Machine-learning-driven Wearable Healthcare for Dementia:

A Review of Emerging Technologies and Challenges

Akio Sashima

Human Augmentation Research Center, National Institute of Advanced Industrial Science and Technology (AIST),

Kashiwa II Campus, University of Tokyo, 6-2-3 Kashiwanoha, Kashiwa, Chiba, 277-0882, Japan

Keywords:

Healthcare, Dementia, Wearable, Machine Learning, Digital Biomarker.

Abstract:

As personal mobile devices, such as smartphones and smartwatches, are increasingly commoditized, it has

become easier to measure individual physiological and physical states and record them continuously. Ap-

plying machine learning techniques to the data, we can detect early signs of diseases in older people, such

as dementia, and predict probabilities of future disorders. This review paper describes the machine learning

technologies in realizing wearable healthcare for older people. First, we survey the literature on machine-

learning-driven wearable technologies for the early detection of dementia. Second, we discuss issues of the

datasets for constructing ML models. Third, we describe the need for a service framework to collect longitudi-

nal data through continuous monitoring of the user’s health status. Finally, we discuss the socially acceptable

implementation of the service framework.

1 INTRODUCTION

Dementia is caused by brain disorders and diseases,

such as Alzheimer’s Disease (AD), Dementia with

Lewy Bodies, Vascular dementia, Frontotemporal de-

mentia, Parkinson’s disease with dementia (PDD).

Because there is no cure for dementia, early detec-

tion of the symptoms is essential to prevent disease

exacerbation.

Many medical screening methods for dementia

have been proposed (Turner et al., 2020; Thabtah

et al., 2020). Cognitive assessment tools (Cordell

et al., 2013; Giebel and Challis, 2016) are often used

for detection by investigating cognitive domains (at-

tention, language, memory, and visuospatial function)

because of their simplicity. The tools are essential for

the diagnosis of dementia in the hospital. However,

most persons with dementia (PwDs) do not notice a

decline in their cognitive function in the early stages.

They do not go to the hospital until noticeable demen-

tia symptoms are progressing. Thus, detecting the in-

dividual cognitive decline at home is an important is-

sue. A home screening system that informs doctors

about user changes is required.

To capture changes in the cognitive functions of

individuals, the use of digital biomarkers has drawn

https://orcid.org/0000-0002-6414-6506

attention in recent years. They collect, track, and an-

alyze patients’ social, behavioral, and physiological

states using sensors, such as cameras, accelerometers,

barometers, GPS, and microphones (Kourtis et al.,

2019; Piau et al., 2019; Husebo et al., 2020). Be-

cause recent wearable devices (e.g., smartwatches)

have these sensors, they can be used to detect dig-

ital biomarkers of users’ unnoticeable diseases. As

many people use smartwatches, there is a large med-

ical impact on the early detection of individual cog-

nitive decline at home. In addition, recently, machine

learning (ML) technologies are also used to detect de-

mentia(Tsang et al., 2020; Tanveer et al., 2020). It is

still a new challenging research area to apply ML to

detecting early signals of dementia from the sensor

data on commercially available wearable devices.

In this short review, we survey the literature on

using wearable and ML technologies to detect early

symptoms of dementia and discuss its current status

and future challenges. First, we conduct a literature

review of relevant articles from major publishing sites

and academic libraries. Second, we discuss issues of

constructing the datasets for ML models. Third, we

point out the need for a framework construction to

collect longitudinal data through continuous monitor-

ing of the user’s health status. Finally, we describe the

challenges of making such a data collection frame-

work socially acceptable.

864

Sashima, A.

Machine-learning-driven Wearable Healthcare for Dementia: A Review of Emerging Technologies and Challenges.

DOI: 10.5220/0010973900003123

In Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2022) - Volume 5: HEALTHINF, pages 864-871

ISBN: 978-989-758-552-4; ISSN: 2184-4305

2 WEARABLE DEVICES AND

DIGITAL BIOMARKERS FOR

DEMENTIA

In this section, we review the literature on detecting

early symptoms of dementia by applying ML meth-

ods to various types of digital biomarkers. We con-

ducted a literature review by examining relevant ar-

ticles from major publishing sites and academic li-

braries. We selected appropriate literature related to

the digital biomarkers sensed by the wearable devices.

The summary of the literature is shown in Table 1.

2.1 Physiological Signals

Rim et al. (Rim et al., 2020) provided a review paper

on DL applications in physiological signal data. For

early detection of AD, electroencephalogram (EEG)

has recently become a promising area of AD (Al-

Jumeily et al., 2014; D. Kim and K. Kim, 2018). Bi

et al. (Bi and Wang, 2019) proposed an EEG spectral

image classiﬁcation with a multi-task learning strat-

egy based on a convolutional high-order Boltzmann

machine. Since current wearable EEG devices allow

long-term noninvasive recording of brain signals out-

side of a laboratory (Casson, 2019), ML-driven wear-

able EEG is an important research area for the early

detection of dementia.

2.2 Human Activities

Patients of MCI and PwDs are characterized by decre-

ments in instrumental activities of daily living (IADL)

(Sacco et al., 2012; Lindbergh et al., 2016). PwDs

often show speciﬁc patterns of behavioral and psy-

chological symptoms of dementia, such as hyperac-

tivity (van der Linde et al., 2016). By monitoring and

recognizing their daily activities using wearable de-

vices, we may be able to detect the speciﬁc behav-

ioral patterns of dementia. Besides, the amount of

social interaction and physical activity in the lifestyle

of older people directly affects cognitive decline that

may progress to dementia (Kelly et al., 2017; Najar

et al., 2019). Thus, human activity recognition (HAR)

can be used as an assessment tool for estimating the

risk of dementia.

HAR using wearable devices is a popular appli-

cation area of ML. To capture human motion, we can

use sensors in smartphones or smartwatches placed on

the user’s body. The sensors include accelerometers,

gyroscopes, and inertial measurement units. By ap-

plying various ML methods to the sensor data, such as

CNN with transfer learning (Akbari and Jafari, 2019),

we can recognize the activities of the user correctly.

Although using HAR with ML could be useful

for the early detection of dementia, currently, there

is little available literature on this topic. Bringas et

al. (Bringas et al., 2019) proposed a method that

processes accelerometer data of Alzheimer’s patients

and a CNN that classiﬁed the stage of the disease.

They applied it in a case study with thirty-one patients

with AD, in which the classiﬁcation success rate was

ninety-one percent. Li et al. (Li et al., 2018) showed

time-aware Toeplitz inverse covariance-based cluster-

ing (Hallac et al., 2017) and CNN for predicting AD

using actigraphy data provide a solution for continu-

ously monitoring changes of physical activity of sub-

jects in daily living environments. Abnormal behav-

iors related to the stage of dementia can be detected

by monitoring daily life patterns. Arifoglu et al. (Ar-

ifoglu et al., 2020) applied graph convolutional net-

works to HAR, and abnormal behaviors related to de-

mentia were detected using activity recognition con-

ﬁdence probabilities. Using long short-term memory

(LSTM) networks, Zhan et al. (Zhan and Haddadi,

2019) proposed a system to predict patients’ activities

and timing to enable caregivers to provide timely and

appropriate care. Okada et al. (Okada et al., 2019)

used IoT sensors and mobile robots to monitor daily

activities and interactions of PwDs at home. They

proposed an ML-based estimation method, e.g., Ran-

dom Forest (RF), for the dementia stages based on

the sensor data. Gonz

alez D

ıaz et al. (Gonz

alez D

ıaz

et al., 2013) showed a support vector machine (SVM)

based method of recognizing IADL from an egocen-

tric camera view as a context for Alzheimer’s disease

research.

2.3 Gait Patterns

Gait analysis is a study of the body movements during

walking, that is, human locomotion. It is known that

mobility disorders, such as freezing of gait (FOG), are

often identiﬁed as early symptoms of AD and PDD

(Shull et al., 2014; Block et al., 2016). Mc Ardle et

al. (Mc Ardle et al., 2018; Mc Ardle et al., 2020)

proposed an application of wearable technology as a

clinical tool to differentiate the subtypes of dementia.

Xie et al. (Xie et al., 2019) showed that a sensor-

based wearable device for gait measurement might be

a convenient tool for screening cognitive impairment

called amnestic mild cognitive impairment.

To identify irregularities in gait patterns in older

persons with cognitive decline and dementia, moni-

toring their gaits using ML-driven mobile and wear-

able devices can be used to detect digital biomark-

ers of dementia. Xu et al. (Xu et al., 2018) pro-

posed an improved subsequence dynamic time warp-

Machine-learning-driven Wearable Healthcare for Dementia: A Review of Emerging Technologies and Challenges

865

ing, a pattern matching method of two time-sequence

data to detect FOG, a typical symptom of PDD.

Rodr

ıguez-Mart

ın et al. (Rodr

ıguez-Mart

ın et al.,

2017) proposed detecting FOG using SVM through

a single waist-worn triaxial accelerometer. Zhang et

al. (Zhang et al., 2020) presented an ML-based PD

diagnostic model that exploited PD pathological in-

formation from two independent accelerometers and

gyroscope records.

2.4 Eye Movements

Tracking eye movements can be a practical diagnos-

tic tool in assessing dementia (Crawford et al., 2005;

Marandi and Gazerani, 2019). Subtle impairments

in cognitive inhibition of people in the early stages

of AD can be detected using relatively simple eye-

tracking paradigms (Wilcockson et al., 2019; Carr and

Grover, 2020).

Recent progress in eyeglass-type wearable devices

has shown the potential of wearable eye-tracking for

mental health monitoring in daily life settings (Vidal

et al., 2012; Liu et al., 2019; Li et al., 2020). Pavisic et

al. (Pavisic et al., 2017) proposed an ML approach us-

ing a classiﬁcation method based on the smooth pur-

suit of raw eye-tracking data and signiﬁcant correla-

tions between eye-tracking metrics and standard vi-

sual cognitive estimates of young-onset Alzheimer’s

disease. To assess apathy in patients with AD, Chung

et al. (Chung et al., 2018) used recurrent neural net-

works to detect differences between visual scanning

behaviors on emotional and non-emotional stimuli to

classify apathetic and non-apathetic AD patients.

2.5 Linguistic Features

Dementia can affect a person’s speech, language,

and conversational interaction capabilities (Peelle and

Grossman, 2008; Colman and Bastiaanse, 2011;

Reilly et al., 2011). With the recent advances in

speech recognition systems on mobile devices, it may

be possible to record and analyze the speech of older

people to detect and assess abnormalities in their lan-

guage and conversations.

The Alzheimer’s Dementia Recognition through

Spontaneous Speech Challenge (Luz et al., 2020) pro-

vided a benchmark dataset of spontaneous speech. In

the challenge, different approaches to the automated

recognition of AD based on the dataset were com-

pared. Searle et al. (Searle et al., 2020) analyzed

spontaneous speech datasets and compared perfor-

mance across numerous classiﬁcation models of AD

and prediction of MMSE scores. They showed that an

SVM model and a “DistilBERT” model (Sanh et al.,

2020) showed good prediction performance.

Beltr

an et al. proposed an ML-based approach us-

ing the microphones of wearable sensors to detect au-

dible cues of problematic behaviors, such as tapping

and mumbling (Beltr

an et al., 2019). They classiﬁed

the audio signals based on the hidden Markov model

and SVM. Rosas et al. (Rosas et al., 2019) analyzed

the lexicon (mental dictionaries and the ability to un-

derstand complex words) and the speech ﬂuency of

PwDs. They proposed two ML algorithms to auto-

matically classify the presence/absence of dementia.

Troger et al. (Tr

oger et al., 2017) showed an ML-

based dementia screening tool trained on the French

Dem@Care corpus (Karakostas et al., 2016). Utiliz-

ing vocal features, they conﬁrmed the prediction ac-

curacy of 89%. Karlekar et al. (Karlekar et al., 2018)

proposed a CNN-LSTM model based on the Demen-

tiaBank dataset (Boller and Becker, 2005) to classify

AD. Orimaye et al. (Orimaye et al., 2018) also used

the DementiaBank dataset to realize a combination of

deep neural networks and deep language models for

classifying diseases. Several types of research (Chi-

naei et al., 2017; Pan et al., 2019; Kong et al., 2019;

Di Palo and Parde, 2019) apply ML to the Demen-

tiaBank dataset. Zhou et al. (Zhou et al., 2019) ap-

ply a natural language processing approach to extract

lifestyle habits from free-text electronic health record

data. They found that patients with AD were exposed

to more potential risk factors than the comparison

group. Such a method implemented on a smartphone

can be a novel assessment tool for estimating the risk

of dementia.

3 DISCUSSIONS

Based on the results of the previous section, we dis-

cuss the status and issues of the studies.

3.1 Use of Clinical Datasets

Although there are many studies on ML-driven mo-

bile/wearable technologies, there are still few appli-

cations to early detection of dementia. A reason is

that only a few datasets are appropriate as follows.

A dataset on oral speech during the clinical exam-

ination of dementia, called DementiaBank, is avail-

able for linguistic analysis. As the dataset has al-

ready been used for international competitions, the

latest deep learning algorithms have been applied ac-

tively for linguistic analysis. In gait analysis research,

some researchers have developed datasets of gait dis-

orders, such as PD; they have focused on the diagno-

sis and treatment of gait disorders and applied ML to

WHC 2022 - Special Session on Wearable HealthCare

866

Table 1: Summary of the studies on ML and digital biomarkers.

Analysis Sensor Method Literature

Physiological signal EEG PCA (Al-Jumeily et al., 2014)

DNN (D. Kim and K. Kim, 2018)

convolutional Boltz-

man machine

(Bi and Wang, 2019)

Human activities accelerometer CNN (Bringas et al., 2019)

actigraphy TICC&CNN (Li et al., 2018)

door&motion sensor GCN (Arifoglu et al., 2020)

activity sensor LSTM (Zhan and Haddadi, 2019)

location&interaction

sensor

RF, SVM, GBDT, etc. (Okada et al., 2019)

egocentric camera SVM (Gonz

alez D

ıaz et al., 2013)

Gait accelerometer micro features and

correlation

(Mc Ardle et al., 2018;

Mc Ardle et al., 2020; Xie

et al., 2019)

dynamic time warping (Xu et al., 2018)

SVM (Rodr

ıguez-Mart

ın et al.,

2017; Zhang et al., 2020)

Eye movement head-mounted infrared

eye tracker

Correlation&HMM (Pavisic et al., 2017)

display mounted infrared

eye tracker

RNN (Chung et al., 2018)

Linguistic analysis spontaneous speech (mi-

crophone)

SVM, GBDT, CRF

&DNN

(Searle et al., 2020)

SVM, HMM (Beltr

an et al., 2019)

3-Layer NN, SVM (Rosas et al., 2019)

SVM (Tr

oger et al., 2017)

CNN, LSTM-RNN,

CNN&LSTM-RNN

(Karlekar et al., 2018)

D2NNLM (Orimaye et al., 2018)

detecting them. Thus, it is desirable to develop clin-

ical datasets applied to other analyses, such as eye-

tracking data of PwDs, when they see their doctors.

3.2 Daily Activity Records for Datasets

HAR is the research area where ML has been most ap-

plied in mobile wearable technology. As sensors built

in a smartphone or smartwatch are designed to sense

the user behavior, it is natural to use the sensor data

for ML. HAR can be an excellent tool for the early

detection of dementia by recognizing typical behav-

ioral patterns, such as wandering and changes in the

patterns.

However, there has been little research on diag-

nosing dementia using HAR. That is because it is dif-

ﬁcult to observe and accumulate the ordinal behav-

ioral patterns of patients in the context of conven-

tional medical diagnoses. On the other hand, with the

widespread use of smartphones, it is becoming possi-

ble to sense and record the daily activities of older

people. If the data can be used, it may be possi-

ble to detect differences in behavior before and after

the onset of dementia. The daily activity records for

datasets, including continuous HAR data, need to be

developed for diagnosing dementia.

4 CHALLENGES

This section discusses future challenges to overcome

the issues described in the before section.

4.1 ML-driven Data Collection

Framework

Due to the difﬁculty in constructing datasets in daily

life, not much research has been conducted on ML-

driven wearable techniques for the early detection of

dementia. As current smartphone sensors can collect

human motions and conversations from a technical

point of view, a key issue is designing a service frame-

work that collects longitudinal data by monitoring the

user’s health status before the detection.

Machine-learning-driven Wearable Healthcare for Dementia: A Review of Emerging Technologies and Challenges

867

Because the service framework is different from

ordinal diagnosis processes based on inspection data

conducted in medical institutions, the service should

be designed from an information service point of

view. For example, large web companies already pro-

vide several ML-driven consumer services based on

user models using longitudinal data of the users. They

adjust the models by comparing predicted results with

actual results and by measuring the interventions’ ef-

fect. In addition, based on longitudinal personal data,

the service can construct more precise user models

and ﬁnd more appropriate interventions for each per-

son.

We believe that a future direction for early detec-

tion of dementia is creating a data collection frame-

work like web marketing technologies. The frame-

work should be a daily life support system that older

people use continuously. For example, it would pro-

vide speech recognition services to mediate other In-

ternet services. Moreover, it successively assesses

users’ health status based on the ML model. The

learning processes may be lifelong cycles, and the

model would evolve with the service continuously.

The data collection framework has the following

positive effects. First, the longitudinal data obtained

by the services will become the datasets to help diag-

nose the disease. High-quality longitudinal personal

data can construct a precise diagnostic and interven-

tional model for each person. Second, the detection

results can be validated with subsequent data for more

accurate future results. Third, voluntary participation

in the service may positively impact the users’ self-

healthcare. How to encourage voluntary participation

is an essential perspective considering social imple-

mentation.

4.2 Social Acceptance for Data

Collection Framework

Clinicians and patients need to understand the data

collection framework, which will help the framework

get social acceptance. Many people have a high de-

gree of conﬁdence in traditional hospital diagnoses

and prefer to receive their diagnosis in a hospital.

Therefore, making clinicians and patients like to use

the sensing data at home medically is a key issue.

An answer can be providing incentives to them to

use the framework. For example, expectations and

interests in telemedicine services are rapidly increas-

ing due to the COVID-19 pandemic (Smith et al.,

2020). There is a growing need to shift to a new med-

ical service style. Clinicians and patients stay in dif-

ferent spaces, such as their homes, to remotely pro-

vide/receive the service. As telemedicine services are

becoming more popular, home-healthcare is also be-

coming a common medical service.

It is also important to provide daily information

services on wearable devices, such as speech recog-

nition service described at the before section, to sup-

port older people in the post-onset phase of dementia.

Useful information services can be incentives for us-

ing the framework.

There are more direct incentive approaches, such

as ﬁnancial incentives for physical activity (Barte and

Wendel-Vos, 2017). Recently, implicit behavioral in-

centives such as nudges (Last et al., 2021) have also

drawn attention. The design of incentives to encour-

age users to use the ML-based healthcare framework

is a future challenge.

4.3 Personal Data and Privacy

A critical challenge for implementing the ML-driven

data collection framework is how to protect users’

privacy. Because it collects personal data from sen-

sors close to the human body (Atlam and Wills, 2020;

Kapoor et al., 2020), it may raise ethical challenges

(Maher et al., 2019; Chang et al., 2019; Burr et al.,

2020).Moreover, it is designed to focus on support-

ing individuals including those who have physical and

cognitive disabilities.

To prevent the issues associated with the current

healthcare services, some researchers have proposed

personal data management models focusing on user

privacy (Hasida, 2014; B. C. Singh et al., 2019; An-

ciaux et al., 2019). In the models, the users can man-

age their data on their own data storage, and they can

choose service providers to access their data and set

access limits.

5 CONCLUSIONS

We surveyed the literature on ML-driven wearable

technologies for early detection of dementia. We

found that the utilization and creation of datasets is an

essential issue for realizing the technologies. We de-

scribed that the datasets should be accumulated based

on an ML-driven data collection framework as a con-

tinuous healthcare service. We also discussed the is-

sues on socially acceptable implementation of the ser-

vice framework. We hope that such a data collection

framework becomes a part of future medical service

infrastructure to support users’ long-term health.

WHC 2022 - Special Session on Wearable HealthCare

868

ACKNOWLEDGEMENTS

This work was supported in part by JST CREST, JP-

MJCR18A4.

REFERENCES

Al-Jumeily, D., Iram, S., Hussain, A. J., Francois-Benois,

V., and Fergus, P. (2014). Early Detection Method of

Alzheimer’s Disease Using EEG Signals. In Huang,

D.-S., Han, K., and Gromiha, M., editors, Intelligent

Computing in Bioinformatics, Lecture Notes in Com-

puter Science, pages 25–33, Cham. Springer Interna-

tional Publishing.

Anciaux, N., Bonnet, P., Bouganim, L., Nguyen, B.,

Pucheral, P., Sandu Popa, I., and Scerri, G. (2019).

Personal Data Management Systems: The security

and functionality standpoint. Information Systems,

80:13–35.

Arifoglu, D., Charif, H. N., and Bouchachia, A. (2020). De-

tecting indicators of cognitive impairment via Graph

Convolutional Networks. Engineering Applications of

Artiﬁcial Intelligence, 89:103401.

Atlam, H. F. and Wills, G. B. (2020). IoT Security, Pri-

vacy, Safety and Ethics. In Farsi, M., Daneshkhah, A.,

Hosseinian-Far, A., and Jahankhani, H., editors, Dig-

ital Twin Technologies and Smart Cities, pages 123–

149. Springer International Publishing, Cham.

B. C. Singh, B. Carminati, and E. Ferrari (2019). Privacy-

aware Personal Data Storage (P-PDS): Learning how

to Protect User Privacy from External Applications.

IEEE Transactions on Dependable and Secure Com-

puting, pages 1–1.

Barte, J. C. M. and Wendel-Vos, G. C. W. (2017). A Sys-

tematic Review of Financial Incentives for Physical

Activity: The Effects on Physical Activity and Related

Outcomes. Behavioral Medicine, 43(2):79–90.

Beltr

an, J., Navarro, R., Ch

avez, E., Favela, J., Soto-

Mendoza, V., and Ibarra, C. (2019). Recognition of

audible disruptive behavior from people with demen-

tia. Personal and Ubiquitous Computing, 23(1):145–

157.

Bi, X. and Wang, H. (2019). Early Alzheimer’s disease

diagnosis based on EEG spectral images using deep

learning. Neural Networks, 114:119–135.

Block, V. A. J., Pitsch, E., Tahir, P., Cree, B. A. C., Allen,

D. D., and Gelfand, J. M. (2016). Remote Physical

Activity Monitoring in Neurological Disease: A Sys-

tematic Review. PLOS ONE, 11(4):e0154335.

Boller, F. and Becker, J. (2005). Dementiabank database

guide. University of Pittsburgh.

Bringas, S., Salom

on, S., Duque, R., Monta

na, J. L.,

and Lage, C. (2019). A Convolutional Neural

Network-Based Method for Human Movement Pat-

terns Classiﬁcation in Alzheimer’s Disease. Proceed-

ings, 31(1):72.

Burr, C., Taddeo, M., and Floridi, L. (2020). The Ethics of

Digital Well-Being: A Thematic Review. Science and

Engineering Ethics.

Carr, D. B. and Grover, P. (2020). The Role of Eye Tracking

Technology in Assessing Older Driver Safety. Geri-

atrics, 5(2):36.

Casson, A. J. (2019). Wearable EEG and beyond. Biomed-

ical Engineering Letters, 9(1):53–71.

Chang, V., Xu, X., Wong, B., and Mendez, V. (2019). Eth-

ical Problems of Smart Wearable Devices. COM-

PLEXIS, page 9.

Chinaei, H., Currie, L. C., Danks, A., Lin, H., Mehta, T.,

and Rudzicz, F. (2017). Identifying and Avoiding

Confusion in Dialogue with People with Alzheimer’s

Disease. Computational Linguistics, 43(2):377–406.

Chung, J., Chau, S. A., Herrmann, N., Lanct

ot, K. L.,

and Eizenman, M. (2018). Detection of Apathy in

Alzheimer Patients by Analysing Visual Scanning Be-

haviour with RNNs. In Proceedings of the 24th

ACM SIGKDD International Conference on Knowl-

edge Discovery & Data Mining, pages 149–157, Lon-

don United Kingdom. ACM.

Colman, K. and Bastiaanse, R. (2011). Language Process-

ing in Parkinson’s Disease Patients Without Dementia.

In Dushanova, J., editor, Diagnostics and Rehabilita-

tion of Parkinson’s Disease. InTech.

Cordell, C. B., Borson, S., Boustani, M., Chodosh, J.,

Reuben, D., Verghese, J., Thies, W., Fried, L. B., and

Medicare Detection of Cognitive Impairment Work-

group (2013). Alzheimer’s Association recommen-

dations for operationalizing the detection of cogni-

tive impairment during the Medicare Annual Wellness

Visit in a primary care setting. Alzheimer’s & Demen-

tia, 9(2):141–150.

Crawford, T. J., Higham, S., Renvoize, T., Patel, J., Dale,

M., Suriya, A., and Tetley, S. (2005). Inhibitory con-

trol of saccadic eye movements and cognitive impair-

ment in Alzheimer’s disease. Biological Psychiatry,

57(9):1052–1060.

D. Kim and K. Kim (2018). Detection of Early Stage

Alzheimer’s Disease using EEG Relative Power with

Deep Neural Network. In 2018 40th Annual In-

ternational Conference of the IEEE Engineering in

Medicine and Biology Society (EMBC), pages 352–

355.

Di Palo, F. and Parde, N. (2019). Enriching Neural Mod-

els with Targeted Features for Dementia Detection.

arXiv:1906.05483 [cs].

Giebel, C. and Challis, D. (2016). Sensitivity of the Mini-

Mental State Examination, Montreal Cognitive As-

sessment and the Addenbrooke’s Cognitive Examina-

tion III to everyday activity impairments in dementia:

An exploratory study. International Journal of Geri-

atric Psychiatry, 32.

Gonz

alez D

ıaz, I., Buso, V., Benois-Pineau, J., Bourmaud,

G., and Megret, R. (2013). Modeling instrumental

activities of daily living in egocentric vision as se-

quences of active objects and context for alzheimer

disease research. In Proceedings of the 1st ACM Inter-

national Workshop on Multimedia Indexing and Infor-

mation Retrieval for Healthcare - MIIRH ’13, pages

11–14, Barcelona, Spain. ACM Press.

Hallac, D., Vare, S., Boyd, S., and Leskovec, J. (2017).

Toeplitz Inverse Covariance-Based Clustering of Mul-

tivariate Time Series Data. In Proceedings of the 23rd

Machine-learning-driven Wearable Healthcare for Dementia: A Review of Emerging Technologies and Challenges

869

ACM SIGKDD International Conference on Knowl-

edge Discovery and Data Mining, pages 215–223,

Halifax NS Canada. ACM.

Hasida, K. (2014). Personal life repository as a distributed

PDS and its dissemination strategy for healthcare ser-

vices. In AAAI Spring Symposia.

Husebo, B. S., Heintz, H. L., Berge, L. I., Owoyemi, P.,

Rahman, A. T., and Vahia, I. V. (2020). Sensing

Technology to Monitor Behavioral and Psychological

Symptoms and to Assess Treatment Response in Peo-

ple With Dementia. A Systematic Review. Frontiers

in Pharmacology, 10:1699.

Kapoor, V., Singh, R., Reddy, R., and Churi, P. (2020). Pri-

vacy Issues in Wearable Technology: An Intrinsic Re-

view. SSRN Electronic Journal.

Karakostas, A., Briassouli, A., Avgerinakis, K., Kompat-

siaris, I., and Tsolaki, M. (2016). The Dem@Care

experiments and datasets: A technical report.

Karlekar, S., Niu, T., and Bansal, M. (2018). Detecting

Linguistic Characteristics of Alzheimer’s Dementia

by Interpreting Neural Models. In Proceedings of

the 2018 Conference of the North American Chap-

ter of the Association for Computational Linguistics:

Human Language Technologies, Volume 2 (Short Pa-

pers), pages 701–707, New Orleans, Louisiana. Asso-

ciation for Computational Linguistics.

Kelly, M. E., Duff, H., Kelly, S., McHugh Power, J. E.,

Brennan, S., Lawlor, B. A., and Loughrey, D. G.

(2017). The impact of social activities, social net-

works, social support and social relationships on the

cognitive functioning of healthy older adults: A sys-

tematic review. Systematic Reviews, 6(1):259.

Kong, W., Jang, H., and Carenini, G. (2019). A Neural

Model for Predicting Dementia from Language. In

Machine Learning for Healthcare Conference., pages

270–286.

Kourtis, L. C., Regele, O. B., Wright, J. M., and Jones, G. B.

(2019). Digital biomarkers for Alzheimer’s disease:

The mobile/wearable devices opportunity. npj Digital

Medicine, 2(1):9.

Last, B. S., Buttenheim, A. M., Timon, C. E., Mitra, N., and

Beidas, R. S. (2021). Systematic review of clinician-

directed nudges in healthcare contexts. BMJ Open,

11(7):e048801.

Li, J., Li, H., Umer, W., Wang, H., Xing, X., Zhao, S.,

and Hou, J. (2020). Identiﬁcation and classiﬁcation

of construction equipment operators’ mental fatigue

using wearable eye-tracking technology. Automation

in Construction, 109:103000.

Li, J., Rong, Y., Meng, H., Lu, Z., Kwok, T., and Cheng,

H. (2018). TATC: Predicting Alzheimer’s Disease

with Actigraphy Data. In Proceedings of the 24th

ACM SIGKDD International Conference on Knowl-

edge Discovery & Data Mining, pages 509–518, Lon-

don United Kingdom. ACM.

Lindbergh, C. A., Dishman, R. K., and Miller, L. S. (2016).

Functional Disability in Mild Cognitive Impairment:

A Systematic Review and Meta-Analysis. Neuropsy-

chology Review, 26(2):129–159.

Liu, Z. X., Liu, Y., and Gao, X. (2019). Using Mobile Eye

Tracking to Evaluate the Satisfaction with Service Of-

ﬁce. In Marcus, A. and Wang, W., editors, Design,

User Experience, and Usability. Practice and Case

Studies, pages 183–195, Cham. Springer International

Publishing.

Luz, S., Haider, F., de la Fuente, S., Fromm, D.,

and MacWhinney, B. (2020). Alzheimer’s Demen-

tia Recognition through Spontaneous Speech: The

ADReSS Challenge. arXiv:2004.06833 [cs, eess,

stat].

Maher, N. A., Senders, J. T., Hulsbergen, A. F., Lamba, N.,

Parker, M., Onnela, J.-P., Bredenoord, A. L., Smith,

T. R., and Broekman, M. L. (2019). Passive data col-

lection and use in healthcare: A systematic review of

ethical issues. International Journal of Medical Infor-

matics, 129:242–247.

Marandi, R. Z. and Gazerani, P. (2019). Aging and eye

tracking: In the quest for objective biomarkers. Future

Neurology, 14(4):FNL33.

Mc Ardle, R., Del Din, S., Galna, B., Thomas, A., and

Rochester, L. (2020). Differentiating dementia dis-

ease subtypes with gait analysis: Feasibility of wear-

able sensors? Gait & Posture, 76:372–376.

Mc Ardle, R., Morris, R., Hickey, A., Del Din, S., Koychev,

I., Gunn, R. N., Lawson, J., Zamboni, G., Ridha, B.,

Sahakian, B., Rowe, J., Thomas, A., Zetterberg, H.,

MacKay, C., Lovestone, S., Rochesteron, L., Team (,

D., and Study, F. P. (2018). Gait in Mild Alzheimer’s

Disease: Feasibility of Multi-Center Measurement in

the Clinic and Home with Body-Worn Sensors: A Pi-

lot Study.

Najar, J.,

Ostling, S., Gudmundsson, P., Sundh, V., Johans-

son, L., Kern, S., Guo, X., H

allstr

om, T., and Skoog, I.

(2019). Cognitive and physical activity and dementia:

A 44-year longitudinal population study of women.

Neurology, 92(12):e1322–e1330.

Okada, S., Inoue, K., Imai, T., Noguchi, M., and Kuwa-

mura, K. (2019). Dementia Scale Classiﬁcation Based

on Ubiquitous Daily Activity and Interaction Sensing.

8th International Conference on Affective Computing

and Intelligent Interaction (ACII), page 7.

Orimaye, S. O., Wong, J. S.-M., and Wong, C. P. (2018).

Deep language space neural network for classifying

mild cognitive impairment and Alzheimer-type de-

mentia. PLOS ONE, 13(11):e0205636.

Pan, Y., Mirheidari, B., Reuber, M., Venneri, A., Blackburn,

D., and Christensen, H. (2019). Automatic Hierarchi-

cal Attention Neural Network for Detecting AD. In

Interspeech 2019, pages 4105–4109. ISCA.

Pavisic, I. M., Firth, N. C., Parsons, S., Rego, D. M., Shake-

speare, T. J., Yong, K. X. X., Slattery, C. F., Pa-

terson, R. W., Foulkes, A. J. M., Macpherson, K.,

Carton, A. M., Alexander, D. C., Shawe-Taylor, J.,

Fox, N. C., Schott, J. M., Crutch, S. J., and Prima-

tivo, S. (2017). Eyetracking Metrics in Young Onset

Alzheimer’s Disease: A Window into Cognitive Vi-

sual Functions. Frontiers in Neurology, 8:377.

Peelle, J. E. and Grossman, M. (2008). Language Pro-

cessing in Frontotemporal Dementia: A Brief Review.

Language and Linguistics Compass, 2(1):18–35.

Piau, A., Wild, K., Mattek, N., and Kaye, J. (2019). Cur-

rent State of Digital Biomarker Technologies for Real-

Life, Home-Based Monitoring of Cognitive Function

for Mild Cognitive Impairment to Mild Alzheimer

WHC 2022 - Special Session on Wearable HealthCare

870

Disease and Implications for Clinical Care: System-

atic Review. Journal of Medical Internet Research,

21(8).

Reilly, J., Troche, J., and Grossman, M. (2011). Lan-

guage processing in dementia. In The Handbook of

Alzheimer’s Disease and Other Dementias, volume 7.

Wiley Online Library.

Rim, B., Sung, N.-J., Min, S., and Hong, M. (2020). Deep

Learning in Physiological Signal Data: A Survey.

Sensors, 20(4):969.

Rodr

ıguez-Mart

ın, D., Sam

a, A., P

erez-L

opez, C., Catal

A., Moreno Arostegui, J. M., Cabestany, J., Bay

es,

A., Alcaine, S., Mestre, B., Prats, A., Crespo, M. C.,

Counihan, T. J., Browne, P., Quinlan, L. R.,

OLaighin,

G., Sweeney, D., Lewy, H., Azuri, J., Vainstein,

G., Annicchiarico, R., Costa, A., and Rodr

ıguez-

Molinero, A. (2017). Home detection of freezing

of gait using support vector machines through a sin-

gle waist-worn triaxial accelerometer. PLOS ONE,

12(2):e0171764.

Rosas, D. S., Arriaga, S. T., and Fernandez, M. A. A.

(2019). Search for Dementia Patterns in Transcribed

Conversations using Natural Language Processing. In

2019 16th International Conference on Electrical En-

gineering, Computing Science and Automatic Control

(CCE), pages 1–6, Mexico City, Mexico. IEEE.

Sacco, G., Joumier, Darmon, Dechamps, Derreumaux, Lee,

Piano, Bordone, Konig, Teboul, David, Guerin, Bre-

mond, and Robert (2012). Detection of activities

of daily living impairment in Alzheimer’s disease

and mild cognitive impairment using information and

communication technology. Clinical Interventions in

Aging, page 539.

Sanh, V., Debut, L., Chaumond, J., and Wolf, T. (2020).

DistilBERT, a distilled version of BERT: Smaller,

faster, cheaper and lighter. arXiv:1910.01108 [cs].

Searle, T., Ibrahim, Z., and Dobson, R. (2020). Com-

paring Natural Language Processing Techniques for

Alzheimer’s Dementia Prediction in Spontaneous

Speech. arXiv:2006.07358 [cs, eess, stat].

Shull, P. B., Jirattigalachote, W., Hunt, M. A., Cutkosky,

M. R., and Delp, S. L. (2014). Quantiﬁed self and

human movement: A review on the clinical impact of

wearable sensing and feedback for gait analysis and

intervention. Gait & Posture, 40(1):11–19.

Smith, A. C., Thomas, E., Snoswell, C. L., Haydon, H.,

Mehrotra, A., Clemensen, J., and Caffery, L. J. (2020).

Telehealth for global emergencies: Implications for

coronavirus disease 2019 (COVID-19). Journal of

Telemedicine and Telecare, 26(5):309–313.

Tanveer, M., Richhariya, B., Khan, R. U., Rashid, A. H.,

Khanna, P., Prasad, M., and Lin, C. T. (2020).

Machine Learning Techniques for the Diagnosis of

Alzheimer’s Disease: A Review. ACM Transactions

on Multimedia Computing, Communications, and Ap-

plications, 16(1s):1–35.

Thabtah, F., Peebles, D., Retzler, J., and Hathurusingha,

C. (2020). Dementia medical screening using mo-

bile applications: A systematic review with a new

mapping model. Journal of biomedical informatics,

111:103573.

oger, J., Linz, N., Alexandersson, J., K

onig, A., and

Robert, P. (2017). Automated speech-based screening

for alzheimer’s disease in a care service scenario. In

Proceedings of the 11th EAI International Conference

on Pervasive Computing Technologies for Healthcare

- PervasiveHealth ’17, pages 292–297, Barcelona,

Spain. ACM Press.

Tsang, G., Xie, X., and Zhou, S.-M. (2020). Harnessing the

Power of Machine Learning in Dementia Informat-

ics Research: Issues, Opportunities, and Challenges.

IEEE Reviews in Biomedical Engineering, 13:113–

129.

Turner, R. S., Stubbs, T., Davies, D. A., and Albensi, B. C.

(2020). Potential New Approaches for Diagnosis of

Alzheimer’s Disease and Related Dementias. Fron-

tiers in Neurology, 11:496.

van der Linde, R. M., Dening, T., Stephan, B. C. M., Prina,

A. M., Evans, E., and Brayne, C. (2016). Longitudinal

course of behavioural and psychological symptoms of

dementia: Systematic review. British Journal of Psy-

chiatry, 209(5):366–377.

Vidal, M., Turner, J., Bulling, A., and Gellersen, H. (2012).

Wearable eye tracking for mental health monitoring.

Computer Communications, 35(11):1306–1311.

Wilcockson, T. D., Mardanbegi, D., Xia, B., Taylor, S.,

Sawyer, P., Gellersen, H. W., Leroi, I., Killick, R.,

and Crawford, T. J. (2019). Abnormalities of saccadic

eye movements in dementia due to Alzheimer’s dis-

ease and mild cognitive impairment. Aging, 11(15).

Xie, H., Wang, Y., Tao, S., Huang, S., Zhang, C., and Lv,

Z. (2019). Wearable Sensor-Based Daily Life Walk-

ing Assessment of Gait for Distinguishing Individuals

With Amnestic Mild Cognitive Impairment. Frontiers

in Aging Neuroscience, 11:285.

Xu, C., He, J., Zhang, X., Wang, C., and Duan, S. (2018).

Template-Matching-Based Detection of Freezing of

Gait Using Wearable Sensors. Procedia Computer

Science, 129:21–27.

Zhan, Y. and Haddadi, H. (2019). Activity prediction for

improving well-being of both the elderly and care-

givers. In Proceedings of the 2019 ACM Interna-

tional Joint Conference on Pervasive and Ubiquitous

Computing and Proceedings of the 2019 ACM Inter-

national Symposium on Wearable Computers - Ubi-

Comp/ISWC ’19, pages 1214–1217, London, United

Kingdom. ACM Press.

Zhang, H., Deng, K., Li, H., Albin, R. L., and Guan, Y.

(2020). Deep Learning Identiﬁes Digital Biomark-

ers for Self-Reported Parkinson’s Disease. Patterns,

1(3):100042.

Zhou, X., Wang, Y., Sohn, S., Therneau, T. M., Liu, H.,

and Knopman, D. S. (2019). Automatic extraction and

assessment of lifestyle exposures for Alzheimer’s dis-

ease using natural language processing. International

Journal of Medical Informatics, 130:103943.

Machine-learning-driven Wearable Healthcare for Dementia: A Review of Emerging Technologies and Challenges

871