the latter. The Aging oriented

approach is focusing on genetic and biological features and their gradual loss because of

the process of aging.

Figure 3. Generation oriented and Aging oriented approach Framework. Adapted from

Liu.S. 2012.

Declining and maintaining. While aging-oriented approach focus on the gradual loss in

abilities that come with age and suggest design approaches to compensate for those

deficits, the generation-oriented approached focuses on the abilities and knowledge that

remain and provides design recommendations to make use of them.

Historical context. The aging-oriented approach tends to be isolated from historical

context and external influences, whereas the generation-oriented approach recognizes the

importance of context and considers each generation to be shaped by its historical

circumstances. This makes the generation-oriented approach dynamic, as it requires

ongoing updates to our understanding of the knowledge and experience of older adults as

new generations enter that age range.

To summarize, the aging-oriented design approach has had moderate success in crafting

user interfaces for elderly users, but it faces certain shortcomings. On the other hand, the

generation-oriented approach holds promise for refining and advancing the aging-

oriented technique. Therefore, a framework that integrates these both approaches in a

systematic way is highly desirable for achieving the conclusive design solutions. (Liu,

2012)

If the target user’s capabilities are not properly understood in interface design, it can lead

to frustration, accidents, or even exclusion. When the requirements of specific interface

features surpass user’s abilities, design exclusion occurs, as the capability-demand

relationship suggests. For developing an interface that is easy to operate and approachable

for older people, it’s crucial to consider all the main characteristics and capabilities of this

user group in the design process.

Designing user interface for older people is commercially popular so there is gap in

literature for designing and evaluating design concepts. To address the gap in current

research, emphasize the need for a deeper understanding of what it means to be an older

adult, and offer a more nuanced approach for developing interfaces tailored to older

people. In Liu’s article, they are proposing a framework that combines the generation-

oriented and aging-oriented approach.

The suggested framework in the article, could act as a theoretical framework that

illustrates the design process and reasoning pathway. To Begin the process of designing

the user interface, it is important to thoroughly understand the characteristics of the target

user group, which in this case are older adults. The designers will follow two paths in the

framework diagram: the aging-oriented and the generation-oriented approaches.

To gain extensive understanding of the characteristics of older adult user, the designers

should follow both paths outlined in the framework. This involves applying both the

aging-oriented and the generation-oriented approach. After completing this process, early

design solutions based by each approach can be generated. It’s worth noting that these

preliminary design solutions may differ from one another. To arrive at a final design

solution, a dialogue comparing the two preliminary products is necessary. This design

solution informed by the comparison dialogue will be created at the end of the process.

(Liu, 2012)

5.2 Generation-oriented approach

On the opposite side of the diagram of figure 13, the focus is on developing design

solutions through a generation-oriented approach. This strategy centres on gaining insight

into older adults by exploring the unique experiences and knowledge shaped by their

generational background.

The first in this generation-oriented approach framework is innate knowledge, followed

by sensorimotor knowledge, then knowledge specific to an individual’s culture and

finally expertise. Additionally, there is a level of tool knowledge that encompasses the

sensorimotor, cultural and expertise levels of knowledge (Liu.S 2012).

Within the suggested framework, knowledge is divided into three categories, the

sensorimotor, cultural, and tools level. The sensorimotor level includes concepts such as

population stereotypes, image schemas and affordance. Knowledge can differ

significantly between cultures, including differences in population stereotypes, colour,

values, and language. At the tool level, there is knowledge about specific product features

such as operation, appearance, location and but also mental models and symbols (Liu.S

2012).

In the proposed framework, experience is divided into three categories. Expertise with a

wide variety of products spanning multiple types and domains, familiarity with other

products of the same type or domain, and specific experience with the product being

evaluated. All three categories of experience are included in the framework.

In the proposed framework, historical context is considered an important factor in the

generation-oriented approach. This is because each generation is shaped by the unique

historical context in which they grow up and live, which includes social, cultural,

environmental and technological characteristics. Studying how user interfaces have

developed over time, as well as the objects and interactions that senior adults have

experienced during key developmental periods, is important for understanding their

behaviours and preferences. In particular, the formative period (ages 10-25) is a critical

time for the acquisition of norms, values, and skills that shape future of behaviours and

preferences (Liu.S 2012).

5.3 User interface for individuals with limited tech familiarity

When designing GUIs for people with limited tech familiarity, it is important to consider

their specific needs and abilities. User may have visual, cognitive, or even physical

limitations that can affect their ability to use the interface. For example, they may have

difficulty reading small text or distinguishing between similar colours. They may also

have slower processing speed or memory deficits that can impact their ability to learn and

use the interface (Liu, 2012).

To address these challenges, it is important to consider the principles of universal design,

which aim to make products and environments accessible to all users, including those

with disabilities. Some strategies for designing GUIs for users with limited tech

familiarity, include using larger text and high contrast colours, providing clear and simple

instructions, and allowing for customization of the interface to meet the needs of the

individual user.

Designers can use simulation kits, helping designers gain insight into varying levels of

physical and sensory functions. On the other hand, currently there are lack of proper tools

for simulating elderly users’ prior knowledge and experience (Liu, 2012).

5.4 Dialogue and design solution

The final product of this framework will be successful if we have a dialogue comparing

the solutions produced using the aging-oriented and the generation-oriented approach. To

reach a final design solution using the framework, it is important to encourage and

conduct a dialogue that includes careful analysis, evaluation and discussion of the specific

goals and needs, as well as the specific design challenges and situations (Liu, 2012).

5.5 Persuasive system designing on software development

According to Oinas-Kukkonen, Interactive information technology that seeks to change

people’s behaviour or attitudes is called persuasive technology. Historically, persuasion

has referred to communication designed to impact other’s actions and decisions. The

Internet, smartphones, and related technologies open possibilities for persuasive

communication through their ability to reach users effortlessly. In addition, the Web-

based systems are particularly suitable for persuasive interaction, since they consolidate

the benefits of personal and broad audience communication. Persuasive technology may

prove especially valuable in fields like healthcare software applications that motivate

people to adopt healthy habits, potentially preventing or delaying medical issues and

easing the burden on public healthcare systems. (Oinas-Kukkonen & Harjumaa 2009)

As in Oinas-Kukkonen article persuasive systems are defined as “Computerized software

or information systems designed to reinforce, change or shape attitudes or behaviours or

both without using coercion or deception”. This definition comes with three potential

effects for persuasive system. The shaping or change of attitudes, the voluntary

reinforcement and/or behaviours. A reinforcing effect refers to the strengthening of

existing attitudes or behaviours, making them harder to change. A changing effect refers

to the modification of a person’s response to a particular issue. A shaping effect refers to

the creation of a new pattern for situation that did not previously exists. In many cases,

communication that results in a shaping effect may be more effective that communication

that targets for a changing effect. (Oinas-Kukkonen, 2009)

Oinas-Kukkonen introduces in their article for the Persuasive System Design (PSD)

model with seven postulates behind Persuasive System, which is based on other research,

conceptual analysis, and their empirical work. (Oinas-Kukkonen, 2009)

First postulate is that “Information technology is never neutral”, but rather constantly

influencing people’s attitudes and behaviours. This is like how teachers persuade students

in school, and it is not inherently negative. Persuasion can be viewed as a process rather

than single event. Persuading a user is a multifaceted and complex task that may change

over time based on factors such as the user’s goals. (Oinas-Kukkonen, 2009)

The second postulate maintains that “people like their views about the world to be

organized and consistent.” This is bases in theories of commitment and cognitive

consistency. Systems that foster commitment can enhance persuasive outcomes. The idea

of commitment propose that persuasive systems could offer the option for individuals to

make private or public commitments to engage in the desired behaviour. One way to

implement this is by providing a convenient way to send a text message or email to one’s

family, friends, or colleagues. (Oinas-Kukkonen, 2009)

The third postulate states that “direct and indirect routes are key persuasion strategies.”

Someone who thoughtfully analyses the persuasive communication is likely to be

persuaded through the straightforward path, in contrast, a person demonstrating

diminished evaluative ability and uses simple cues or stereotypes may be swayed by the

indirect route. Both direct and indirect processes may operate at the same time and be

supported by various features of the software system. Direct persuasion has been more

effective in the long run, but in today’s age of information overload, people are often

forced to rely on indirect cues more frequently due to the overwhelming amount of

information they must process. (Oinas-Kukkonen, 2009)

The fourth postulate states that “Persuasion is often incremental”. It is more effective to

gradually introduce new actions to someone through a series of progressive

recommendations, rather than trying to persuade them to adopt a new behaviour all at

once. A persuasive system should therefore aim to guide people towards a desired

behaviour by encouraging them to take small, incremental steps. For example, a healthy

eating app might first encourage users to include some vegetables in their meals, and then

later suggests filling half of their plate with vegetables. It is important to consider ethical

implications of incremental persuasion and ensure that the ultimate goal is transparent

and clear at every step. (Oinas-Kukkonen, 2009)

The fifth postulate states “Persuasion through persuasive systems should always be

open”. It is important to openly acknowledge the designer’s bias in a persuasive system.

For example, simulations can be powerful tools for persuasion, but if the designer’s bias

is not clearly stated, they may lose credibility or even mislead users. Additionally, it is

essential to make sure that the content of the persuasive system is based on accurate and

truthful information, as the overall goal is to encourage voluntary changes in attitudes or

behaviors. (Oinas-Kukkonen, 2009)

The sixth postulate is “Persuasive systems should aim at unobtrusiveness”. In order to

meet user expectations, it is important that persuasive systems do not disrupt or divert

users attention while they are engaged in their primary tasks with the system. The

fundamental idea of unobtrusiveness also requires careful consideration of when it is

appropriate to present persuasive features. Using persuasive features at inappropriate

times, such as suggesting exercise to someone who is sick or reminding someone to take

pharmacy products for high blood pressure during a demonstration at the lecture, can lead

to negative outcomes. (Oinas-Kukkonen, 2009)

The seventh postulate states that “Persuasive systems should aim at being both useful and

easy to use”. In example, at really helping the user’s needs. To be effective, a persuasive

system must process a range of characteristics that contribute to a positive user

experience. These include ease of access, responsiveness, error free operation,

convenience, and high-quality information, along with attractiveness and user loyalty. It

is important to note that these are general qualities that all software should strive for and

are not unique to persuasive systems. If a system is difficult to use or provide not value

to user, it is unlikely to be persuasive. (Oinas-Kukkonen, 2009)

5.6 EU regulations on software in medical devices

When developing software for medical device in EU, manufacturer must consider several

requirements from the standards and from Medical Device Regulation (MDR). Main

standards for medical device software are IEC 62304, which is for medical device

software and software’s life cycle process and IEC 82304-1, which is mainly for Health

software and general requirements for product safety. Also, there are other standards like

as ISO 13485 which is standard for Quality Management Systems (QMS) and ISO 14971

standard for risk management. These standards need to be considered also, in a software

development process. Even though standards have been written much of a waterfall model

in mind, they still do allow agile methodologies and development sprints, but require

approvals, documentation and reviews. These standards are not usually freely distributed,

and developers and medical device manufacturers are required to buy them.

6. Measurement methods

This section describes the various technologies used to collect data for the stroke project.

6.1 Fundus photography

The eye can be said to be structural extension of the brain and spinal cord. In addition, it

has structural properties that are characteristics of the central nervous system. The retina

or fundus is anatomically and developmentally an extension of the central nervous

system. It is made of retinal ganglion cells whose axons form the optic nerve. The optic

nerve fibres in turn, are in fact axons of the central nervous system. So, it is not wrong to

say as London et al says, “eyes are the mirror of soul- or if not a mirror, then at least

window to the brain”. Several well-recognized neurodegenerative conditions that affect

spinal cord and brain are also noticeable in the eye, especially in the fundus. In fact, ocular

symptoms are in some cases the first signs of central nervous system disorders and can

therefore be used as a prophylactic aid in early-stage diagnosis (London, Benhar &

Schwartz, 2013)

Figure 4. Optomed Aurora Camera.

Photo from Personal picture gallery

Figure 5. Research nurse using Camera.

Photo by Riitta Laitela

In cerebrovascular disorders, changes occur in the small blood vessels of the retina that

are imaginable. However, it is typical that the changes are not seen at an early stage, but

only when the blood-brain barrier has been damaged and the brain suffered permanent

damage, for example because of oedema (Cheung et al. 2013). It is therefore reasonable

to assume that early detection of small vascular changes could justify initiating

prophylactic treatment for Cerebrovascular disorder.

Tools based on AI are already being used in various areas of healthcare. In the context of

Retinal imaging, Artificial intelligence-based algorithms are used to identify changes

associated with, for example diabetic retinopathy. However, it is known that algorithms

can detect much more of the retina than is detectable to the human eye. For example,

Poplin and colleagues developed algorithm that identified several factors influencing risk

of cardiovascular disease from digital retinal photography, like age, sex, and diastolic and

systolic blood pressure (Poplin et al, 2018)

In a recent review article, Esteva and colleagues speculated that retinal will be

increasingly provide information on non-ocular diseases, but in particular cardiovascular

and central nervous system diseases or systemic diseases (Esteva et al, 2021)

In this Stroke Project, the retina of cerebrovascular disorder patients, in other words

fundus is imaged with Optomed Aurora digital fundus camera, manufacture by Optomed.

The images show changes in retinal vasculature as a function of time, as well as possible

other changes. The aim of the observed changes is to develop an algorithm that utilizes

artificial intelligence, which could be used on the one hand to monitor recovery but

possibly also to detect early changes that anticipate cerebrovascular disorder.

6.2 EEG and ECG

EEG provides information on the electrical activity of nerve cells (neurons) in the cortex

on brain. The electrical activity of the brain manifests itself as variations in the voltage

and frequency of the waves. Rhythmic voltage fluctuations are thought to be due to the

rhythmic oscillation of neural networks. The measured signals are in the microvolt range

and frequencies are usually between the 0.5 and 100 Hz. The brains electrical activity is

broadly divided into four main frequency bands, Delta (0-3 Hz), Theta (4-7 Hz), Alpha

(8-13 Hz) and Beta waves (14-30 Hz). Each frequency range has its own characteristics

and their changes in cerebrovascular disorders are studied.

In general, with EEG measurements, slow frequencies <0.5 Hz are filtered out because

they are thought to be caused something non-neural activity. In Stroke project in

particular, low frequency direct current DC-EEG voltage fluctuations <0.5 Hz are also

interest because they reflect the strongest mechanism of cerebral glia circulation that

deliver changes in permeability across the blood-brain barrier affecting the previously

mentioned faster neural rhythm (Kiviniemi et al. 2017). In cerebrovascular disorder, these

low frequency functions can be radically altered when vascular function fails. The

measurements are done with Bittium’s commercial NeurOne DC-EEG measuring device,

which can also be used to measure Electrocardiography (ECG) signal which is reflecting

electrical activity of the heart.

6.3 NIRS

The NIRS device was developed at the University of Oulu and it enables continuo us

measurement of cerebral blood flow, in particular synchronous measurement of the

concentrations of deoxy haemoglobin (Hb) and oxygen haemoglobin (HbO),

cerebrospinal fluid (CSF) concentrations (Myllylä et al., 2018).

Figure 6. NIRS setup testing. Photo

Personal picture gallery

Figure 7. Bittium NeurOne. Photo

Personal picture gallery

6.4 Speech and video recording

Research aim is to find early anomaly symptom patterns and incidents, for example

aphasia, face nerve paralysis and motorized problems from video image of the patient and

recorded speech. Algorithm based on video image monitors facial muscle asymmetry and

changes in blood flow by distinguishing remote photoplethysmography (rPPG) signals

from several areas from face.

There is little literature material about detecting stroke from video image and most of

material is concentrated on detecting muscle paralysis from video image, like for example

(Thevenot 2017). Paralysis symptoms can be caused by several different factors like for

example, infection, trauma, tumours, stroke and different syndromes like Guillan-Barre

and Moebius. Differential diagnosis between these is not easy or obvious. Different

methods have been developed to identify asymmetries between the sides of the face in

facial stroke. In these tests, the test subject is usually asked to perform different

expressions like smiling. Image processing can also be used to measure the distances of

anatomical points, facial features and to compare facial aspects. In addition, facial

movement can be measured to assess facial muscle activation.

Video based studies have used House-Brackman method as a reference, which allows

physicians to assess the severity of stroke at six different levels (I-VI) from normal

symmetrical movement (Level I) to complete paralysis (Level VI). House-Brackman is

the most used point scale for measuring facial nerve function (Niziol et al., 2015).

However, the method has been criticized for not being able to accurately measure facial

mobility. The method is used to evaluate the face generally as whole and not the different

parts of the face like example forehead, nose, eyes and mouth separately. In addition,

there is little consensus among different physicians. Banks et al. Have developed the

“Electronic Facial Paralysis Assessment” tool, which has been validated using several

different measurement scales, but can only be used to detect facial paralysis. (Banks et

al., 2015).

Recently has been published a study that used video first time to assess stroke. The video

was recorded in the hospital emergency room. The study achieved 93% sensitivity for

classification as stroke / non-stroke and 79% accuracy from the video which was less than

four minutes. The method used in the study, utilized a combination of visual and audio

data (Yu et al., 2020).

Stroke project aims at better accuracy compared to Yu et al’s research, by using among

other things like more advanced methods of artificial intelligence and machine learning,

as well as a new kind of data combination. Unlike previous studies, in addition to video

and audio signals, Stroke project utilizes photoplethysmography signals and examines the

added value of pooled data in classifying stroke patients.

Figure 8. Video and speech measurement.

Photo by Riitta Laitela

6.5 Accelerometer

According to recommendations of Toimia database (Terveysportti 2022) the balance of a

stroke patient can be assessed by the following indicators, ABC survey (The Activities-

Specific Balance Confidence Scale) (Powell, Myers 1995), posture swing on power plate,

the Berg’s balance test (Berg et al. 1989), and the Timed up and go test (TUG) (Podsiadlo

& Richardson 1991). The ABC query is based on only questions, which means it does not

require special facilities or equipment. Measuring postural oscillation on power plate, on

the other hand requires a separate power plate, computer, and software to perform the test

and it is usually done by a physiotherapist. The Berg Balance test as well as the TUG tests

also require professional to do the tests and the Berg test takes about 15 minutes and the

TUG a little less than that.

VTT has developed an Accelerometer based method for assessing the risk of falling based

on a person’s walking style. In developing this method VTT has used Berg’s balance test

as reference meter (Similä et al. 2017). However, this method has not yet been tested with

stroke patients. Acceleration based balance assessment provides an opportunity to speed

up balance assessment at the outpatient clinic and on the other hand, it also provides a

cost-effective opportunity to monitor rehabilitation of a stroke patient, especially with

regard to walking rehabilitation.

Stroke project investigates the applicability of accelerometer-based balance assessment

in the balance of stroke patients based on their walking style traits. Project utilizes on

their study an application which is previously developed by VTT and uses it to collect

acceleration sensor data from stroke patients about their walking style. Stroke project uses

on its research clinical measures as reference measures. The result of this method is

compared to for example, with the results of the ABC survey (Powell & Myers, 1995)

and the results of mRS meter (Van Swieten et al., 1988). Stroke project is going to refine

the index composed of features suggesting imbalances which is more suitable for patients

with cerebrovascular disorders.

Figure 9. MoveSense Balance Sensor. Photo

from personal picture gallery

7. Materials and measurement protocols

In this part is described the data collection protocols and measurement methods that were

used.

7.1 Data collection and research methods

Data collection is carried out as a national multicentre study in two university hospitals

in Finland. At Oulu university hospital (Oulun Yliopistollinen Sairaala , OYS) and

Kuopio’s university hospital ( Kuopion Yliopistollinen Sairaala, KYS ). The research is

based on direct observations in which the aim is to monitor the conditions of subjects and

the effects of changes in the subjects. The Stroke Project is not an intervention study that

actively seeks to change the conditions of the subjects, but an hybrid study. Main research

methods are quantitative and qualitative. In my thesis from Stroke project data, only

quantitative data collection from bioelectrical sensors method is used. This is prospective

patient series study that aims to gather data for the development of new technologies and

analytical methods, as well as to determine the accuracy of the methods compared to

existing clinical methods. In addition, control data is collected as a reference point, which

can be examined by separate invitation example at OYS’s premises. The control data and

the data collected for differential diagnosis cover three different groups, which can be

examined by separate invitation, example at OYS’s premises. Subject groups are listed in

Table 1.

Table 1. List of test subjects

Subject group

Planned amount of test subjects

Healthy controls

Vascular dementia

Migraine

Stroke

100

The data collection is preceded by a month-long testing of the measurement protocol, in

which the functionality of the protocol is verified in practice and modified if necessary.

At this stage, the research nurses will be trained by researchers to make the measurement

protocol and the staff of the department involved in the data collection will be informed

about the forthcoming research and its purpose. Patients are recruited for 6 months in the

hospital’s cerebrovascular disorder department and the follow-up period is three months.

The data collection protocol is evaluated and if necessary, updated after three months of

recruitment to ensure sufficient and correct data.

The purpose of the study is to extract from the hospital systems the information that

cerebrovascular disorder patients currently that is collected during their treatment, as well

as to collect data with the surveys, sensors and devices which are developed in the project.

The collection of data has no effect on the treatment received by the patient or on the

treatment decisions made for him or her. The study nurse or research assistant will take

the following measurements on the patient.

 Video and speech recording by mobile application which has been developed for

this purpose

 Balance measurement with a wearable motion sensor and mobile applicatio n

which has been developed for this purpose.

 Measurement of the electroencephalogram (DC-EEG) and optical brain imaging

with Near Infrared Spectroscopy (NIRS) with head mounted measuring device

 Cardiac electrocardiogram (ECG) measurement with a chest electrode.

 Fundus photography with Optomed Aurora camera.

 Questionaries on background factors, cerebrovascular disorder risk factors,

assurance of functional balance and patient experience related to their health

(Patient Reported Outcome Measure, PROM)

 Patients at risk for cerebral infarction (ABCD2) and severity of injury from

cerebrovascular disorder (Modified Ranking Scale, mRS)

Patients medical condition will be considered, and measurements shall be performed only

if the patient’s condition permits. The opinion of the Ethics Committee of the Northern

Ostrobothnia Hospital District is sought for the study, and it is submitted to the Ethics

committee of the Northern Savonia Hospital District for information. In addition, the

equipment used in the study, will be reported to Fimea. Following the favourable opinion

of the Ethics committee, registry research permits are also applied for from OYS and

KYS for the collection for registry material.

7.2 Research subjects in Stroke project

7.2.1 Inclusion criteria for research subjects

Participants were included in the study if they were over 18 years old and had a suspicion

of transient ischemic attack (TIA) or stroke, classified under ICD-10 codes I60-69 and

G45, as well as ICD-8 and ICD-9 codes 430-438. In addition to the main study group, the

University of Oulu and Oulu University Hospital (OYS) collected data for both a control

group and differential diagnostics group. Control group participants were required to be

healthy individuals over 18 years old without any chronic diseases or regular medication.

The differential diagnostics group included subjects with either severe migraine or

vascular dementia. Migraine patients had to be over 18 years old and experience at least

15 migraine days per month, while dementia patients had to be over 18 years old with a

diagnosis of vascular dementia.

7.2.2 Exclusion criteria

Individuals in the exclusion criteria group included prisoners, forensic psychiatric

patients, minors, individuals with disabilities, and pregnant or breastfeeding women, in

accordance with the Medical Research Act (488/1999). Additionally, healthy control

participants were required to have no regular medication, smoking habits, or other

addictions, however, the use of blood pressure and cholesterol medication was permitted.

7.2.3 Consent of subjects

The written, informed voluntary consent of the subjects is requested at the time of

recruitment. Subjects shall be provided with a written explanation of their rights, the

purpose, nature and methods the study, as well as any risks and inconveniences (Medical

research Act 488/1999). Participants have the right at any stage of the study to suspend

their participation without giving any reason affecting their treatment or services. Consent

forms will be signed in duplicate, one of which will remain with the patient and the other

will be kept by the study nurse in a prearranged locker on hospital premises.

The ability of memory-impaired patients to be recruited for differential diagnosis to give

informed consent, may be impaired. If the patient is under guardianship, consent to

participate in the study will be asked from the guardian. If the patient is not subject to

guardianship, the patient’s physician will assess the patient’s ability to consent to

participate in the study. If the treating physician considers that the patient’s conscious

ability to consent is impaired, consent will be asked from the relative. Relatives are asked

to assess whether the patient would have given their consent before the memory illness

and to decide on behalf of the patient. If the treating physician considers that the patient’s

ability to give informed consent has not been impaired, consent will be asked from the

patient himself.

7.3 Sampling

KYS’s cerebrovascular disorder ward has an estimated 500-600 patients annually and

their average length of hospital stay is 4 days. Correspondingly the cerebrovascular

disorder ward at OYS has 600-700 annually and average length of hospital stay is 3.5

days. Patients are actively treated in both hospitals, as well as patients referred for

rehabilitation, will come for a follow-up visit after 3 months. Follow-up checking can

also be done by phone call. The number of these patients in both hospitals is estimated at

150-200 per year.

The sample consists of all patients who meet the admission criteria coming to

cerebrovascular ward during the recruitment period, so that it is estimated that the

research nurses will be able to recruit and perform the planned measurements for one

patient per day within working hours while not disturbing the daily treatments of the

patients. During the estimated 6-month recruitment period, it is possible for patients with

TIA or Stroke suspicion to reach N=50-100 per hospital, for a total N= 100-200 patients.

The number of patients that can be reached will be specified during protocol testing. In

addition, the University of Oulu and OYS measure control and differential diagnosis data,

N=100 -125, which are called on by email-lists, like for example Oulu.fi and ppshp.fi and

based on research invitations to local newspapers.

7.4 Recruitment

When a patient who meets the admission criteria arrives at the cerebrovascular disorder

ward, the ward physician will present the research in the context of the doctor’s round

and provide the patient with a paper version of the subject’s bulletin and consent form.

The study nurse will later inquire about the patient’s willingness to participate in the

research. If recruitment is not done at doctor’s round, the study nurse will go recruit new

patients from the ward by him or herself and ask from ward’s nurses for suitable patients

and check that they meet the recruitment criteria. Then research nurse will go present the

research for patient and if patient is willing to participate for the research, he/she gives

the subject’s bulletin for the patients and consent form. The patient is asked for voluntary,

individualized, informed, and unambiguous written consent prior to the measurements.

During the recruitment process, all patients invited to the study and reasons for their

refusals are recorded to assess the generalizability of the results.

In the control group, healthy controls are collected either from patients’ relatives of from

mailing lists using attached separate invitation. Patients with severe migraine are recruited

from the neurology department for those patients whose diagnosis has not been fully

confirmed and or whose symptoms have required departmental care. In addition, migraine

patients can be recruited from the emergency department if necessary. Patients with

vascular dementia are recruited from the Department of Neurology Memory Clinic

7.5 Data collection protocols

If the patient agrees to the study, the study nurse will perform the following measureme nts

on the patient once while they are in a Stroke ward as soon as possible after they have

entered the ward. The study nurse with the help of the ward staff, will evaluate when the

patient has appropriate time to participate in the measurements, in which case he or she

will not have any procedures or examinations related to normal care. Measurement

protocols are described in following sub chapters.

1. Balance measurement using a wearable motion sensor at the waist and mobile

application developed by VTT for this purpose. The measurement is performed

on those patients who can walk independently for distance of at least 10 meters.

2. Video and speech recording with a mobile application developed by VTT for this

purpose.

3. Combined measurement in which DC-EEG, NIRS, ECG and Accelerometer

signals are measured in addition to video and speech.

4. Fundus photography by using Optomed Aurora camera.

5. Questionnaires either as a paper form filled in by the patient him or herself or by

interviewer.

6. Meters made by a study nurse.

a. TIA patients stroke risk (ABCD2)

b. Severity of disability caused by stroke with Modified Ranking Scale

(mRS)

7.6 Balance protocol

Patient’s dynamic balance is assessed using a walk test. The walk test is performed on

patients who can walk about 10 meters independently. For the walk test, a 3D acceleration

sensor is attached to the patient’s waist and the measurement data is transferred wirelessly

to the mobile device. Mobile device has an easy-to-use mobile application which has been

developed for this purpose by VTT, where measurement is started and stopped. Mobile

application does not store any personal information of patients.

1. Study nurse evaluates patients’ ability to walk independently about 10 meter

distance. If patient is able to perform walking test, study nurse will explain the

course and purpose of the measurement.

2. Study nurses attach accelerometer sensor (Movesense, size 36,6mm x 10,6mm,

weight 10g) around patients waist with fastening belt that comes with a sensor.

3. Study nurse will type patients previously determined patient code into application.

4. Study nurse follows instructions from the application, which gives guidelines to

study nurse for checking there is 10-meter space for walking, like for example

hallway. When patient is ready, study nurse starts the measurement from the

application by pressing “Aloita” button and stops measurement by pressing

“Stop” button. Measurement can be do again if needed.

Balance measurement takes about 5 minutes. Only study nurse will use mobile applicatio n

for measuring balance and patient does not get any feedback from the balance test.

7.7 Video and speech protocol

Video and speech recording is done by consent of patient. Any other persons, like for

example hospital staff, other patients, relatives and so on, are not visible on video

recording.

1. Research nurse presents test protocol for a patient.

2. Research nurse inputs predefined patient identification number to a mobile

application developed for this purpose.

3. Research nurse follows instructions given by an application and starts the

recording.

4. Patient is asked to perform following tasks.

a. Patient is asked to smile, show teeth, and raise eyebrows a few times.

b. Patient is asked to extend the other hand first to 90-degree angle and keep

eyes closed and keep hand extended 10 seconds and then perform same

tasks with other hand and then extend both hands to 90-degree angle with

eyes closed for 10 seconds.

c. Patient is asked to read aloud sentences which are presented to him in

paper.

d. Patient is asked to explain in aloud what is happening on a paper which is

showed to patient.

e. Patient is asked to follow with his eyes only an object which is moved to

left and right.

f. Patient is asked to lie down and asked to raise first other leg to 30-degree

angle and keep it erected for 5 seconds and repeat the same task with other

leg.

g. If Patient is able, he is asked to stand still for 30 seconds.

In part C and D showed sentences and pictures are shown in figure 10. Tasks are chosen

by following meters used in clinical work. National Institutes of Health Stroke Scale

(NIHSS), FPSS, CPSS, Los Angeles Prehospital Stroke Scale (LAPSS), and FAST/BE-

FAST.

1. When tasks are finished, recording is stopped from mobile application.

Video and speech recording is used only by research nurse and collected videos can only

view through mobile application. Mobile Application requires personal access code.

Estimated time for video on speech protocol is 5 minutes.

Figure 10, Picture From Stroke tutkimussuunnitelma

7.8 Video speech protocol including additional measurement

devices

7.8.1 Additional measurement devices used

In this protocol, DC-EEG, NIRS, Accelerometer, ECG and video is collected

simultaneously. Protocol is using following equipment’s and sensors.

 Bittium NeurOne and DC-EEG cap, which has also followed sensors attached to

it.

o NIRS

o 3x Accelerometer

o ECG sensors

 Video and speech recording with mobile application

7.8.2 Test protocol with additional measurement devices

1. Research nurse introduce test protocol to patient.

2. Research nurse enters predetermined patient identification on application

developed for measurement on laptop and in mobile application which has been

developed for recording video and speech.

3. Research nurse dress headcap for patient which includes DC-EEG, NIRS, and

Accelerometers.

4. Research nurse dress ECG sensors to patient’s chest.

5. Research nurse starts DC-EEG, NIRS, Accelerometer and ECG sensor recording

on computers and on mobile application by instructions from mobile application.

6. Patient is first asked to sit 5 minutes, then next lie down for 5 minutes and then sit

down 5 minutes again.

7. Patient is asked for breath and hold breath for a moment for CBF reserves

(Cerebral Blood Flow) calibration. Patient is asked to breath normally first 30

seconds and then hold breath for 30 seconds.

8. Patient is requested to perform following task.

a. Patient is asked to smile, show teeth, and raise eyebrows a few times.

b. Patient is asked to extend the other hand first to 90-degree angle and keep

eyes closed and keep hand extended 10 seconds and then perform same

tasks with other hand and then extend both hands to 90-degree angle with

eyes closed for 10 seconds.

c. Patient is asked to read aloud sentences which are presented to him in

paper.

d. Patient is asked to explain in aloud what is happening on a paper which is

showed to patient.

e. Patient is asked to follow with his eyes only an object which is moved to

left and right.

f. Patient is asked to lie down and asked to raise first other leg to 30-degree

angle and keep it erected for 5 seconds and repeat the same task with other

leg.

g. If Patient is able, he/she is asked to stand still for 30 seconds.

In part C and D showed sentences and pictures are shown in figure 10. Tasks are chosen

by following meters used in clinical work. National Institutes of Health Stroke Scale

(NIHHS), FPSS, CPSS, Los Angeles Prehospital Stroke Scale (LAPSS), and FAST/BE-

FAST.

1. When task is finished, recording is stopped from laptop and mobile application.

2. Research nurse removes sensors from a patient.

3. Research nurse disposes ECG sensors and cleans and disinfects headcap.

Patient part of protocol is estimated to take 1 hour, consisting of ten steps as follows:

 Steps 1-4: 30 minutes

 Steps 5-9: 15 minutes

 Step 10: 20 minutes

An additional step, which does not apply to the patient and consists of cleaning the

equipment, is estimated to take 20-30 minutes.

7.9 Fundus photography with Optomed Aurora

Photographs of the patient’s retina, in other words fundus are taken with Optomed Aurora

camera. The images are fist saved to cameras memory card, from where they are

transferred to the laptops hard drive and then transferred to secure network drive for the

study for later use. The photograph does not require for example the use of pupil dilators

or other prior preparations. The photographing event corresponds to normal

photographing with a digital camera. The inconvenience to the patient is limited to the

use of the flash.

1. Research nurse presents the test protocol to patient.

2. Research nurse creates in camera the folder named “study” and enter there the

patients pre-determined unique identification number ( ID ).

3. Research nurse takes photograph by following instructions.

4. At the end of the measurement, research nurse will move the images from camera

to separately specified storage space.

Estimated time for measurement protocol is 5 minutes. In addition, cleaning the camera

will take from research nurse approximately 1 minute. This step does not apply to the

patient time.

8. Development of the user interface for

immediate measurement by NIRS

Presently, the process of conducting measurements with NIRS device is challenging,

compounded by software that primarily caters to researchers. Consequently, the software

encompasses numerous functions irrelevant to nurses or lay users, rendering it

challenging to utilize effectively without requisite expertise.

Purpose of new interface in prototype software was to create easy to use user interface

for people who does not have knowledge or skills to use current interface efficiently.

8.1 Users feedback

Some of the requirements came from feedback of foreign researchers who are using our

NIRS device and our software. They were also in need of interface, which would be easy

to use and do not have anything extra information, which could confuse users. Currently

GUI has been developed mainly for researchers, and because of that, it can be challenging

to use efficiently without proper knowledge. One target group for future users from the

feedback, is for example research nurses.

8.2 Technical description

Hardware software has been developed with C and main GUI has been developed by QT

software and logic and functionality is done with python version 3.9. Main libraries used

in Python are, Pyqt5, Scipy, Pyqtgraph, Matplotlib, Numpy. Python programming has

been done with Visual studio Code and later with Pycharm Community Edition. Version

control has been done with a GIT. The GUI Window was created using QT Designer,

PyCharm and Python version 3.13.

8.3 The primary concept GUI

The primary concept behind the interface design was involved implementing a single

button to streamline most functions, allowing users to initiate measurements with just one

press. This guideline does follow persuasive design seventh postulate of easy to use. The

default time of 5 minutes was selected as it aligns with the typical duration used as a

baseline in measurements with NIRS devices. To enhance user’s clarity, a countdown

timer was integrated to distinctly display the remaining measurement time. This

deliberate choice aims to ensure that for example elderly home users can easily follow

and comprehend the remaining time for the measurement, reducing the likelihood of

prematurely stopping the process.

Layout design adhered to the principles of the Golden Cut, strategically placing more

significant GUI functions towards centre and middle of the interface. This approach

enhanced the visual appeal for the users. The pivotal functions, such as the START button

and timer indicator, command a more prominent presence through larger sizes and fonts.

This helps users who have challenges on hand motoric skills to press start button easily.

Additional information-providing functions are predominantly situated in the corners of

the interface. One of the design ideas is based on generation-oriented approach, as the

GUI uses the same principles as those found in blood pressure medical device interfaces,

which are widely used. This makes GUI very familiar and easy to understand and user-

friendly.

Colour scheme primarily employs black, white and gray to ensure the visibility of main

functions for potential users with colour blindness. The only instance of colour use is in

indicating signal quality, which colours ranging from red to yellow to green.

8.4 Functions provided

Following functions were already created for the original version of the GUI, and I chose

to use these functions in my GUI.

 Hr monitoring

 HRV monitoring

 Signal quality

 Measurement recording

 Controlled timer

Figure 11 shows the GUI, where HR and HRV were incorporated to provide nurses with

additional insights into the physiological conditions of test subjects. HR is valuable for

detecting potential health issues in test subjects, offering a means to identify other

underlying health conditions and for normal users, it gives also interesting information of

users HR.

Signal quality indicators were incorporated as per the request to provide additional

insights into the signal’s quality. This enhancement aids users in identifying whether there

is a necessity to adjust the position of the head unit.

Start button will handle Data recording, by using default filename with timestamp.

Pressing it again will stop data recording before timer goes to zero.

Configure timer button was added, so GUI can be used for different test protocols where

time would be different.

Figure 11. Screenshot of GUI Window

Software developed for a medical device, or intended to function as a medical device(e.g.,

operating a device, providing a user interface, or standalone software), is subject to strict

regulation under the European Union’s MDR. Such software is typically classified as a

Class IIa medical device under the MDR. If software or device is considered as wellness

product, then MDR regulation does not effect on it. In my thesis, I will be considering my

software as a wellness product in this thesis, because getting called a medical device,

software needs to have audited by notified body and pass all the regulations of the MDR,

so I won’t be taken in consideration most of the requirements which goes under MDR. I

introduced shortly some of the requirements in chapter 5.6, but I won’t go too deeply on

them, because that can be topic to whole new thesis.

Developing graphical user interface (GUI) can be challenging task, as it requires

balancing the needs and goals of the user with the constraints and capabilities of the

system. To create effective and user-friendly GUIs, it is important to follow best practices

and guidelines for design and evaluation article (Nielsen, Molich 1990).

Nielsen and Molich’s article (Nielsen, Molich 1990) introduces a method called heuristic

evaluation, which involves reviewing the interface against set of usability heuristics.

These heuristics, such as visibility of system status and consistency and standards,

provide a set of criteria for assessing the quality of the interface. Using heuristic

evaluation as a part of the design process can help identify problems and areas for

improvement in the interface. According to the article, most people probably perform

heuristic evaluation based on their own common sense and with their intuition.

Another important consideration in GUI design is the use of appropriate input and output

methods, like predict how much time is required for a user to move their pointer to a

target on the screen. This can be useful in designing efficient and effective input methods,

such as selection buttons and menu items article (Nielsen, Molich 1990).

In addition to usability and input/output methods, it is also important to consider the

overall appearance and layout of the interface. Some guidelines for designing user

interface software can include the use of good visual design principles, such as hierarchy

and balance and the use of appropriate widgets and layout techniques article (Nielsen,

Molich 1990).

It can be also helpful to use tools and resources to assist in the GUI design process. Like

for example various user interface design tools and their features, including prototyping,

simulation, and usability testing capabilities. Using such tools can help enhance the

design process and ensure that finished product is of high-quality (Nielsen, Molich 1990).

9. Results

In this section, I present results related to RQ1 in chapter 9.1 and RQ2 in chapter 9.2.

9.1 Feasibility of the multimodal setup for stroke data collection in

terms of easy to use and measurement time

The following table is based on my experience with making measurements using the

presented methods during Stroke project. The times provided are mostly estimations

based on my experience of how long it took us to complete the measurements

Table 2. Timetable for measurements used in stroke project

Method

Min

Time/min

Max

Time/min

Other information

Fundus camera

30+

We stopped trying to capture a

successful picture after 30 minutes of

unsuccessful attempts.

MoveSense

Software did cause sometimes

problems. Could not connect to sensor

successfully always.

EEG

2-3 people were working together

always at the same time to prepare EEG.

fNIRS

We were using MRI compatible version

of NIRS device with optical fibers and

LabView software which made using

the device slower

Fundus camera was quite small and lightweight. Learning to use it was also quite easy,

but taking good pictures did take some time and practice. It was not always easy to get

good pictures, because, for example, different problems people had in their eyes. The

older the test subject was, the more problematic it was to get good pictures. We

encountered a few test subjects, from we did not get any good pictures, and we had to

skip that part of the test.

MoveSense was very small in size and very easy to use. The only problem with it was

that we had challenges getting connected to mobile applications. Sometimes we needed

to remove the battery and try with better luck with a full power battery.

NeurOne EEG device did take a long time to prepare for measurements. Sometimes it

was bit uncomfortable for test subjects, so when test subjects were not in good shape and

were getting already too tired, we sometimes divided measurements to two days and did

easier measurements during the first day and at second day we did EEG measureme nts.

Couple time test subjects commented about preparing EEG headcap as “head massage”.

While preparing headcap with elderly subjects, we needed to be more careful, because

the skin is no longer that durable as compared with younger test subjects. NeurOne also

required quite a lot space and was not easy to move between the hospital and University

side of the building.

The NIRS device was housed in a large metal box, which was quite heavy. Due to the

long fibers, it was not easy to operate. The holders were not connected to each other,

making it difficult to maintain an exact 3 cm distance between them. We had to measure

the distance manually using a ruler, and holders moved quite easily during position

changes from sitting to supine.

Software required some experience to operate effectively and to recognize signal quality

issues. At times, the fibers pressed a bit too hard on the test subject’s foreheads, and we

received a few comments indicating that the holders felt uncomfortable.

With our newly developed wearable multimodal fNIRS device, we can significantly

reduce preparation time and minimize the stress associated with using multiple

measurement devices for stroke patients. This innovation will enable faster (as shown in

Table 3) and more reliable stroke diagnosis in future, as doctors will have access to

multimodal data to help differentiate stroke from symptoms caused by other medical

conditions.

9.2 Software interface and device development results when using

NIRS interface

Quickly diagnosing a stroke is crucial, as it directly impacts recovery and rehabilitatio n

outcomes. In the ambulance, a simple FAST test is typically used to assess stroke, and it

is a reliable method for rapid diagnosis. However, when symptoms overlap with

conditions like migraine or vascular dementia, additional tools may be needed to

distinguish a stroke from less acute conditions. The 4-hour windows has been established

as a key benchmark for determining when stroke treatment should begin from the start of

the symptoms, as initiating treatment within this timeframe can significantly improve

recovery and rehabilitation outcomes for stroke patients.

If emergency medical personnel in the ambulance can diagnose a stroke early, they can

initiate treatment right away, if feasible, and notify the hospital in advance. This allows

the hospital to prepare for the patient's arrival and begin treatment quickly as possible.

On emergency situations, time is of the essence. There may be instances where patients

are unconscious or unable to respond to verbal cues. In such cases, a NIRS device could

be invaluable, as it can be quickly deployed to assess whether the patient is experiencing

a stroke, or at the very least, help rule out one potential cause from the diagnosis.

The software interface (figure 11.) is now easier and faster to operate than our regular UI.

Starting the measurement is quicker and does not require high technical skills to configure

settings. Also, software has been developed from LabView program to Python, which

makes it easier to use on different computers and GUI has been improved also to show

more important data for researchers. The name of GUI could be changed in future,

because users could be also other than nurses or elderly home users.

Devices have been developed to be smaller, which makes it possible to use as a wearable

device. Connectors are reduced to a minimum, which makes it easier to operate. Head

unit needs now only one connector which makes using the device easier and lowers the

mistakes on connecting head unit sensors.

Figure 12 shows multimodal test measurement done together with our wearable NIRS

device with ACM sensors and EEG data from clinically approved medical device

NeurOne. Data has manually synchronized, which can be seen on the picture figure 12

by three artefacts. Manual synchronization was done by tapping different sensors three

time at the same time.

Figure 12. Multimodal measurement with NIRS, ACM and EEG

signals with manual synchronization artefact.

Overall, I conducted over 100+ measurements in this project using different versions of

our software. I made tests on myself to measure how long it would take to start program

and until recording is started. With four tests, I got an average time of 54 seconds with

my GUI. With our current version of software, it would take about 30-60 seconds longer

time. This time would be probably bit longer if users don’t have much experience on

using the device and software, which I did have quite a lot. In my tests, I had all cables

disconnected. Having them already connected to the device, or replacing them with

wireless solutions, would significantly reduce startup time, making the device easier to

operate, especially in ambulances.

I made short tests with six subjects from our research group. The test was the same as I

did myself. I gave verbal instructions to test subjects who had lower experience on

conducting measurement, on how to use devices. All the cables were disconnected, and

test subjects used my head to install headband. The test started when program was clicked

to start and ended when the start button was pressed on the interface. Results are shown

in Table 3.

Table 3. Speed test for starting measurement with new GUI and wearable NIRS device

Test Subject

Level of

experience

Starting the

measurement

time (s)

Measurement

time (s)

Estimation

of time for

removing

the device

from test

subjects

head (s)

High

300

Low

124

300

Low

144

300

Low

300

Low

300

Low

300

Average

300

Standard deviation

30,49044

0,534522

The test result did show clear differences between experienced users and less experienced

users. Two users tried the test a second time and they managed to decrease their time by

30 seconds and the other by 59 seconds. For my opinion, this indicates that our device

and software is easy to learn so operating time can be lowered significantly with very

little practice. Compared to MRI compatible NIRS devices operating time, with our

wearable device, shows significant results on operating time. On the test we run 5-minute

standard test and then the removing device part consisted of removing the headband from

the head. Compared to Stroke project measurements with NIRS and EEG where it could

take sometimes even 1 hour to prepare time for starting measurements, we have made

significant improvements on how quickly we can start measurements.

In the future, one goal is to reduce preparation time to under one minute. This can be

achieved by improving how the head unit is connected to the test subject’s forehead,

making the device wireless, and simplifying the user interface.

Succes rate on measurements with proto3 version is quite good. For experienced users

the acceptable collected data is about 75%. When the protocol is calm, the possibility of

getting good signal is high.

For example, from one measurement, where 30 subjects were measured. There were 3

protocols used, which were supine to stand, sit to stand and breath hold protocols. In

supine to stand protocol, 22-25 out of 30 were at least acceptable, with few of them having

differences between left and right channel. In Sit to stand protocol, 22-25 out of 30 were

at least acceptable, with few of them having differences between left and right channel.

In Breath hold protocol, 21-29 out of 30 were at least acceptable, with few of them having

differences between left and right channel.

Compared to different techniques used in the Stroke project, our wearable solution is

significantly faster. For example, with NeurOne, preparing time was usually about 20

minutes when three people were working with a patient. With the Fundus camera, there

was much more variability because taking pictures of older patients was more difficult

compared to younger people. The quickest attempt took about 5 minutes, but sometimes

we had to stop trying because we couldn’t get good quality pictures. After about 20

minutes, we usually stopped and moved on to the next measurement.

The Movesense sensor has its limitations because patients are not always able to walk

due to paralysis. Video recording is also not practical in ambulances or emergency care,

as it requires more space and is currently a bit slow to use. However, it may have good

potential in the future with more efficient AI software.

Our developed fNIRS device is capable of quickly and easily starting brain monitoring

and recording for stroke patients.

10. Discussion

Every sixth Finnish people will have once in their lifetime a cerebrovascular accident,

also known as a stroke. Stroke is third expensive disease group, which yearly costs is

estimated to be 1.6 billion euros. Stroke kills in every year about 4500 people in Finland,

which makes it third most common cause of death. It is also notable that stroke can cause

permanent disadvantage to half of survivors. Permanent disadvantage causes great danger

for disability for work and because of that, it has big impact on social costs. For stroke

treatment, fast diagnose and starting of treatment is essential.

There are different kinds of stroke, like TIA which usually is only temporary and last

usually under one hour and mostly 2-15 min, and ICH and SAV. About 10-20% of TIA

patients will get stroke under 90 days if correct treatment is not given. If patient needs

dissolution treatment, time window is about four hours from starting of symptoms. Also,

doctors need comprehensive information of starting symptoms, previous diagnoses,

medication and other risk factors. This critical pathway needs new technologies which

helps faster treatment for stroke patients.

Estimation is that 8/10 of strokes could be prevented with early intervention on risk

factors like obesity, high blood pressure, diabetes, alcohol usage, mid-body obesity,

smoking, heart-based diseases, stress, bad diet, lack of exercise and high blood fat values.

Risk factors are known well, but more research information is needed for how to monitor

different risk factors and how background factors can effect on patients’ recovery. Stroke

risk factors identification and anticipation could help on reducing following health and

economy costs considerably.

New research is showing that stroke and glymphatic system disturbance are closely

related. It is also found that glymphatic system can be disturbed by migraine and

dementia. Differential diagnostics can be challenging, and these different diseases should

be separate from different blood circulatory diseases. One of the goals in Stroke project

is to develop new methods to separate blood circulatory and glymphatic system

disturbances.

The Stroke project was quite a big project with several different collaborators from two

different University hospitals and several different companies from the private sector.

Also, during the project, we had two big things that did an effect on the project, and the

biggest was a global pandemic, which influenced project planning, how the work needed

to be done and how we were needed to communicate with each other. We did not meet

face to face, so communication was done remotely when possible and planning that was

its own work. There was also a strike with nurses, and this did influence collecting the

data.

In KYS research room which were planned for project, did change because reorganization

of wards in KYS. Also, in OYS we had to find new space for doing measurements during

the project. In Oulu our room was not big enough for our needs, so we did have small

problems placing all the measuring devices well enough in the room, especially when

patients had to change positions. We had all three (two research nurses and me) helping

during these position changes. Everyone had some cables to handle or devices to carry

during movement or taking care of moving other objects away or correct position, for

example examination bed to correct position. Few times the electrodes did move also, and

we had to correct sensor positions back to correct place and optical fiber cables got also

some damage during the project while patient was changing position, even though we

tried to be very careful when handling optical fibers, they still got slightly damaged and

we had to replace them once during the project. Also, our NIRS device needed LED

replacements once and we had some smaller problems with other devices which did affect

how we could do measurements.

The data collection model changed during the project planning period. Originally the plan

was to use one research nurse, but when we tested protocols before starting actual

measurements with a project leader and with research nurse, we quickly found out that it

would be very difficult to do all the measurements alone. Handling all the different

devices and taking care of a patient’s safety at the same time would have been too much

for one person. Taking care of patients’ safety was a priority one for us and because

conditions with stroke patients could change from walking on their own to moving in bed,

it did bring own challenges when working with patients. Sometimes we could not do all

the measurements because of the condition of the test subject.

There were challenges on recruiting patients and different recruiting methods between

KYS and OYS. In KYS physicians at ward did present the project to suitable patients and

asked their interest to join as a test subject for research, and in OYS we recruited patients

with research nurses from the ward by asking from nurse’s possible patients which were

suitable for being our test subjects and then we went discussing with patients and tried to

recruit them. In Oulu we could only get patients for our measurements in the late

afternoon when normal daily routines in the ward have ended. Luckily most of the people

we were asking to participate were interested in joining in our research.

They’re existed variations in patient recruitment between OYS and KYS. While KYS had

a limited number of TIA patients, OYS predominantly enrolled TIA patients. TIA patients

posed less complexity as test subjects since they typically exhibited normal mobility.

Later in Oulu we recruited control group subjects by sending emails to hospitals mailing

lists. Challenge recruiting control group test subjects were requirements, which were at

start preferably over 60 years old, healthy and with no medication. This was then changed

so that some medicines were allowed, like for example blood pressure medicines which

did help recruit test subjects.

The plan on the project was also to recruit test subjects with severe migraine and people

with dementia. On this goal we failed, and we managed to measure only one test subject

with migraine. There were challenges in the recruitment process for both groups and also,

we were running out of time on the project, and because of that, priority was stroke

patients and control group.

Obtaining a new server for data analysis presented significant hurdles, particularly in

navigating legal considerations. Eventually, we identified suitable servers provided by

CSC. Since the project plan lacked an allocated budget for ICT service expenses, this

posed a major challenge in determining how researchers could gain access to the data

collected. The lengthy duration of this process prevented researchers from assessing the

data’s quality during collection. Access to the data was granted only after our data

collection had concluded, leaving us unaware of its suitability for research purposes. All

these caused that project to apply for more time.

Data recording needs precise naming for files. Even though we designed to use codes on

naming the files, there were still too many differences on naming files. Even that we

practiced the test protocols with KYS and OYS side research nurses, there were still some

differences, like for example NeurOne had different sampling rate. Luckily this was not

a problem. Also, i noticed afterwards that I should have been more active with KYS

research nurses to ask if they need help with using devices and I should have visited there

more often to practice test protocols more with them. When I asked for comments from

them after we finished collecting data, they told us that they were not confident on using

our NIRS device. More training would have increased their confidence in using our

device.

The need for new interface did arise from the user feedback and from the future plans to

use devices in different user groups outside of researchers, like home users and nurses for

example. Main idea was to create user interface which is very easy to use and still have

main functions of the main program. The concept of a single-button interface draws

inspiration from devices such as blood pressure monitors, where the user interface

operates seamlessly with just one button.

Techniques used in stroke project are not at the moment very reliable for detecting

strokes. Stroke symptoms can be detected from the subject’s face, but confirmation still

needs other tests, like for example, MRI. That is where AI can be useful in future to detect

strokes from video recording, when it can be used easily with, for example, by mobile

phones.

Eye camera technics have a problem with a different eye. From my experience, taking

good pictures can be challenging, because the patient can’t stay still, they can’t keep their

eyes open enough or they have some other problems with their eyes which make taking

good pictures really challenging. It is more suitable for use in a controlled environment

and with subjects who don’t have, for example, cataracts.

Detecting stroke from balance is a novel technique, but it has some limitations. Where

biggest one is that subjects need to able to walk by themselves without assistance and

results could give wrong diagnoses if the subject already had problems in walking by

some other reasons like for example injury in leg or hip which effects on walking. For my

opinion, this technique won’t be very reliable in near future.

NIRS technique looks most promising from techniques used in stroke projects. NIRS

device has been developed to be wearable and easy to use, so it can fit easily in

ambulances or use in emergency care. Preliminary results from NIRS data analysis are

promising, however, this analysis work was not part of this thesis but will be later

published in a journal.

Tests in our research group show that our device can be utilized quickly with a little

training. Our test group had big variation on how skilled users were doing measurements.

Actual device was known to all, but only one of the test subjects had done actual

measurements with our device and others had mostly experience of being test subject.

One of the test subjects was developing our device software and had mostly experience

on testing software changes with the device.

Development of NIRS devices to be wearable and developments on software and head

unit made measurements easier and faster. This also helps to lower the stress on patients

because the time needed for making measurements is shorter and head unit feels more

comfortable.

We can increase the starting time in future with, for example, making the new GUI start

as a only interface for users. This would increase the starting time considerably. I have

also tested with our newer prototype version, but for this thesis, I did not include our latest

prototype version, and it already shows that it is quicker and easier to use. We are also

planning in future to develop devices to be wireless and using Android applications.

These changes can make device to more quicker to use. Also connecting device to

forehead is having improvements, which would help in future use in emergency

situations, making installment of device to patient faster and safer, which would help

using the device for example in ambulances or emergency care in hospitals.

For my opinion, NIRS technology does have bright future as a medical device and also

as a wellbeing device in consumer markets.

11. Conclusion

The development of the NIRS device from desktop version to wearable format has

expanded its potential applications across various settings. This advancement enables its

use not only in hospital environments but also in consumer markets. Additionally, as the

device becomes more compact and the software more user-friendly, ne clinical

applications within hospital settings continue to emerge.

With a smaller size, lighter weight, and improved holder connections to the forehead, the

NIRS device has become more practical for use in various hospital wards, including

emergency care. Additionally, a simplified GUI allows even people with limited

technological familiarity to operate the device with minimal training.

In the future, wearable NIRS technology could serve as a rapid diagnostic tool for stroke

detection, particularly in ambulances, where first responders can quickly differentiate

stroke from migraine or vascular dementia symptoms. This early diagnosis would enable

faster treatment initiation, improving patient outcomes and aiding stroke rehabilitation.

Additionally, quicker intervention and enhanced recovery could contribute to reducing

the overall burden on healthcare costs.

For future research applications, further development of the NIRS device is needed,

particularly in the holder design, to enable quicker and easier attachment to the forehead

without requiring an additional headband for stability. While the device is already

wearable, future iterations will be smaller and lighter, enhancing usability and comfort

for extended measurements, such as in sleep research.

Development is also underway to make the NIRS device wireless and operable via mobile

software, further enhancing its usability. Future software improvements will focus on

increasing user-friendliness, reducing the need for extensive training, and making the

device more accessible for a wider range of users. Signal analysis could also be developed

to function in the cloud, enabling automated processing and rapid result delivery to users

mobile software. This would allow users to quickly access metrics such as the clearance

of the glymphatic system during sleep.

In my opinion, wearable NIRS technology has a promising future in healthcare and may

also have potential applications in consumer markets too.

12. References

Ataş, İ., Ersunan, G., Bỉlỉr, Ö., Yavaşỉ, Ö., Altuntaş, M., & Karakullukçu, S. (2024). The

utility of NIRS in follow-up of patients with acute ischaemic stroke treated with IV

thrombolysis and mechanical thrombectomy in the emergency department. Journal of

Thrombosis and Thrombolysis, 57(3), 466-472.

Atula, S. (2019). Vaskulaarinen dementia (verenkiertoperäinen muistisairaus).

Lääkärikirja Duodecim.

Ansado, J., Chasen, C., Bouchard, S., & Northoff, G. (2021). How brain imaging provides

predictive biomarkers for therapeutic success in the context of virtual reality cognitive

training. Neuroscience & Biobehavioral Reviews, 120, 583-594.

Banks CA, Bhama PK, Park J, Hadlock CR, Hadlock TA. Clinician-Graded Electronic

Facial Paralysis Assessment: The eFACE. Plast Reconstr Surg. 2015 Aug;136(2):223e-

230e. doi: 10.1097/PRS.0000000000001447. PMID: 26218397.

Berg, K. O., Wood-Dauphinee, S. L., Williams, J. I., & Gayton, D. (1989). Measuring

balance in the elderly: preliminary development of an instrument. Physiotherapy Canada,

41(6), 303–311.

Biedrzycka, A., & Lango, R. (2016). Tissue oximetry in anaesthesia and intensive care.

Anaesthesiol Intensive Ther, 48(1), 41-8.

Chaudhary, U., Birbaumer, N., & Curado, M. R. (2015). Brain-machine interface (BMI)

in paralysis. Annals of physical and rehabilitation medicine, 58(1), 9-13.

Cheung, C. Y. L., Tay, W. T., Ikram, M. K., Ong, Y. T., De Silva, D. A., Chow, K. Y.,

& Wong, T. Y. (2013). Retinal microvascular changes and risk of stroke: the Singapore

Malay Eye Study. Stroke, 44(9), 2402-2408.

Çınaroğlu, O. S., Bora, E. S., Acar, H., Arıkan, C., Küçük, M., & Kırık, S. (2024). Is near-

infrared spectroscopy a promising predictor for early intracranial hemorrhage diagnosis

in the Emergency Department?. Brazilian Journal of Medical and Biological Research,

57, e13155.

Crofts, A., Kelly, M. E., & Gibson, C. L. (2020). Imaging functional recovery following

ischemic stroke: clinical and preclinical fMRI studies. Journal of Neuroimaging, 30(1),

5-14.

DuoDecim. (n.d.). Akuutin aivoinfarktin kuvantaminen.

https://www.duodecimlehti.fi/duo92542

DuoDecim, Käypä Hoito. (2024). https://www.kaypahoito.fi/hoi50051#s18

Esteva, A., Chou, K., Yeung, S., Naik, N., Madani, A., Mottaghi, A., Liu, Y., Topol, E.

(2021) Deep learning-enabled medical computer vision. Npj Digital Medicine, 4(1):5.

Ferdinando, H., Moradi, S., Korhonen, V., Helakari, H., Kiviniemi, V., & Myllylä, T.

(2022). Spectral entropy provides separation between Alzheimer’s disease patients and

controls: a study of fNIRS. The European Physical Journal Special Topics, 1-8.

Files, B. (2011). An introduction to EEG. Perception

Gao, T., Liu, S., Wang, X., Liu, J., Li, Y., Tang, X., ... & Fan, Y. (2023). Stroke analysis

and recognition in functional near-infrared spectroscopy signals using machine learning

methods. Biomedical Optics Express, 14(8), 4246-4260.

Guo, X., & Dye, J. (2023). Modern Prehospital Screening Technology for Emergent

Neurovascular Disorders. Advanced Biology, 7(10), 2300174.

Hara, Y. (2015). Brain plasticity and rehabilitation in stroke patients. Journal of Nippon

Medical School, 82(1), 4-13.

Hevner, A. R. (2007). A three cycle view of design science research. Scandinavian

journal of information systems, 19(2), 4.

Huang, Y. H., Chen, W. Y., Liu, Y. H., Li, T. Y., Lin, C. P., Cheong, P. L., ... & Wu, C.

C. (2024). Mild cognitive impairment estimation based on functional near‐infrared

spectroscopy. Journal of Biophotonics, 17(1), e202300251.

Johnston, S. C., Rothwell, P. M., Nguyen-Huynh, M. N., Giles, M. F., Elkins, J. S.,

Bernstein, A. L., & Sidney, S. (2007). Validation and refinement of scores to predict very

early stroke risk after transient ischaemic attack. The Lancet, 369(9558), 283-292.

Jurík, M., Siarnik, P., Valovičová, K., Karapin, P., Klobucnikova, K., Turčáni, E., &

Kollar, B. (2021). Cerebral blood flow in stroke patients with sleep apnea: any role of

single-night positive airway pressure therapy?. Neuroendocrinology Letters, 42(7).

Kallela, M., & Lindsberg, P. J. (2012). Miten erotan migreeniauran TIA-kohtauksesta?.

Duodecim, 128(9), 971-977.

Kato, K., Miyata, S., Ando, M., Matsuoka, H., Yasuma, F., Iwamoto, K., ... & Noda, A.

(2017). Influence of sleep duration on cortical oxygenation in elderly individuals.

Psychiatry and Clinical Neurosciences, 71(1), 44-51.

Katti, G., Ara, S. A., & Shireen, A. (2011). Magnetic resonance imaging (MRI)–A

review. International journal of dental clinics, 3(1), 65-70.

Kiviniemi, V., Korhonen, V., Kortelainen, J., Rytky, S., Keinänen, T., Tuovinen, T., ... &

Alahuhta, S. (2017). Real-time monitoring of human blood-brain barrier disruption. PLoS

One, 12(3), e0174072.

Koyanagi, M., Yamada, M., Higashi, T., Mitsunaga, W., Moriuchi, T., & Tsujihata, M.

(2021). The usefulness of functional near-infrared spectroscopy for the assessment of

post-stroke depression. Frontiers in Human Neuroscience, 15, 680847.

Kumar, M. A., Vangala, H., Tong, D. C., Campbell, D. M., Balgude, A., Eyngorn, I., ...

& Albers, G. W. (2011). MRI guides diagnostic approach for ischaemic stroke. Journal

of Neurology, Neurosurgery & Psychiatry, 82(11), 1201-1205.

Kuriakose, D., & Xiao, Z. (2020). Pathophysiology and treatment of stroke: present status

and future perspectives. International journal of molecular sciences, 21(20), 7609.

Liedes H, Mattila J, Lingsma H, et al. Prediction of outcome after traumatic brain injury:

Comparison of disease state index and IMPACT calculator. In: Studies in Health

Technology and Informatics. Vol 224. IOS Press; 2016:175-180.

Li, Y., Ma, Y., Ma, S., Liang, Z., Xu, F., Tong, Y., ... & Li, X. (2020). Asymmetry of

peripheral vascular biomarkers in ischemic stroke patients, assessed using NIRS. Journal

of Biomedical Optics, 25(6), 065001-065001.

Liu, S., & Joines, S. (2012, September). Developing a framework of guiding interface

design for older adults. In Proceedings of the Human Factors and Ergonomics Society

Annual Meeting (Vol. 56, No. 1, pp. 1967-1971). Sage CA: Los Angeles, CA: SAGE

Publications.

London, A., Benhar, I., Schwarz, M. (2013) The retina as a window to the brain – from

eye research to CNS disorders. Nat. Rev. Neurol, 9: 44-53.

Mattila, J., Koikkalainen, J., Virkki, A., Simonsen, A., van Gils, M., Waldemar, G., …

Lötjönen, J. (2011). A Disease State Fingerprint for Evaluation of Alzheimer’s Disease.

Journal of Alzheimer’s Disease, 27(1), 163–176.

Meretoja, A., Kaste, M., Roine, R. O., Juntunen, M., Linna, M., Hillbom, M., ... &

Häkkinen, U. (2011). Direct costs of patients with stroke can be continuously monitored

on a national level: performance, effectiveness, and Costs of Treatment episodes in Stroke

(PERFECT Stroke) Database in Finland. Stroke, 42(7), 2007-2012.

Minoshima, S., Drzezga, A. E., Barthel, H., Bohnen, N., Djekidel, M., Lewis, D. H., ... &

Van Laere, K. (2016). SNMMI procedure standard/EANM practice guideline for amyloid

PET imaging of the brain 1.0. Journal of Nuclear Medicine, 57(8), 1316-1322.

Musen, M. A., Middleton, B., & Greenes, R. A. (2014). Clinical Decision-Support

Systems. In Biomedical Informatics (pp. 643–674). London: Springer London.

Muehlschlegel, S., Selb, J., Patel, M., Diamond, S. G., Franceschini, M. A., Sorensen, A.

G., ... & Schwamm, L. H. (2009). Feasibility of NIRS in the neurointensive care unit: a

pilot study in stroke using physiological oscillations. Neurocritical care, 11(2), 288.

Myllylä, T., Harju, M., Korhonen, V., Bykov, A., Kiviniemi, V., & Meglinski, I. 2018.

”Assessment of the dynamics of human glymphatic system by near‐infrared

spectroscopy.” Journal of biophotonics 11 (8).

Myllylä, T., Kaakinen, M., Vihriälä, E., Jukkola, J., Zhao, Z., Ferdinando, H., ... &

Eklund, L. (2020, May). Functional NIRS study of blood brain barrier disruption when

induced by focused ultrasound and intra-arterial mannitol infusion. In Tissue Optics and

Photonics (Vol. 11363, pp. 74-79). SPIE.

Naseer, N., & Hong, K. S. (2015). fNIRS-based brain-computer interfaces: a review.

Frontiers in human neuroscience, 9, 3.

Nielsen, J. (2005). Ten usability heuristics.

Nielsen, J., & Molich, R. (1990, March). Heuristic evaluation of user interfaces. In

Proceedings of the SIGCHI conference on Human factors in computing systems (pp. 249-

256).

Niziol R, Henry FP, Leckenby JI, Grobbelaar AO. 2015. Is there an ideal outcome scoring

system for facial reanimation surgery? A review of current methods and suggestions for

future publications. J Plast Reconstr Aesthet Surg. 2015 Apr;68(4):447-56. doi:

10.1016/j.bjps.2014.12.015. Epub 2014 Dec 24. PMID: 25589458.

Oinas-Kukkonen, H., & Harjumaa, M. (2009). Persuasive systems design: Key issues,

process model, and system features. Communications of the association for Information

Systems, 24(1), 28.

Olejniczak, P. (2006). Neurophysiologic basis of EEG. Journal of clinical

neurophysiology, 23(3), 186-189.

OpenNIRS Documentation (n.d.) OpenNIRS.org – Modular open hardware for Near

InfraRed Spectroscopy. http://www.opennirs.org/documentation/

Orhii, P. D., Haque, M. E., Fujita, M., & Selvaraj, S. (2023). Advances in PET imaging

of ischemic stroke. Frontiers in Stroke, 1, 1093386.

Poplin, R., Varadarajan, A. V., Blumer, K., Liu, Y., McConnell, M. V., Corrado, G. S.,

Peng, L., Webster, D. R. (2018) Prediction of cardiovascular risk factors from retinal

fundus photographs via deep learning. Nat Biomed Eng. 2: 158-164.

Podsiadlo, D., & Richardson, S. (1991). The timed “Up & Go”: a test of basic functional

mobility for frail elderly persons. Journal of the American Geriatrics Society, 39(2), 142–

148

Rindler, R. S., Allen, J. W., Barrow, J. W., Pradilla, G., & Barrow, D. L. (2020).

Neuroimaging of intracerebral hemorrhage. Neurosurgery, 86(5), E414-E423.

Shneiderman, B., Plaisant, C., Cohen, M. S., Jacobs, S., Elmqvist, N., & Diakopoulos, N.

(2016). Designing the user interface: strategies for effective human-computer interaction.

Pearson.

Similä, Heidi, Milla Immonen, ja Miikka Ermes. 2017. ”Accelerometry-based

Assessment and Detection of Early Signs of Balance Deficits.” Computers in Biology and

Medicine 85: 25-32.

Teplan, M. (2002). Fundamentals of EEG measurement. Measurement science review,

2(2), 1-11.

Thevenot J., Bordallo López M., Hadid A., (2017) A Survey on Computer Vision for

Assistive Medical Diagnosis from Faces, IEEE Journal of Biomedical and Health

Informatics

Thientunyakit, T., Shiratori, S., Ishii, K., & Gelovani, J. G. (2022). Molecular PET

imaging in Alzheimer’s disease. Journal of Medical and Biological Engineering, 42(3),

301-317.

Thomas, R., Shin, S. S., & Balu, R. (2023). Applications of near-infrared spectroscopy in

neurocritical care. Neurophotonics, 10(2), 023522-023522.

van Swieten JC, Koudstaal PJ, Visser MC, Schouten HJ, van Gijn J. 1988 Interobserver

agreement for the assessment of handicap in stroke patients. Stroke. May;19(5):604-7.

Wilkinson, C. M., Burrell, J. I., Kuziek, J. W., Thirunavukkarasu, S., Buck, B. H., &

Mathewson, K. E. (2020). Predicting stroke severity with a 3-min recording from the

Muse portable EEG system for rapid diagnosis of stroke. Scientific Reports, 10(1), 18465.

Yew, K. S., & Cheng, E. M. (2015). Diagnosis of acute stroke. American family

physician, 91(8), 528-536.

Yu, M., Cai, T.,Huang, X., Wong, K., Volpi, J., Wang, J.Z., and Wong, S.T.C., 2020.

Toward Rapid Stroke Diagnosis with Multimodal Deep Learning, MICCAI 2020, LNCS

12263, pp. 616–626.

0 views·68 pages

Multimodal monitoring setup for stroke patient measurements and development of an easy-to-use user interface for the data collection PDF Free Download

Multimodal monitoring setup for stroke patient measurements and development of an easy-to-use user interface for the data collection PDF free Download. Think more deeply and widely.

Uploaded by kara on 4/29/2026

/68

100%