AI-Rehab: A Framework for AI Driven Neurorehabilitation

Training - The Proﬁling Challenge

Yuri Almeida

1,2

, Manisha Sanjay Sirsat

2 b

, Sergi Berm´udez i Badia

1,3 c

and Eduardo Ferm´e

1,2 d

Universidade da Madeira, Portugal

NOVA-LINCS, Portugal

Madeira Interactive Technologies Institute, Portugal

Keywords:

Long Term Care in Cognitive Neurorehabilitation, Proﬁling Challenges, Machine Learning, Belief Revision.

Abstract:

One of the health clinic challenges is rehabilitation therapy cognitive impairment that can happen after brain

injury, dementia and in normal cognitive decline due to aging. Current cognitive rehabilitation therapy has

been shown to be the most effective way to address this problem. However, a) it is not adaptive for every

patient, b) it has a high cost, and c) it is usually implemented in clinical environments. The Task Generator

(TG) is a free tool for the generation of cognitive training tasks. However, TG is not designed to adapt and

monitor the cognitive progress of the patient. Hence, we propose in the BRaNT project an enhancement of

TG with belief revision and machine learning techniques, gamiﬁcation and remote monitoring capabilities to

enable health professionals to provide a long-term personalized cognitive rehabilitation therapy at home. The

BRaNT is an interdisciplinary effort that addresses scientiﬁc limitations of current practices as well as provides

solutions towards the sustainability of health systems and contributes towards the improvement of quality of

life of patients. This paper proposes the AI-Rehab framework for the BRaNT, explains proﬁling challenge in

the situation of insufﬁcient data and presents an alternate AI solutions which might be applicable once enough

data is available.

1 INTRODUCTION

Cognitive deﬁcits are common after brain injury, de-

mentia and normally due to aging. These impact the

performance of activities of daily living and limit peo-

ple’s independence, with a high monetary and societal

cost. Moreover, many cognitive rehabilitation tools

lack ﬂexibility to be adapted to each patient needs.

Cognitive rehabilitation has been shown to be the

most effective way to address this problem. However,

current rehabilitation has some limitations:

• Rehabilitation tools are not adaptive and may not

be adequate for every patient. Hence, a sub-

optimal set of exercises limits the impact of re-

habilitation and reduces engagement in rehabili-

tation;

• Interventions are time consuming and have a high

cost, and are usually implemented in clinical envi-

https://orcid.org/0000-0003-3213-7514

https://orcid.org/0000-0002-5696-3602

https://orcid.org/0000-0003-4452-0414

https://orcid.org/0000-0002-9618-2421

ronments. Thus, the lack of monetary and human

resources prevent public health systems to imple-

ment ideal long-term rehabilitation.

Neuropsychological assessments (NPAs) are use-

ful instruments to evaluate cognitive skills. How-

ever, current cognitive interventions are sub-optimal

in terms of lack of adaptability according to the NPAs

results (Williams and Sims, 2000; Parsons, 2015).

Besides, the tools are not ﬂexible enough to cover the

needs of all patients and traditional therapy has a high

cost both on human and monetary resources (Solana

et al., 2014). Here we propose bridging NPAs and

computational modelling to deliver a highly personal-

ized tool that allows for the creation of interventions

through Information and Communication Technolo-

gies (ICT). A tool that could generate validated cog-

nitive training tasks, parameterizedper patients needs,

which intelligently adjusts difﬁculty over time, mon-

itors changes and communicates with the patient’s

healthcare team. We start by creating an extension

and enhancement of an existing tool, the Task Gen-

erator (TG

) (Faria and Berm´udez i Badia, 2015), to

http://neurorehabilitation.m-iti.org/TaskGenerator/

Almeida, Y., Sirsat, M., Bermúdez i Badia, S. and Fermé, E.

AI-Rehab: A Framework for AI Driven Neurorehabilitation Training - The Proﬁling Challenge.

DOI: 10.5220/0009369108450853

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 5: HEALTHINF, pages 845-853

ISBN: 978-989-758-398-8; ISSN: 2184-4305

845

generate the appropriatecognitive training tasks using

the patient proﬁle as an input. The TG is a free web

based tool for the procedural generation of cognitive

training tasks. TG tasks are parametrized in terms

of training difﬁculty and cognitive demands, but are

not designed to monitor the evolution of the patient,

which is essential to adjust training for the patient’s

evolution. In the BRaNT project, we are develop-

ing an extension of TG with extended and improved

training content. This project uses two artiﬁcial in-

telligence modules (based on belief revision and ma-

chine learning techniques), gamiﬁcation and remote

monitoringcapabilities to enable Health Professionals

(HP) to provide long-term personalized cognitive re-

habilitation therapy at home. AI-Rehab is the frame-

work behind the personalization and adaptation of

training, and relies on the results of validated NPA

used to evaluate patient and create a computational

Cognitive Proﬁle (CP) of the patient. By comparing

this proﬁle with the Normative Data (ND) accord-

ingly to the Socio-Demographic Information (SDI)

of the patient, we infer the patient’s Cognitive Sta-

tus (CS). After this, the HP can decide on a set of

training tasks for that patient. The patient trains at

home and the performance is communicated to Be-

lief Revision Engine (BR-E), which assesses evolu-

tion of the patient and manages task personalization

over time. The proposed system will be able to con-

tact the HP if something unexpected happens. This

framework tackles three important Artiﬁcial Intelli-

gence (AI) challenges: patient proﬁling, task selec-

tion and parameterization, and adaptive training. In

this paper we propose a solution for the patient pro-

ﬁling challenge. In the Section 2, we review some

literature about assessment of cognitive impairment

and AI in cognitive rehabilitation ﬁeld whereas, Sec-

tion 3 gives detailed explanation on each step of pro-

posed AI-Rehab framework. Section 4 describes the

patient’s proﬁling challenge and ﬁnally, its AI solu-

tion is reported in the Section 4.1, including task set-

ting and iterative training.

2 STATE OF THE ART

Cognitive impairments following brain injury are

common, and are present in approximately 70% of

patients in the acute stages of recovery (Morris et al.,

2012), causing problems in activities of daily life and

social participation. Stroke commonly includes focal

disorders such as aphasia and neglect, and more dif-

fuse abnormalities such as slowed information pro-

cessing and executive dysfunction (Cumming et al.,

2013). Cognitive rehabilitation is designed to re-

store, substitute, or compensate for the lost of cog-

nitive abilities, and is the treatment of choice for

these deﬁcits (Bott and Kramer, 2017). The Amer-

ican Congress of Rehabilitation Medicine conducted

systematic reviews on a total of 370 cognitive rehabil-

itation studies for people with acquired brain injury,

published from 1971-2008 (Cicerone et al., 2000; Ci-

cerone et al., 2005; Cicerone et al., 2011). Cogni-

tive rehabilitation was shown to be of greater beneﬁt

than conventional rehabilitation in 94.1% of the stud-

ies. Thus, cognitive rehabilitation is the best avail-

able form of treatment for people with neurocognitive

impairment (Cicerone et al., 2011). Unfortunately,

the efﬁcacy of cognitive training highly depends on

the intensity of treatment over an extended period of

time. The traditional intervention model is very time

consuming for teams to manage personalized rehabil-

itation programs; patients move to the clinical center,

making the duration of the treatment conditional to

patient’s availability; interventions are subject to the

availability of vacancies and transportation (Solana

et al., 2014). This results in a very high cost, com-

promising sustainability, accessibility and scalability,

resulting in a large economic burden to both the health

system and families (Carod-Artal et al., 2000).

The ICT based solutions such as gaming, virtual

reality or computer simulations have been shown to

have an enormous potential for enhancing cognitive

rehabilitation by supporting the ability to carry out

controlled and highly adaptive valid tasks (Berm´udez

i Badia et al., 2016). Over the past few years, sev-

eral computer based solutions have been proposed

to increase the availability and quality of cognitive

training, ﬂooding the marketplace with commercial

brain exercise programs that claim to improve cogni-

tion and have diagnostic abilities (George and White-

house, 2011) such as the CogWeb and the Guttmann

Neuro Personal Trainer. Through these platforms it is

possible to deliver a training program to a patient, an-

alyze results and transfer them to the Hospital Infor-

mation System. Nevertheless, none of these tools ad-

dresses multiple domains of cognitive functioning in

a systematic and quantitative manner relying on vali-

dated NPAs.

2.1 Assessment of Cognitive

Impairment

The Montreal Cognitive Assessment (MoCA)

(Nasreddine et al., 2005) is a well-known test which

is the recommended instrument in Portugal for global

cognitive screening measurement. It addresses (i)

short-term memory, (ii) executive functions, (iii)

visuospatial abilities, (iv) language, (v) attention,

Cognitive Health IT 2020 - Special Session on Machine Learning and Deep Learning Improve Preventive and Personalized Healthcare

846

Figure 1: The AI-Rehab framework for the BRaNT project.

concentration and working memory and (vi) temporal

and spatial orientation (Freitas et al., 2012c). MoCA

has been the subject of systematic research within the

Portuguese population and validation studies were

conducted on speciﬁc clinical groups like Mild Cog-

nitive Impairment and Alzheimer’s Disease (Freitas

et al., 2013), Frontotemporal Dementia (Freitas et al.,

2012a), Vascular Dementia (Freitas et al., 2012b) and

Multiple Sclerosis (Freitas et al., 2018). Many studies

emphasize the psychometric characteristics of the test

(e.g. (Freitas et al., 2012c; Freitas et al., 2015; Freitas

et al., 2014), with norms for Portuguese population

(Freitas et al., 2011). Unfortunately, as a screening

instrument it can detect deﬁcits but not quantify them

accurately. That can only be achieved with domain

speciﬁc NPAs. There are some NPAs such as Free

and Cued Selective Reminding Test, Semantic and

phonemic verbal ﬂuency, Rey-Osterrieth Complex

Figure Test, clinical evaluation of dementia, etc. The

AI-Rehab is designed to receive information from

any NPA.

2.2 Artiﬁcial Intelligence in Cognitive

Rehabilitation

This section review the list of pilot studies of AI for

cognitive impairment and rehab. AI aims to bring

high precision in healthcare by employing computa-

tional intelligence in clinical tasks. Nowadays, the

most popular ML algorithms for structured data are

support vector machine, neural network (NN) and

deep learning (DL) whereas for unstructured data is

natural language processing (Jiang et al., 2017). Sev-

eral statistical learning (Hastie et al., 2009), ML clas-

siﬁcation (Fern´andez-Delgado et al., 2014) and re-

gression (Fern´andez-Delgado et al., 2018) and DL

techniques (LeCun et al., 2015) are available, by us-

ing them a few AI models are developed and these

really contribute for advancement of cognitive ﬁeld.

For instance, the study (Chi et al., 2017) developed

personalized long-term and follow-up models to pre-

dict CS. First is sequential estimation of risk factors

to predict how cognition will change over long time

and second is observationof time-varyingrisk factors.

Likewise, (Ko et al., 2019) developed adaptive Least

Absolute Shrinkage and Selection Operator (LASSO)

model to identify signiﬁcant predictors of multivari-

ate NPAs and demographic variables for prediction of

cerebral amyloid beta abnormal level of status. Mem-

ory dysfunction is a crucial cognitive factor for early

detection of disease and one of the instruments to cal-

culate it is M-CRT. The binary classiﬁcation (Berg-

eron et al., 2019) of cognitive health status (healthy or

unhealthy) and health related question (yes or no) is

modelled logistic regression using demographic data

and M-CRT test score.

Belief revision systems are logical frameworks

for modelling how agents modify their beliefs when

they receivenew information (sometimes inconsistent

with the previous beliefs)

. To integrate the new in-

formation, the agent will have to give up some in-

formation while preserving as much of the old in-

formation as possible. The AGM-framework (Al-

chourr´on et al., 1985) is the most popular framework

to guide the change of belief. The AGM model has ac-

quired the status of a standard model (for an overview,

For the sake of simplicity, we will assume that beliefs,

knowledge and data have the same status

AI-Rehab: A Framework for AI Driven Neurorehabilitation Training - The Proﬁling Challenge

847

see (Ferm´e and Hansson, 2011; Ferm´e and Hans-

son, 2018)). Several algorithms for the implementa-

tion of belief change operations were proposed. Most

of them were constructed to recognize which beliefs

are supported and how, and to perform changes while

minimizing the number of sentences to be changed,

thus preserving the maximum amount of the previ-

ous knowledge. Implementation also include pro-

posals by (Katsuno and Mendelzon, 1991; Williams

and Sims, 2000; Delgrande and Schaub, 2003; Pep-

pas and Williams, 2016). Katsuno and Mendelzon

(Katsuno and Mendelzon, 1991) also focused a lot

of their paper on trying to deﬁne a notion of dis-

tance between Knowledge Bases. This will be useful

later when comparing the CP at different stages. A

core aspect in implementation is the space and time

required for computation. Jin and Thielscher pro-

posed Reinforcement Belief Revision that combines

two desiderata for belief change implementations: It

satisﬁes the standard rationality postulates, and the

time and space required for its implementation can

be assessed. Recently, new studies of implementa-

tion belief revision by Horn Clauses have been initi-

ated (Booth et al., 2010; Delgrande, 2008; Delgrande

and Wassermann, 2010). Pagnucco (Pagnucco, 2006)

and (Zhuang et al., 2007) formalised a way of im-

plementing AGM operations using a knowledge com-

pilation technique involving prime implicates in or-

der to improve computational efﬁciency. The study

(Schwind et al., 2019) has proposed a change for-

mula, that given two know bases (or the same knowl-

edge base at different times) it is possible to determine

what caused the change in the belief set. Belief Re-

vision is still a relative new area of investigation and

there are not many examples we can draw from, so we

hope to contribute on this area.

3 FRAMEWORK

The BRaNT project is an interdisciplinary effort to

create a new set of ICT for rehabilitation at home.

For this project, we propose the AI-Rehab framework,

which includes the steps shown in Figure 1. There

are clearly three distinct challenges in the framework.

First and foremost we need to ﬁnd an optimal way

to consolidate whole data from all the different NPAs

into a consistent proﬁle, that should be easily inter-

preted by HP. This is the Proﬁling Challenge. The

second challenge refers to the task selection and set-

tings deﬁnition. How will the system guarantee that

the tasks are always optimally conﬁgured for a partic-

ular patient’s proﬁle? To achieve this, we will create

a Belief Revision Engine (BR-E) to generate a pre-

dictable proﬁle given a set of difﬁculty and follow the

ﬂow theory to keep the patient always engaged with

the training task in hand (Nakamura and Csikszentmi-

halyi, 2009). Finally, the third challenge is the adap-

tive training at home. The goal of this challenge is to

keep adapting the predicted proﬁle and the game set-

tings to keep the patient in ﬂow, all at the comfort of

his home, without updated NPAs data. The process of

the AI-Rehab framework (Figure 1) can be explained

as follows:

1. HPs use validated NPAs to assess the patient;

2. The result of NPAs data are then injected in a

database and mapped to values by applying equa-

tion 1 from section 4.1, which creates a normal-

ized patient’s CP where each factor (or cognitive

domain) is normalized from 0 to 100%;

3. Once the CP is created, it is compared with Nor-

mative Data (ND) to deﬁne the CS, so called

Identification, and for that it is necessary to in-

corporate new information of the patient (SDI)

along with ND;

4. The HP can then interpret the CS and specify

training objectives (For instance, A, B and C with

their percentages, shown in the Figure 1) for that

particular proﬁle, where the set of number of pa-

rameters of the training tasks are suggested. Si-

multaneously, the initial difﬁculty of the training

is set based on the CP. At this time the BR-E tries

to predict the expected results of the training task

(ERTT);

5. The patient performs a gamiﬁed training task at

home and performance is communicated to the

BR-E. This compares the results of training with

ERTT and identiﬁes:

(a) If patient performance is in the accordance to

expectation,

(b) Data shows patient evolution or involution, the

patient proﬁle is updated and a new difﬁculty is

set,

data is detected, and the HP is contacted to re-

assess the situation and train BR-E.

6. Finally, the HP will do a new NPA reassessment,

after intervention is completed to quantify the im-

pact of cognitive training with AI-Rehab and de-

cide if more therapy is required.

Hence, we aim at combining the advantages of ICT

with a participatory design approach involving health

professionals (such as rehabilitation physicians, ther-

apists and neuropsychologists) to develop a novel

portable tool for the generation of cognitive rehabili-

tation training for the home use (Paulino et al., 2018;

Cognitive Health IT 2020 - Special Session on Machine Learning and Deep Learning Improve Preventive and Personalized Healthcare

848

Paulino et al., 2019). This tool will be a free and

worldwide accessible for clinicians, able to generate

patient’s proﬁle and personalized cognitive rehabili-

tation programs in digital form instead of paper-and-

pencil format. It will be composed by a set of stan-

dardized rehabilitation tasks gathered from clinical

settings and parameterized through a participatory de-

sign approach and will be able to procedurally gener-

ate a large number of tasks by specifying the values of

their intrinsic parameters. It addresses the following

task types as described below (Faria et al., 2018):

• Knowledge: Memory of Stories; Cancellation;

Questions of General Knowledge; Find Loca-

tions; Image Pairs.

• Comprehension: Differences between Similar

Scenarios; Categorization; Synonymous and

Antonyms; Association.

• Application: Mazes; Problem Resolution; Tan-

gram; Numeric Sequences; Navigation.

• Analysis: Action Sequencing; Visual Memory;

Puzzles; Word Search.

• Evaluation: Differentiation between Coherent and

Incoherent situations; Comprehension of Con-

texts.

These tasks will be implemented in an interactive dig-

ital environment, shaped as tasks through a gami-

ﬁcation approach to deliver an immediate feedback

and reinforcement on progress (Wilson and McDon-

agh, 2013), which is an important element to increase

the motivation and avoid dropouts. Besides, adaptive

training is the last challenge of the AI-Rehab frame-

work. After each iteration of the task training, the BR-

E will calculate a new temporary proﬁle and cognitive

situation. These will be compared with the ERTT cal-

culated before to see if any adjustment needs to be

done. The authors (Katsuno and Mendelzon, 1991)

deﬁned on their paper several operators that can be

used to calculate the differences between knowledge

bases. With this difference calculated, we can deter-

mine how that patient evolved during the previous it-

eration and we are able to use this information to ad-

just the parameters of the task. In this paper we will

only focus on the computational patient proﬁling.

4 PATIENT PROFILING

CHALLENGE

This step starts with the task of speciﬁcation the rel-

evant NPAs for a comprehensive evaluation of cogni-

tion. Screening tests are brief assessment triage tools

to identify patients cognitively at-risk that require a

fuller evaluation. Based on information from the clin-

ical process, which includes medical data, interview

data and scores on cognitive screening tests (such as

MoCA), the patient examined may be referred for a

more comprehensive NPA aimed at rehabilitation ob-

jectives. The NPA, instead, is a standard part of inte-

grated medical and psychological care, and is a nec-

essary step to implement and further evaluate rehabil-

itation procedures. The NPA has several objectives,

namely: (i) to identify and characterize cognitiveabil-

ities and activities of daily life, personality, emotional

functioning and behaviours of the person and to deﬁne

changes in these domains in comparison to the level

of premorbid functioning; (ii) to document and quan-

tify the nature and severity of cognitive and functional

deﬁcits, symptoms and signs present, potentially as-

sociated with pathology, in the context of examining

the structural and functional integrity of brain func-

tioning, differentiating what is the associated decline

to the normal aging and deﬁcits associated with cere-

bral dysfunction / neurodegenerative pathology; (iii)

to deﬁne a baseline in various domains of the cog-

nitive, emotional and behavioural functioning, which

can be examined in a longitudinal register, through

repeated evaluations, thus enabling monitoring of the

clinical evolution of the person, recovery of functions,

response to the intervention (e.g., rehabilitation, psy-

chotherapy, pharmacological) or the progression of

the disease; (iv) to identify personal resources and

preserved functions that are equally useful for plan-

ning and implementing rehabilitative/ therapeuticand

preventive intervention procedures, as well as evalu-

ating their effectiveness, with the aim of promoting

the person’s well-being and quality of life. Further

step is devoted to the factorization of the outcome

of these assessments and translation to a formal lan-

guage that allows creating automatically (by artiﬁcial

intelligence techniques) a proﬁle of the patient that

can be used for deﬁning a neurorehabilitation ther-

apy. This will give us the CP, shown in the Figure

1. The Identification part is more complicated. We

need to create a CS using the ND and the SDI, in or-

der to interpret how a particular patient CP compares

to the ND. This will produce the CS in the Figure 1,

which represents, in percentage, where the CP is lo-

cated compared to the rest of the population with sim-

ilar SDI. Hence, facilitating the interpretation of the

data by the HP.

4.1 Proposed Solution

This section reports on the AI solutions for AI-Rehab

framework. First and the most important challenge is

patient proﬁling. Then we need to solve the task set-

AI-Rehab: A Framework for AI Driven Neurorehabilitation Training - The Proﬁling Challenge

849

ting and iterative training challenge, which will not

be tackled in this paper. Here, we propose some tech-

niques without applying them on data which can be

useful to automatize the neurorehabilitation system.

The second step in AI-Rehab framework from Sec-

tion 3 is to map NPAs values to a consolidated CP

into the interval 0 to 100 and the equation 1 is used

for the Mapping. Note that it is formulated owing to

no data.

∑

i=1

∑

j=1

Norm(NPA

).W

kij

(1)

Where,

kij











1/p, where p is count of factors,

contributing to CP

0, otherwise

NPA

= factor from a neuropsychological as-

sessment

= cognitive domain K from CP

Norm= normalization function, interval is 0 and 100

kij

= weight function, interval is 0 to 1, which sums

to 1

m= number of times a F appeared in NPAs

n= number of NPAs

The way this formula works is that we take ev-

ery factor j from each NPA i (lets say, memory from

MOCA for instance) and multiply by the weight that

factor has on determining that speciﬁc cognitive do-

main (memory in this example). At the start, we will

inquire HPs to provide us with weights for each fac-

tor, as it seems a better solution to follow the pro-

fessionals intuition than to attribute arbitrary weights.

Once we have enough data, we can start using AI

algorithms to determine the weights which we have

explained at the end of this section. The weight W

should belong in the interval of 0 to 1. Given that, the

‘Norm’ function will give us a value between 0 and

100 and the sum of the weights is 1, the output CP

will be a value between 0 and 100, corresponding to

the weighted factor of the CP (in this example, mem-

ory). This is the Mapping process from the Figure 1.

Finally, the CP

is the aggregated result of a battery

of tests, instead of only one test, and CP is a set of

cognitive domain k.

The ﬁnal step is to contextualize this proﬁle. If

we interpret the CP using the ND from all the NPAs

taking into account the patient SDI, we will get a stan-

dardized cognitive proﬁle. This process will generate

a proﬁle that is compared to the population data. This

is the step which is needed for the identiﬁcation of

deﬁcits and the outcome of this is CS. Once again,

these process will be improved as we get actual data

from patients using the system.

SCP

∑

i=1

Norm(ND

,SDI)

(2)

SDI= patient’s socio-demographic information

ND= normative data of k for each NPA used to

calculate CP

SCP

= standardized cognitive proﬁle K from ND and

SDI

Norm= normalization function, interval is 0 and 100

n= number of NPAs

First, it is important to mention that we do not

have access to the normative value for the factors of

each NPA, we just have it for the result of the whole

NPAs. Also, since it is normative data, there is no

need to consider weights as it is already embedded in

the ND value itself. The value represented by the ND

and SDI pair is the average result of someone in the

closest socio-demographic group as the patient, since

the SDI is relative to her/him. This can be observed

for MoCA in (Freitas et al., 2018) for the Portuguese

population. For simplicity, lets consider only MoCA

and memory for an example. The result of this for-

mula would be the average of someone with a similar

SDI as the patient. By doing a simple cross multipli-

cation between the CP memory score and the result

of this function, we can get a relative value of the pa-

tient when compared with the ND. This value would

be, in this example, the memory domain in the CS of

the patient, where 50% means that you are as good

as it is expected for you group, 75% means you are

25% above average or 100% means you are twice as

good as expected. Eventually, we will have collected

patient’s CS with all the factors so HP can interpret

them and take clinical decision with ease. Eventually,

it gives solution for proﬁling challenge without data.

∑

i=1

∑

j=1

Norm(NPA

).W

kij

(3)

Where, W

kij

belongs to (0≤W≤1) and

∑

i=1

kij

= 1

which is being determined by some AI techniques.

Once we get enough data, we can have AI solu-

tion for the W problem, appeared in the equation 1.

The equation 3 is a proposal solution for the weights

computation. Some statistical learning, ML or DL

techniques will apply on NPAs data to obtain highly

optimal W in the Mapping process, such as Princi-

pal Component Analysis, random forest or neural net-

work. Besides, if the number of F and NPA grow

over time then the system performance may decline

and the data will suffer with high dimentionality. To

overcome such problem, feature selection techniques

Cognitive Health IT 2020 - Special Session on Machine Learning and Deep Learning Improve Preventive and Personalized Healthcare

850

like Principal Component Analysis, LASSO, Ridge

or t-Distributed Stochastic Neighbor Embedding can

be used. These techniques generate highly inﬂuential

parameters without loosing much information. Sub-

sequently, the next step is to create the Belief Revision

Engine (BR-E), a computational infrastructure based

on the Believe Revision that will enable the accurate

prediction of CPs for patients, over the iterative train-

ing process. The BR-E has three objectives:

1. At each iteration, register the difference between

the ERTT and the real result;

2. At each iteration, check for inconsistencies;

3. End ﬁnally, once the who training is completed,

after n iterations, it will compare the n

predicted

proﬁle with the new real proﬁle after the NPA re-

assessment.

The ﬁrst point will allow us the tweak the difﬁculty

of the settings, with the goal of keep the patient in

ﬂow (Nakamura and Csikszentmihalyi, 2009). It has

been proven that people at this level of concentration

and immersion are at their most effective, which will

lead to better rehabilitation results. If after each it-

eration, the new predicted proﬁle is inconsistent, by

any reason, then an alarm must be risen and the HP

must be contact to evaluatethe situation. Finally, once

the patient has completed the trained and has been

re-evaluated, we can compare the n

predicted result

with the new real result. This will allow us the eval-

uate the performance of the system and to see if it

is performing as we want. The study (Schwind et al.,

2019) can help us to understand where the system pre-

dicted wrong, given the ﬁnal result. Once all is done,

the loop restart until the HP decides that no more ther-

apy is necessary.

5 CONCLUSION

The BRaNT project goal is the development of a

novel cognitive rehabilitation tool to allow the health

professional the monitoring and adaptation of treat-

ment at home. To do this, we propose the usage of

Artiﬁcial Intelligence (AI) to improve the existing TG

tool. BRaNT proposes the usage of machine learn-

ing, deep learning and believe revision framework is

able to assess the patient’s deﬁcits through the us-

age of the results battery of validated NPAs, generate

gamiﬁed cognitive training tasks adjusted to each pa-

tient proﬁle, and support the continuum of healthcare

from the clinic to the home with a distributed archi-

tecture with remote monitoring capabilities. Conse-

quently, this paper proposes the AI-Rehab framework

for AI driven neurorehabilitation training and identi-

ﬁes three challenges: patient’s cognitive proﬁle, task

settings and iterative training. The present work only

focuses on proﬁling challenge and proposes a solu-

tion for Mapping and Identification process. Once

we have enough data, we will apply machine learning

algorithms. As for the BR techniques, (Katsuno and

Mendelzon, 1991) research will be crucial to mea-

sure the distance between the several proﬁles gathered

during the process and (Schwind et al., 2019) work

will help us understand what actually changed and

what needs to be adapted for the next training loop.

Presently, the framework targets only cognitive do-

mains. However, in future, it can easily be extended

to cognitive sub domains or to other domains such as

ﬁtness.

ACKNOWLEDGEMENTS

We want to thank the BRaNT team for their support

and stimulating discussions. This research is sup-

ported by BRaNT - Belief Bevision applied to Neu-

rorehabilitation Therapy [project number PTDC/CCI-

COM/30990/2017],ﬁnanced by FCT - Fundac¸˜ao para

a Ciˆencia e a Tecnologia. EF is partially supported

by UID/CEC/04516/2019.SBB is partially supported

by MACBIOIDI: Promoting the cohesion of Mac-

aronesian regions through a common ICT platform

for biomedical R - D - i” (INTERREG program

MAC/1.1.b/098)

REFERENCES

Alchourr´on, C. E., G¨ardenfors, P., and Makinson, D.

(1985). On the logic of theory change: Partial meet

contraction and revision functions. The journal of

symbolic logic, 50(2):510–530.

Bergeron, M. F., Landset, S., Tarpin-Bernard, F., Ashford,

C. B., Khoshgoftaar, T. M., and Ashford, J. W. (2019).

Episodic-memory performance in machine learning

modeling for predicting cognitive health status classi-

ﬁcation. Journal of Alzheimer’s Disease, 70(1):277–

286.

Berm´udez i Badia, S., Fluet, G. G., Llorens, R., and

Deutsch, J. E. (2016). Virtual reality for sensorimo-

tor rehabilitation post stroke: Design principles and

evidence. In Neurorehabilitation technology, pages

573–603. Springer.

Booth, R., Meyer, T., Varzinczak, I., and Wassermann, R.

(2010). Contraction core for horn belief change: pre-

liminary report. Unpublished manuscript.

Bott, N. T. and Kramer, A. (2017). Cognitive rehabilitation.

Encyclopedia of Geropsychology, pages 544–551.

AI-Rehab: A Framework for AI Driven Neurorehabilitation Training - The Proﬁling Challenge

851

Carod-Artal, J., Egido, J. A., Gonz´alez, J. L., and Varela de

Seijas, E. (2000). Quality of life among stroke sur-

vivors evaluated 1 year after stroke: experience of a

stroke unit. Stroke, 31(12):2995–3000.

Chi, C.-L., Zeng, W., Oh, W., Borson, S., Lenskaia, T.,

Shen, X., and Tonellato, P. J. (2017). Personalized

long-term prediction of cognitive function: Using se-

quential assessments to improve model performance.

Journal of biomedical informatics, 76:78–86.

Cicerone, K. D., Dahlberg, C., Kalmar, K., Langenbahn,

D. M., Malec, J. F., Bergquist, T. F., Felicetti, T., Gi-

acino, J. T., Harley, J. P., Harrington, D. E., et al.

(2000). Evidence-based cognitive rehabilitation: rec-

ommendations for clinical practice. Archives of phys-

ical medicine and rehabilitation, 81(12):1596–1615.

Cicerone, K. D., Dahlberg, C., Malec, J. F., Langenbahn,

D. M., Felicetti, T., Kneipp, S., Ellmo, W., Kalmar, K.,

Giacino, J. T., Harley, J. P., et al. (2005). Evidence-

based cognitive rehabilitation: updated review of the

literature from 1998 through 2002. Archives of physi-

cal medicine and rehabilitation, 86(8):1681–1692.

Cicerone, K. D., Langenbahn, D. M., Braden, C., Malec,

J. F., Kalmar, K., Fraas, M., Felicetti, T., Laatsch, L.,

Harley, J. P., Bergquist, T., et al. (2011). Evidence-

based cognitive rehabilitation: updated review of the

literature from 2003 through 2008. Archives of physi-

cal medicine and rehabilitation, 92(4):519–530.

Cumming, T. B., Marshall, R. S., and Lazar, R. M. (2013).

Stroke, cognitive deﬁcits, and rehabilitation: still an

incomplete picture. International Journal of stroke,

8(1):38–45.

Delgrande, J. and Wassermann, R. (2010). Horn clause

contraction functions: Belief set and belief base ap-

proaches. In Twelfth International Conference on the

Principles of Knowledge Representation and Reason-

ing.

Delgrande, J. P. (2008). Horn clause belief change: Con-

traction functions. In KR, pages 156–165.

Delgrande, J. P. and Schaub, T. (2003). A consistency-

based approach for belief change. Artiﬁcial Intelli-

gence, 151(1-2):1–41.

Faria, A. L. and Berm´udez i Badia, S. (2015). Development

and evaluation of a web-based cognitive task gener-

ator for personalized cognitive training: a proof of

concept study with stroke patients. In Proceedings

of the 3rd 2015 Workshop on ICTs for improving Pa-

tients Rehabilitation Research Techniques, pages 1–4.

ACM.

Faria, A. L., Pinho, M. S., and Berm´udez i Badia, S. (2018).

Capturing expert knowledge for the personalization

of cognitive rehabilitation: Study combining com-

putational modeling and a participatory design strat-

egy. JMIR rehabilitation and assistive technologies,

5(2):e10714.

Ferm´e, E. and Hansson, S. O. (2011). Agm 25 years. Jour-

nal of Philosophical Logic, 40(2):295–331.

Ferm´e, E. and Hansson, S. O. (2018). Belief Change: In-

troduction and Overview. Springer.

Fern´andez-Delgado, M., Cernadas, E., Barro, S., and

Amorim, D. (2014). Do we need hundreds of classi-

ﬁers to solve real world classiﬁcation problems? The

Journal of Machine Learning Research, 15(1):3133–

3181.

Fern´andez-Delgado, M., Sirsat, M., Cernadas, E., Alawadi,

S., Barro, S., and Febrero-Bande, M. (2018). An

extensive experimental survey of regression methods.

Neural Networks.

Freitas, S., Batista, S., Afonso, A. C., Sim˜oes, M. R.,

de Sousa, L., Cunha, L., and Santana, I. (2018). The

montreal cognitive assessment (moca) as a screening

test for cognitive dysfunction in multiple sclerosis.

Applied Neuropsychology: Adult, 25(1):57–70.

Freitas, S., Prieto, G., Sim˜oes, M. R., and Santana, I.

(2014). Psychometric properties of the montreal cog-

nitive assessment (moca): an analysis using the rasch

model. The Clinical Neuropsychologist, 28(1):65–83.

Freitas, S., Prieto, G., Sim˜oes, M. R., and Santana, I.

(2015). Scaling cognitive domains of the montreal

cognitive assessment: an analysis using the partial

credit model. Archives of Clinical Neuropsychology,

30(5):435–447.

Freitas, S., Sim˜oes, M. R., Alves, L., Duro, D., and Santana,

I. (2012a). Montreal cognitive assessment (moca):

validation study for frontotemporal dementia. Journal

of Geriatric Psychiatry and Neurology, 25(3):146–

154.

Freitas, S., Sim˜oes, M. R., Alves, L., and Santana, I. (2011).

Montreal cognitive assessment (moca): normative

study for the portuguese population. Journal of clin-

ical and experimental neuropsychology, 33(9):989–

996.

Freitas, S., Sim˜oes, M. R., Alves, L., and Santana, I.

(2013). Montreal cognitive assessment: validation

study for mild cognitive impairment and alzheimer

disease. Alzheimer Disease & Associated Disorders,

27(1):37–43.

Freitas, S., Simoes, M. R., Alves, L., Vicente, M., and

Santana, I. (2012b). Montreal cognitive assessment

(moca): validation study for vascular dementia. Jour-

nal of the International Neuropsychological Society,

18(6):1031–1040.

Freitas, S., Simoes, M. R., Marˆoco, J., Alves, L., and San-

tana, I. (2012c). Construct validity of the montreal

cognitive assessment (moca). Journal of the Interna-

tional Neuropsychological Society, 18(2):242–250.

George, D. R. and Whitehouse, P. J. (2011). Marketplace of

memory: what the brain ﬁtness technology industry

says about us and how we can do better. The Geron-

tologist, 51(5):590–596.

Hastie, T., Tibshirani, R., and Friedman, J. (2009). The el-

ements of statistical learning: data mining, inference,

and prediction. Springer Science & Business Media.

Jiang, F., Jiang, Y., Zhi, H., Dong, Y., Li, H., Ma, S., Wang,

Y., Dong, Q., Shen, H., and Wang, Y. (2017). Arti-

ﬁcial intelligence in healthcare: past, present and fu-

ture. Stroke and vascular neurology, 2(4):230–243.

Katsuno, H. and Mendelzon, A. O. (1991). Propositional

knowledge base revision and minimal change. Artiﬁ-

cial Intelligence, 52(3):263–294.

Cognitive Health IT 2020 - Special Session on Machine Learning and Deep Learning Improve Preventive and Personalized Healthcare

852

Ko, H., Ihm, J.-J., Kim, H.-G., Initiative, A. D. N., et al.

(2019). Cognitive proﬁling related to cerebral amy-

loid beta burden using machine learning approaches.

Frontiers in aging neuroscience, 11.

LeCun, Y., Bengio, Y., and Hinton, G. (2015). Deep learn-

ing. nature, 521(7553):436–444.

Morris, K., Hacker, V., and Lincoln, N. B. (2012). The

validity of the addenbrooke’s cognitive examination-

revised (ace-r) in acute stroke. Disability and rehabil-

itation, 34(3):189–195.

Nakamura, J. and Csikszentmihalyi, M. (2009). Flow the-

ory and research. Handbook of positive psychology,

pages 195–206.

Nasreddine, Z. S., Phillips, N. A., B´edirian, V., Charbon-

neau, S., Whitehead, V., Collin, I., Cummings, J. L.,

and Chertkow, H. (2005). The montreal cognitive as-

sessment, moca: a brief screening tool for mild cogni-

tive impairment. Journal of the American Geriatrics

Society, 53(4):695–699.

Pagnucco, M. (2006). Knowledge compilation for belief

change. In Australasian Joint Conference on Artiﬁcial

Intelligence, pages 90–99. Springer.

Parsons, T. D. (2015). Ecological validity in virtual reality-

based neuropsychological assessment. In Encyclope-

dia of Information Science and Technology, Third Edi-

tion, pages 1006–1015. IGI Global.

Paulino, T., i Badia, S. B., and Cameir˜ao, M. (2019). Us-

ability evaluation of an integrative exergaming system

for the senior population. In 2019 5th Experiment In-

ternational Conference (exp. at’19), pages 286–291.

IEEE.

Paulino, T., Mu˜noz, J., Berm´udez i Badia, S., and Cameir˜ao,

M. S. (2018). Design of an integrative system for con-

ﬁgurable exergames targeting the senior population.

In International Conference on Human Systems Engi-

neering and Design: Future Trends and Applications,

pages 287–292. Springer.

Peppas, P. and Williams, M.-A. (2016). Kinetic consistency

and relevance in belief revision. In European Confer-

ence on Logics in Artiﬁcial Intelligence, pages 401–

414. Springer.

Schwind, N., Inoue, K., Konieczny, S., Lagniez, J.-M., and

Marquis, P. (2019). What has been said? identifying

the change formula in a belief revision scenario. In

Proceedings of the Twenty-Eighth International Joint

Conference on Artiﬁcial Intelligence, IJCAI-19, pages

1865–1871. International Joint Conferences on Artiﬁ-

cial Intelligence Organization.

Solana, J., C´aceres, C., Garc´ıa-Molina, A., Opisso, E.,

Roig, T., Tormos, J. M., and G´omez, E. J. (2014). Im-

proving brain injury cognitive rehabilitation by per-

sonalized telerehabilitation services: Guttmann neu-

ropersonal trainer. IEEE journal of biomedical and

health informatics, 19(1):124–131.

Williams, M.-A. and Sims, A. (2000). Saten: An object-

oriented web-based revision and extraction engine.

arXiv preprint cs/0003059.

Wilson, A. and McDonagh, J. (2013). Application of the

principles of gamiﬁcation to facilitate acquisition of

self-management skills in young people with long-

term medical conditions. In European Conference on

Games Based Learning, page 579. Academic Confer-

ences International Limited.

Zhuang, Z. Q., Pagnucco, M., and Meyer, T. (2007). Imple-

menting iterated belief change via prime implicates.

In Australasian Joint Conference on Artiﬁcial Intelli-

gence, pages 507–518. Springer.

AI-Rehab: A Framework for AI Driven Neurorehabilitation Training - The Proﬁling Challenge

853