clinical study adaptive personalized training games for...

Clinical StudyAdaptive Personalized Training Games for Individual andCollaborative Rehabilitation of People with Multiple Sclerosis

Johanna Renny Octavia12 and Karin Coninx2

1 Industrial Engineering Department Parahyangan Catholic University Ciumbuleuit 94 Bandung 40141 Indonesia2 Expertise Centre for Digital Media-tUL-iMinds Hasselt University Wetenschapspark 2 3590 Diepenbeek Belgium

Correspondence should be addressed to Johanna Renny Octavia johannaunparacid

Received 23 December 2013 Revised 10 March 2014 Accepted 13 March 2014 Published 28 May 2014

Academic Editor Alessandro De Mauro

Copyright copy 2014 J R Octavia and K Coninx This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited

Any rehabilitation involves people who are unique individuals with their own characteristics and rehabilitation needs includingpatients suffering fromMultiple Sclerosis (MS)The prominent variation ofMS symptoms and the disease severity elevate a need toaccommodate the patient diversity and support adaptive personalized training to meet every patientrsquos rehabilitation needs In thispaper we focus on integrating adaptivity andpersonalization in rehabilitation training forMSpatientsWe introduced the automaticadjustment of difficulty levels as an adaptation that can be provided in individual and collaborative rehabilitation training exercisesfor MS patients Two user studies have been carried out with nine MS patients to investigate the outcome of this adaptation Thefindings showed that adaptive personalized training trajectories have been successfully provided to MS patients according to theirindividual training progress which was appreciated by the patients and the therapistThey considered the automatic adjustment ofdifficulty levels to provide more variety in the training and to minimize the therapists involvement in setting up the training Withregard to social interaction in the collaborative training exercise we have observed some social behaviors between the patients andtheir training partner which indicated the development of social interaction during the training

1 Introduction

Multiple Sclerosis (MS) is an autoimmune chronic andprogressive disease of the central nervous systemof humansPeople withMS suffer fromdamaged nerves which lead to theprogressive interference with functions that are controlled bythe nervous system such as vision speech walking writingand memory causing severe limitations of functioning indaily life To date no cure has been found for MS yetTherefore for MS patients the aim of rehabilitation trainingis different compared to any other disease Rehabilitationtraining will not completely recover MS patients however itmay improve their functional mobility and quality of life

People who are in need of any kind of rehabilitationare individuals with their own characteristics and needsAlthough they might be subjected to the same backgroundcause for rehabilitation the stage of their condition or theseverity of their disease may differ which requires different

treatments and forms of rehabilitation In the case of MS theindividual differences among its patients are quite prominentdue to the great variation of MS symptoms and the impactlevels of the disease For example their physical abilitieswhich are largely influenced by the degree ofmuscleweaknessexperienced by the patients differ a lot The differencein physical abilities has an impact on the course of therehabilitation training Some training tasks may be difficultfor some patients because of their limited capabilities due totheir high degree of muscle weakness while other patientsexperience less problems in performing those tasks

During rehabilitation each patient progresses in differentways thus the training exercises must be tailored to eachindividual differently For example the difficulty of an exer-cise should increase faster for those who are progressing wellcompared to those who are having trouble performing theexerciseThe condition of patientsmay also change over timeit can deteriorate according to the progress of the disease or

Hindawi Publishing CorporationBioMed Research InternationalVolume 2014 Article ID 345728 22 pageshttpdxdoiorg1011552014345728

2 BioMed Research International

it can improve as a result of the treatment and rehabilitationeffortsThis change of condition should be taken into accountto ensure providing the right level of rehabilitation trainingto the patient at the right time Therefore due to thediversity among MS patients it would be unwise to offer thesame rehabilitation training to every patient This situationdemands a suited personalized rehabilitation training tomeet every patientrsquos needs

To acquire a good result of rehabilitation it is necessaryto maintain patient motivation Generally rehabilitationinvolves the same training exercises that should be performedrepetitively and for a long period of time Using games asthe training platform has been considered to maintain andenhance patientsrsquo motivation during rehabilitation especiallycollaborative games in which social interaction is incorpo-rated However the usage of game-like training exercise is notthe ultimate solution Some patients may feel less motivatedwhen finding the game to be too easy or too difficultor when reaching a certain point in the training wherethey become bored with the game Therefore rehabilitationtraining should be set at an appropriate level of challenge ordifficulty to maintain the motivation of patients This raisesthe need of adaptivity in the rehabilitation training to ensurethe effectiveness of the rehabilitation

Our work focuses on the idea of integrating adaptivityand personalization into the rehabilitation training for MSpatients and investigating to what extent the adaptive person-alized training may contribute to a successful neurologicalrehabilitation This paper firstly discusses the survey ofrelated work concerning adaptation in rehabilitation train-ing followed by a description of our rehabilitation systemdeveloped to support personalized rehabilitation trainingfor MS patients We further elaborate on the investigationof integrating the adaptivity of automatic difficulty leveladjustment in the rehabilitation training for MS patientsincluding the user studies carried out with a group of MSpatients to acquire their feedback

2 Related Work

In this section we first provide a brief overview of researchwhich has studied adaptation in rehabilitation training Wethen discuss several works which have investigated the usageof virtual environments as the platform for rehabilitationtraining Lastly we describe a number of studies whichcombined both and attempted to integrate adaptation invirtual environments for rehabilitation training

21 Adaptation in Rehabilitation Training The integration ofadaptation in rehabilitation training has been the focus ofseveral studies [1ndash4] These studies have mainly investigatedthe integration of adaptivity in robot-assisted rehabilitationtraining which aimed at providing a personalized trainingto the patients according to their individual characteristicsneeds and abilities Moreover the adaptivity is intendedto facilitate an automated training system to minimize thetherapistrsquos effort in manually adjusting the rehabilitationtraining

Jezernik et al [1 4] studied the adaptation in rehabilita-tion training of locomotion for stroke and spinal cord injuredpatients An automated treadmill training system was intro-duced using a robotic rehabilitation device to increase thetraining duration and reduce the physiotherapistsrsquo effort Aclinical study on six spinal cord injured patients showed thatthe treadmill training with adaptive gait patterns increasesthe motivation of the patient and gives himher the feelingthat they are controlling themachine rather than themachinecontrolling them Kahn et al [2] described the integrationof adaptive assistance into guided force training as part ofthe upper extremity rehabilitation for chronic stroke patientsAn adaptive algorithm was developed to individually tailorthe amount of assistance provided in completing the trainingtask This algorithm has been evaluated with one patientin a two-month training program which showed significantimprovements in the patientrsquos arm function reflected by theperformance increase of functional activities of daily livingsuch as tucking a shirt and stabilizing a pillow Kan et al [3]presented an adaptive upper-limb rehabilitation robotic sys-tem for stroke patients which accounts for the specific needsand abilities of different patients Using the decision theo-retic model the system autonomously facilitates upper-limbreaching rehabilitation by tailoring the exercise parametersand estimating the patientrsquos fatigue based on the observationin hisher compensation or control ofmovementsThe systemperformance was evaluated by comparing the decisionsmadeby the system with those of a human therapist Overall thetherapist agreed with the system decisions approximately65 of the time and also thought the system decisions werebelievable and could envision this system being used in botha clinical and home setting

22 Virtual Environments for Rehabilitation Training An-other emerging technology that has been widely applied inrehabilitation is the virtual environment technology Vir-tual environments provide a variety of potential benefitsfor many aspects of rehabilitation training Schultheis andRizzo [5] discussed a number of advantages for the use ofvirtual environments in rehabilitation One key benefit isthat virtual environments are rather naturalistic or ldquoreal-liferdquo environments which may allow the users or patientsto be immersed within the environment and ldquoforgetrdquo thatthey are in a rehabilitation training session Within virtualenvironments patients are also facilitated to perform therehabilitation training tasks using 3D interaction which givesa close resemblance to the actual movements in the realworld Schultheis and Rizzo [5] also pointed out that virtualenvironments facilitate the design of individualized trainingenvironments where therapists can better tailor the trainingexercises based on an individualrsquos abilities and needs andthey can also easily apply gradual increments of difficulty andchallenge Another benefit is that virtual environments allowthe introduction of gaming factors into the rehabilitationscenario to enhance motivation of patients

Holden [6] provided a thorough overview of the useof virtual environments in the field of motor rehabilitationIn the context of motor rehabilitation several studies have

BioMed Research International 3

shown that patients with motor impairments are able to traintheir motor skills in virtual environments and transfer theseabilities to the real world Furthermore these studies havealso indicated that motor learning in a virtual environmentmay be superior to that in a real environment for examplea study described in Webster et al [7] which comparedtwo groups of patients with stroke and unilateral neglectsyndrome in a wheelchair training program The groupwhich had trained in a virtual environment hit significantlyfewer obstacles with their wheelchair during the real worldobstacle avoidance test than the group which only trainedin a real environment Another study described in Jaffeet al [8] investigated chronic stroke patients in a trainingto avoid obstacles during walking Patients using a virtualenvironment-based training showed greater improvement ina fast paced velocity test in comparison to the patients whotrained in the real world

In Holden [6] an extensive description of the variousvirtual environments that have been developed for reha-bilitation purposes of patients was provided This includesrehabilitation for stroke (upper and lower extremity trainingand spatial and perceptual-motor training) acquired braininjury Parkinsonrsquos disease orthopedic disorders vestibulardisorders (balance training) wheelchair mobility trainingand functional activities of daily living training A review onthese research initiatives showed that a wide variety of clinicalapplications using virtual environments has been developedand tested Mainly the virtual environments consist of scenesthat were designed to be simple therapeutically meaningfultasks to the targeted patients such as making coffee lifting acup to the mouth using an automated teller machine (ATM)and way-finding Virtual environments have been considerednot as a treatment for motor rehabilitation in itself but moreas a new technological tool that can be exploited to enhancemotor rehabilitation training

Over the past few years there is an increasing researchinterest in the development of virtual environments for usein stroke rehabilitation Virtual environments are consideredbeneficial in stroke rehabilitation because they enable moreprecisely controlled training settings intensive practice witheasier repetition of tasks automatic recording of trainingprogress and more enjoyable and compelling interactionfor the patients Some researchers developed virtual envi-ronments based on the literal translation of activities ofdaily living such as self-feeding bathing and dressing orgrooming For example Edmans et al [9] developed a virtualenvironment to train a self-feeding task (ie making a hotdrink) An evaluation study with 50 stroke patients showedthat their performance of hot drink-making in the real worldand in the virtual environment were correlated which mayindicate the usefulness of such virtual environment as arehabilitation tool for stroke patients

Other researchers chose to enclose the rehabilitationtasks in game-like exercises with the purpose of addingfun and challenges into rehabilitation training to increaseor maintain the patientrsquos motivation For instance Alankuset al [10] developed a series of home-based stroke reha-bilitation games which make use of inexpensive devicesfeasible for home use such as Wii remotes and webcams

Two examples of the games are the helicopter game inwhich the patients have to maneuver a flying helicopterusing their arm to avoid hitting buildings and to collectfuel cells in the air and the baseball catch game in whichthe patients have to differentiate balls and control thebaseball glove to catch baseballs and avoid basketballs Thepreliminary evaluation with one therapist and four strokepatients showed that the games motivated the patients andthey used the right motions required for the rehabilitationtraining Saposnik et al [11] described the first randomizedclinical trial on the effectiveness of virtual reality usingthe Wii gaming system namely VRWii on the arm motorimprovement in stroke rehabilitationThe results showed thatthe VRWii gaming technology represented a feasible safeand potentially effective alternative to facilitate rehabilitationtherapy and enhance motor function recovery in strokepatients

23 Adaptation in Virtual Environments for RehabilitationTraining As previously discussed in Section 21 it is essentialto provide a personalized rehabilitation training which issuited according to the individual characteristics of patientsPatients involved in rehabilitation have a wide range of needsand abilities which may benefit from integrating adaptivityinto their rehabilitation training Adaptivity enables offeringa tailored training with minimal efforts from the humantherapists as it decreases the dependence on them to continu-ously monitor the patientrsquos progress and manually adjust thetraining program

Several studies mentioned in Section 22 have acknowl-edged the potential of developing virtual environment appli-cations for the purpose of rehabilitation training Theseapplicationsmay also benefit from adaptivity since it allows todynamically adjust the parameters of the virtual environmentas the training tool to provide a suited personalized trainingto every patient based on hisher current needs and abilitiesIt may also be necessary to integrate adaptivity in virtualenvironments for rehabilitation training due to the fact thatthe complex 3D interaction may introduce extra difficultiesand eventually influence patientsrsquo performance during thetraining sessions Adaptivity may reduce this effect by adjust-ing the virtual environment according to the patientrsquos level ofinteraction

This section discusses several previous studies whichhighlighted the integration of adaptation in virtual environ-ments that were developed for the purpose of rehabilitationtraining Ma et al [12] stated that virtual reality systems forrehabilitation can benefit a group of patients with a greatdiversity through adaptation In their study several adaptivevirtual reality games for rehabilitation of stroke patientswith upper limb motor disorders have been developedThe information of patient performance is used to enableautomatic progression between difficulty levels in the gamesThe elements of the games are designed to be adaptive andto change dynamically according to how well or badly thepatient is performing An initial evaluation from patientsshowed positive feedback since they enjoyed training whileplaying the game and they felt more motivated


Cameirao et al [13] have developed an adaptive virtualreality based gaming system for the upper extremities reha-bilitation of acute stroke patients They proposed a multitaskadaptive rehabilitation system that provides a task-orientedtraining with graded complexity The basic training consistsof a virtual reality game where flying spheres move towardsthe patient These objects have to be intercepted usingthe virtual arms which are controlled by the patientrsquos armmovements As the first task the patients have to performthe hitting task to train range of movement and speed Thesecond task is the grasping task to train finger flexure andthen finally the placing task to train grasp and release is thelast task Based on the individual performance of the patientduring training the difficulty of the task is adapted by mod-ulating several parameters such as the speed of the spheresthe interval of appearance between consecutive spheres andthe range of dispersion in the field The impact of the systemon the recovery of patients was evaluated with 14 patientsThe results suggested that the system induces a sustainedimprovement during therapy with observed benefits in theperformance of activities of daily living

Barzilay and Wolf [14] developed an online biofeedback-based adaptive virtual training system for neuromuscularrehabilitation The system employs an artificial intelligencelearning system that learns from real-time biofeedback andproduces online new patient-specific virtual physiotherapymissions In the training the patient is asked to follow thevirtual tasks (ie trajectories) displayed to himher by mak-ing upper limb movements accordingly The performance isthenmodeled by tracking and recording their kinematics andmuscle activity (measured through electromyography (EMG)signals) using amotion tracking system Based on this trainedmodel the system changes and adapts the task displayedto the patient which results in a new virtual trajectory asthe training exercise This adaptive loop of the system iscontinuously repeated to provide a real-time adaptive reha-bilitation virtual training system for better neuromuscularrehabilitation

According to Holden [6] the use of mixed reality whichcombines physical and virtual environments in rehabilitationtraining can provide adaptive scenes for interactive practiceand feedback that engage the patient physically and mentallyDuff et al [15] presented an adaptive mixed reality rehabilita-tion training system to help improve the reachingmovementsof patients who suffer hemiparesis from stroke The systemprovides real-time multimodal customizable and adaptivefeedback based on the movement patterns of the patientrsquosaffected arm and torso during the reaching movement Thekinematic data is used to assess the movement and adaptthe parameters linked to the feedback presentation and thephysical environment that determines the type of task Thefeedback is provided via innovative visual and musical formsthat present a stimulating enriched environment for thepatientrsquos training For example the audio feedback in theform of music is intended to encourage patients to performthe desired reaching movements The velocity of the patientrsquoshand controls the rhythm of the music when the patient ismoving too slow the music is played with a slower rhythmand when the patient moves with nonsmooth acceleration

or deceleration there will be abrupt changes in the paceof the music Additionally the sequence and intensity oftasks can also be adapted to better address each patientrsquosrehabilitation needs After trainingwith the interactivemixedreality system three chronic stroke patients showed improvedreachingmovements Integratingmixed reality environmentsinto the training was argued to promote an easier bridging inmotor learning between the virtual and physical worlds Inone of their following research Baran et al [16] presentedthe design of a home-based adaptive mixed reality systemfor stroke rehabilitation The system which consists of acustom table chair and media center is designed to be easilyintegrated into any home to provide an engaging long-termreaching task therapy for stroke patients

24 Summary of Related Work In this section we havediscussed several research efforts which have mainly inves-tigated the integration of adaptivity in robot-assisted reha-bilitation training Besides providing a personalized trainingthe adaptivity integrated in the studies was intended tominimize the dependence from the therapists for manuallyadjusting the rehabilitation training Virtual environmenttechnology has been considered to be beneficial and widelyapplied in rehabilitation We also have discussed a numberof research initiatives that used virtual environments as partof the rehabilitation training platform and demonstrated thepatientsrsquo ability to train in virtual environments and transferthe trained skills to the real world Furthermore we havedescribed several studies which have attempted to integrateadaptivity in the virtual environment training system forachieving better results and patientrsquos engagement in reha-bilitation Mostly the adaptivity takes form in automaticallyadjusting the parameters of the virtual environment to alterthe difficulty levels in the rehabilitation exercises based on thepatientrsquos current training performance and training progress

We observed that previous studies mainly focused onrobot-assisted rehabilitation training for stroke patients Inour work we aim to develop a haptic-based rehabilitationsystem (combining robot-assisted rehabilitation and virtualenvironments technologies) to support a systematic and per-sonalized upper limb rehabilitation training for MS patients

3 Rehabilitation for Multiple Sclerosis

Rehabilitation for people withMSmainly aims for improvingtheir functional mobility and quality of life rather thanattaining their full recuperation from the disease This is dueto the fact that MS is an incurable illness In the past physicaltraining was not advised for MS patients due to the opinionthat it would advance the deterioration Now howeverperforming physical training is often part of therapy andrehabilitation efforts for MS patients in parallel with takingmedications A number of studies have shown beneficialeffects of physical training inMS regarding lower limbmusclestrength exercise tolerance level functional mobility (iebalance and walking) and quality of life while no harmfuleffects were reported [17 18]


Little research has investigated the therapeutic value ofupper limb in particular arm rehabilitation for MS patientsUpper limb rehabilitation is considered important sinceupper extremity dysfunction strongly influences the capacityof MS patients to perform several activities of daily living(ADL) such as self feeding (eg eating filling or drinkinga cup of coffee) bathing grooming dressing and takingmedications

31 Robot-Assisted Upper Limb Rehabilitation Keys to a suc-cessful neurological rehabilitation are training duration andtraining intensity [19] As time dedicated to upper limbtraining may be limited in a formal training session addi-tional therapeutic modalities may be necessary to enableMS patients to train independently of the therapist Robot-assisted rehabilitation and virtual environment technologieshave been considered to be promising to provide an effectiveindependent upper limb rehabilitation training [20 21]Robot-assisted rehabilitation allows high-intensity repetitivetask-specific and interactive treatment of the impaired upperlimb Virtual environments facilitate patients to perform thetraining tasks with their upper limb using 3D interactionwhich gives a close resemblance to the real movements in thereal world

In the context of a European research project (INTER-REG-IV program ldquoRehabilitation robotics IIrdquo) we investigatethe effects of robot-assisted upper limb rehabilitation trainingforMSpatientsOurwork combined two technologies robot-assisted rehabilitation and virtual environments by firstinvestigating the feasibility of using a phantom haptic devicefor arm rehabilitation in MS patients [22 23] Further acomplete haptic-based rehabilitation system I-TRAVLE wasdeveloped to support a systematic and personalized upperlimb rehabilitation training for MS patients [24]

I-TRAVLE (Individualized Technology-supported andRobot-Assisted Virtual Learning Environments) consists of ahardware and software system setup as depicted in Figure 1Themain component of the hardware system is a haptic robotthe MOOG HapticMaster that functions as both input andoutput devices As an input device it allows patients to inter-act with the software applications that deliver the trainingexercises As an output device it provides haptic feedbackduring the training by guiding or hindering the patientrsquosarm movement with exerted forces The HapticMaster isequipped with a peripheral device the ADL Gimbal wherethe patientsrsquo hand is placed and secured using the attachedbrace while performing the training exercises A large displaya full HD 4010158401015840 Samsung TV screen is used as a visual displayto project the training exercises and is placed behind theHapticMaster approximately 15m in front of the patient Acomplete description of the I-TRAVLE hardware systemwiththe adjustments made for the context of MS training can befound in de Weyer et al [25]

The main components of the software system of ourI-TRAVLE system are the training exercises the patientinterface the therapist interface and the central databaseTwo types of training exercises were developed namelythe basic and the advanced training exercises We will

Figure 1 I-TRAVLE system setup

(a) Basic training

(b) Advanced training

Figure 2 The patient interface

discuss the training exercises in Section 32 The patientinterface as shown in Figure 2 is created to enable patientsto independently start performing and navigating throughthe exercises predefined by the therapist Figure 2(a) showshow the patient interface looks like in the basic training andFigure 2(b) shows the interface for the advanced trainingAnother interface was developed for the therapist as shownin Figure 3 to manage the user data personalize therapysessions and review the data logged from the performedtraining exercises The central database stores all the dataabout the training exercises and patientsrsquo performance Amore detailed description of the I-TRAVLE software systemcan be found in Notelaers et al [26]


Figure 3 The therapist interface

Several other researches have been done in the course ofthe I-TRAVLE system development regarding the influenceof visual aspects on the capabilities of MS patients Peoplewith MS often suffer from visual system disorders and cog-nitive dysfunctions which may influence their capabilitieswhile navigating in a virtual environment Van den Hoogenet al [27] investigated the impact of visual cues such asshading on navigation tasks in a virtual environment forMS patients The study showed that the addition of shadebelow patientsrsquo current position in the virtual environmentimproves speed during the task and reduces the time spenton the task Van den Hoogen et al [28] discussed a studyto comparatively test the effectiveness of two characteristicsof virtual environments namely the stereo visualization andthe graphic environment in rehabilitation training whenutilizing a 3D haptic interaction device The experimentresults showed that the use of stereoscopy within a virtualtraining environment for neurorehabilitation of MS patientsis most beneficial when the task requires movement in depthFurthermore the use of 25D in the graphic environmentimplemented showed the highest efficiency and accuracy interms of patientsrsquo movements

32 Individual Training Exercises for Upper Limb Rehabilita-tion To keep up the motivation of patients and strive for asuccessful rehabilitation trajectory it is essential to give themtraining exercises that are meaningful in supporting theirfunctional recovery [29] The development of the I-TRAVLEtraining approach was inspired by the T-TOAT (technology-supported task-oriented arm training) method of Timmer-mans et al [30] T-TOAT is a technology supported (butnot haptic) training which allows integration of daily tasksSimilar to the T-TOATmethod I-TRAVLEdivides an activityof daily living (ADL) into skill components and trains theskill components first every component separately and laterseveral components combined

The I-TRAVLE training exercises were designed basedon the skill components that patients need to train relatedto their upper limb rehabilitation Two types of trainingexercises are provided namely basic training exercises whichinclude only one skill component and also advanced trainingexercises which combine multiple skill components Sevenbasic training exercises were developed in I-TRAVLE as

(a) The lifting exercise

(b) The transporting exercise

(c) The rubbing exercise

Figure 4The basic training exercises for upper limb rehabilitation

follows pushing pulling reaching turning lifting trans-porting and rubbing Figure 4 shows three examples of thebasic training exercises lifting (Figure 4(a)) transporting(Figure 4(b)) and rubbing (Figure 4(c))

In the advanced training exercises several skill compo-nents were combined into a game-like training exercise asillustrated in Figure 5 Until now four advanced trainingexercises have been designed as follows penguin paint-ing (Figure 5(a)) arkanoid (Figure 5(b)) egg catching andflower watering Normally in a therapy session a patient firsttrains with several basic training exercises Depending on theprogress of the patient the therapy session can be continuedwith the games (ie advanced training exercises)

One example of the advanced or game-like trainingexercises is the penguin painting game as illustrated inFigure 5(a) This game was designed based on two skillcomponents lifting and transporting Lifting is one of theimportant skill components in the upper limb rehabilitationfor MS patients In the penguin painting game the patienthas to collect as many points as possible within a certaintime period by painting as many penguins as possible withthe right color Figure 6 illustrates how the penguin paintinggame is played On the left side there are two shelves withpenguins waiting to be painted The patient has to select onepenguin from a shelf (Figure 6(a)) and paint it according to


(a) The penguin painting game

(b) The arkanoid game

Figure 5 The advanced training exercises for upper limb rehabili-tation

the color of its belly To paint the patient needs to bring thepenguin to the corresponding buckets by dipping it into thebottom to paint the lower part of the penguin (Figure 6(b))and continue dipping it into the top bucket to paint the upperpart of the penguin (Figure 6(c)) While painting the patientmust hold the penguin long enough to effectively apply thecolor and train for stabilization in this way At some pointsduring the game a devil that tries to capture the penguinappears (Figure 6(d)) andmust be avoided in order to not losethe penguin already in hand Every time the patient finishespainting a penguin the colored penguin must be transportedto the exit platform on the right side (Figure 6(e))

The aforementioned training exercises discussed in thissection are designed to be part of individual rehabilitationtraining for MS patients This means that the patient willperform the training exercises on hisher own In this casethe role of the therapist is to determine which training exer-cises the patient has to perform and also to personalize thetherapy sessions according to the patientrsquos current conditionand needs In another case the therapist will extend hisherrole to not only manage the therapy sessions but also take anactive part in the therapy sessions by collaborating with thepatient in performing the training exercises This so-calledcollaborative rehabilitation training will be further discussedin the next section

33 Collaborative Training Exercises for Social RehabilitationProgram Typically exercises in robot-assisted rehabilitationinvolve a patient training on hisher own (with a therapistrsquos

supervision) and performing repetitive movements for a cer-tain time period predetermined by the therapist Even whengaming elements are integrated in the training exercisesit is crucial to exploit a variety of motivational techniquesto avoid bored patients and patients that are stopping thetherapy due to the high intensity and repetitiveness ofthe exercises Therefore we explore different motivationalaspects to increase the patientrsquos engagement in the therapy

Social support has been demonstrated to be beneficial forthe engagement andmotivation of patients in a rehabilitationcontext Van den Hoogen et al [32] have performed a studyon rehabilitation needs of stroke patients which indicatedthat social support is critical for patient motivation in orderto adhere to the necessary regime of rehabilitation exercisesin the chronic phase of stroke In addition to using gamespatientrsquos motivation during rehabilitation can be maintainedand further enhanced through the incorporation of socialinteraction into the training exercises offered in the rehabili-tation program

During rehabilitation patients can receive social supportfrom different groups in their social network not only fromtheir familymembers and friends but also fromother supportgroups such as their therapists caregivers or fellow patientsThese people can extend their form of social support tothe patients by actively participating in the therapy sessionsThis support can be manifested in two different formssympathy and empathy [33 34] Both depend on the typesof relationships built based on shared emotions and senseof understanding between the persons Two possible socialscenarios can exist in this context

(1) Sympathetic This social situation occurs when aperson shows the ability to understand and to supportthe condition or experience of the patient with com-passion and sensitivity For example healthy familymembers and close friends may be able to showsympathy towards MS patientsOne can imagine a situation where a patient isvisited by a family member (eg a daughter) Thisfamily member could show her sympathetic supportthrough her help in the rehabilitation programby per-forming the collaborative training exercise togetherwith the patient

(2) Empathetic This social situation happens when aperson shows the ability to coexperience and relate tothe thoughts emotions or experience of the patientwithout them being directly communicated Otherfellow patients can easily have empathy towards MSpatients due to the resemblance of their conditionOne can imagine a situation where two patients aresupporting each other by training together at the sametime through the collaborative exercise which makesthe training more pleasant and fun

Social interaction can be incorporated in the rehabili-tation training by providing a social play medium whichrequires a patient to collaborate with hisher supportingpartner in performing the training exercises Thus the ideais that the patient will keep being engaged and stay motivated


(a) Selecting a penguin from a shelf

(b) Painting the lower part of the penguin (c) Painting the upper part of the penguin

(d) Avoiding the devil (e) Releasing the painted penguin to the exitplatform

Figure 6 Illustration of the penguin painting game

in training by collaborating with other people as a trainingpartner By this we extend the individual rehabilitationtraining to a collaborative rehabilitation training where thetraining exercises involve collaboration and participation ofmore than only a single MS patient

Vanacken et al [31] have explored the possibility of socialrehabilitation training by designing a simple collaborativegame-like training exercise as shown in Figure 7 A collab-orative balance pump game was created (Figure 7(a)) wheretwo people ldquoplayrdquo together during the therapy session Thegoal of the game is to collect all stars which represent pointsby hitting them with a ball rolling over a beam Both playershave to collaboratively control and balance the height of

both sides of the beam by pumping each side of it in turnsFigure 7(b) illustrates the setup of the social rehabilitationtraining session in a sympathetic scenario where a patientplays the collaborative balance pump game with a familymember To produce the pumping gesture the patient uses aHapticMaster and the healthy person uses aWiiMote as inputdevices

An informal user study has revealed that most patientsand therapists liked and enjoyed training with the collab-orative balance pump game described in Vanacken et al[31] Inspired by this we developed another collaborativetraining exercise social maze which hasmore game elementsand variations [35] This collaborative training game was


(a) The collaborative balance pump training exercise

(b) Sympathetic social scenario a patient training incompanion of a family member

Figure 7 Example of a collaborative training exercise for MSpatients Vanacken et al [31]

Figure 8 The social maze game

designed based on several skill components (see Section 32)that MS patients need to train related to their upper limbrehabilitation lifting transporting turning pushing andreaching Figure 8 depicts our social maze with all gameelements The goal of this game is to collect all symbolswhich represent points by picking up each symbol andbringing it to the collecting bin The elements of the gamewere purposively designed so that two players (ie a patientand hisher training partner) have to collaborate as such to

(a)Selectinga symbol

(b) Collecting the symbols in the bin

(c) Avoiding thelaser beam

(d)Demolishingthe bomb

(e) Avoiding thedevil

(f)Surmountingthe rotator

(g) Gaining lives with the connecting heart

Figure 9 Illustration of the social maze game

achieve the goal Without collaboration it is impossible tofinish the game

As can be seen in Figure 8 the game area is dividedinto two In each part the player represented by a fish-like avatar can move around hisher own maze to col-lect the symbols Figure 9 illustrates how the social mazegame is performed First the player has to select a symbol(Figure 9(a)) Once the players pick up a symbol they canplace it in the collecting bin (Figure 9(b)) to earn pointsAlong the way there are some obstacles such as the laserbeams that need to be avoided (Figure 9(c)) the bombs thatneed to be demolished (Figure 9(d)) the devil that may notbe encountered (Figure 9(e)) and the rotators that need to besurmounted (Figure 9(f)) Demolishing the bombs demandsa tight collaboration between the players where one playerhas to push the bomb trigger (lifting skill component) inorder to destroy the bomb blocking the way of the otherplayer To pass the rotator the players must enter it andperform the turning movement (turning skill component)to rotate it When the players are hit by the laser beam orthe devil they will lose a life represented by a heart Togain more lives the players must attain the connecting heart(Figure 9(g))


34 Personalization in I-TRAVLE The diversity of MSpatients raises the need for personalization in their reha-bilitation training The differences in physical abilities MSsymptoms and disease progression bring an influence to thepatientsrsquo condition and their rehabilitation needs Thereforeoffering the same training trajectory to MS patients is astraightforward yet unwise approach A personalized train-ing that complies to each patientrsquos characteristics becomes anecessity to accommodate the patient diversity and ensurean effective and satisfactory rehabilitation The I-TRAVLEsystem has been developed with a concern for supportingpersonalization in upper limb rehabilitation training forMS patients An exploratory study on the personalizationaspects has been discussed in Notelaers et al [36] The firstaspect considered in the personalization infrastructure of I-TRAVLE was the workspace determination of patients Dueto the difference in patientsrsquo abilities in using their upperlimb some patients may have a smaller range of motion thanthe others Based on the measurement of their active rangeof motion (AROM) the setup of the training program canbe adjusted accordingly to ensure that the effort required toperform the training exercises is within the capability of thepatient and no impossible or harmful arm movements werenecessary

Within I-TRAVLE normally the therapist manuallydetermines the training program and its exercise parametersin the therapist interface (see Figure 3) to present suitablychallenging individualized rehabilitation training The ther-apist can set up a training program according to the patientrsquosdesire to train certain activities of daily living For examplewhen the patient prefers to train hisher ability of drinkinga cup of coffee (ie self feeding) the therapist can choosethe suitable skill components such as reaching lifting andtransporting Providing customizable parameters of theseskill components ensures personalization to better suit thepatientsrsquo capabilities A personalized training program withthe appropriate difficulty level according to the patientsrsquoabilities can be established For instance to achieve thetraining of more fine-grained arm movements the therapistcan customize the parameters of the lifting exercise to ahigher difficulty Several kinds of parameters are adjustablefor instance related to visual or haptic feedbackThe amountof haptic feedback that is given when the patient has tofollow a certain path in the virtual world influences howeasy or how difficult it is for the patient to stay on the pathFigure 4(a) shows the lifting exercise The green line showsthe ideal path the patient has to follow to bring the greendisc to the target through theHapticMasterWhen the patientdeviates too much from the ideal path the line and discbecome orange and finally red (as shown for the transportingexercise in Figure 4(b) to encourage the patient to correcther path Another example of feedback that is parameterizedis illustrated for the skill component rubbing in Figure 4(c)The patient has to move the disc between the two walls ofthe tube Haptic feedback makes the inside of the walls eithersmooth or course grained so that it is either easy or hard forthe patient to slide the disc next to the walls

Not only the parameters of the skill components can becustomized but also personalization can take place in the

gaming level Figure 5(a) shows the penguin painting gameTo make the lifting easier the weight of the penguins can beadjusted In the arkanoid game (Figure 5(b)) the viscosityof the bar pushing back the ball is also an adjustable hapticparameter

However the customization of exercise parameters isdependent on the therapistrsquos judgement and requires thetherapist to monitor the difficulty levels and preset param-eters during the training session and to intervene to adjust ifneededThere might be times when the therapist is not awareof the need for adjusting the training program according tothe patientrsquos progress or when the therapist finds a limitedtime to set up a new training program customized to thepatientrsquos current ability This dependency can be minimizedwithout a conscious effort from the therapist through theintegration of adaptivity Adaptivity enables the system toautomatically adjust itself in such a way to support thedifferent context of each patientrsquos rehabilitation trainingTailoring the right training challenge in the right timewithoutany interference from the therapist can be provided by auto-matically adjusting the difficulty level of the training exercisesaccording to the patientrsquos current training performance [37]

4 Towards Adaptive PersonalizedRehabilitation Training AutomaticDifficulty Level Adjustment

The focus in this work is how we can enhance our person-alization strategy by integrating adaptivity into the rehabili-tation training both in individual and collaborative trainingexercises forMS patients Due to the diversity ofMS patientsevery patient has a different physical ability and conditionwhich may influence the course of their training For exam-ple one patient might find it very difficult to perform aspecific training task while the other finds it quite easy toperform it We also have to take into account that everypatient progresses differently It could be the case that onepatient may progress slower than the other in performingthe rehabilitation training tasks With adaptation we aim tokeep the patients continue on training at their ease and ensureno major barriers inhibit their interaction during trainingFor that purpose we propose the adaptivity of automaticallyadjusting the difficulty level of the training exercise to beinvestigated in this study

To achieve an optimal training experience for MSpatients we were inspired by the flow theory of Csikszent-mihalyi [38] to keep the balance between difficulty level andpatient performance As illustrated in Figure 10 we wouldlike to make sure that a patient stays within the ldquooptimaltraining zonerdquo where the difficulty level of exercise given toa patient is balanced with his current performance In theoptimal zone the patient will not experience overtraining orundertraining [39] Overtraining happens when the patientis asked to perform the exercises with a high difficulty levelwhile hisher performance is still low thus the patient ismostlikely to find the training too difficult and may not be ableto perform the training On the other hand undertraininghappens when a low difficulty level is given to a patient who


Low

Low

Patient performance High

Hig

hD

ifficu

lty le

vel

Overtraining

Optimal tra

ining zone

Undertraining

Figure 10 Balancing the difficulty level and patient performance

has a high performance which makes the training not thatchallenging anymore

As an optimization strategy to achieve such personalizedtraining we propose providing the ability to automaticallyand dynamically adjust the difficulty of the exercise toavoid boredom provide suitable challenge and minimizethe therapistrsquos involvement as well This can be done bycreating automatic difficulty adjustments according to thepatientrsquos performance and progress in the exercise For thispurpose we need to establish a user model by capturing thepatientrsquos performance metrics (eg task completion timesscores and errors) during the exercises and making use ofthat information to infer the short-term training progress ofthe patient This can be considered as a sort of performance-evaluationmechanism Once established we can put the usermodel into practice by applying it to enable adapting thedifficulty level whenever necessary Based on the informationfrom the user model we can adjust the difficulty of theexercise by making it harder or easier

In this study we investigate the possibility of providingautomatic difficulty level adjustment in two training exercisesof our I-TRAVLE system namely the penguin painting gameand the social maze game through two user studies that wehave carried out with several MS patients

5 Adaptive Difficulty in Individual TrainingThe Penguin Painting Game

The penguin painting game is an individual training exercisethat was designed as part of the upper limb rehabilitationfor MS patients and focuses on training the skill componentsof lifting and transporting The goal of this game is to paintpenguins with the right color as many as possible withina certain time period (see Section 32) At the beginningof the training every patient starts from an initial level asshown in Figure 11 After two consecutive training sessionsthe training progress of the patient is determined based onhisher performance in each session

Figure 11 The penguin painting game initial level

If no significant difference of the performance is shownbetween the training sessions it is considered that the patientis training on an appropriate level and adaptation will notbe triggered If the patient shows a decrease in hisherperformance between the sessions a lower difficulty level willbe automatically offered to the patient in the next session Onthe other hand if an increase of the patientrsquos performanceis shown between the training sessions the system willautomatically provide a level with a higher difficulty in thenext session

We define seven different difficulty levels by altering thefollowing game parameters accordingly (see Table 1)

(i) size of penguin how big the penguins are (smalllarge)

(ii) speed of devil how fast the devil moves (slowmedi-umfast)

(iii) frequency of devil how frequent the devil appears(infrequentnormalfrequent)

(iv) length of stabilization the time required to hold thepenguin still (shortnormallong)

(v) obstacle wall addition of an obstacle wall along theway (noyes)

(vi) amount of coloring buckets how many coloring buck-ets exist (234)

(vii) width of coloring bucket how big the coloring bucketsare (narrowwide)

(viii) exit platform addition of another exit platform whichrequires patients to place the colored penguins to thesize-corresponding platforms (noyes)

As shown in Figure 12 three different levels are designedas the easy levels in the penguin painting game Level minus1(Figure 12(a)) is easier than the initial level Then one leveleasier is Level minus2 (Figure 12(b)) with Level minus3 (Figure 12(c))being the easiest level Figure 14 illustrates the three difficultlevels in the penguin painting game Level 1 (Figure 13(a)) ismore difficult than the initial level Then one level higher isLevel 2 (Figure 13(b)) with Level 3 (Figure 13(c)) being themost difficult level


Table 1 Overview of the penguin painting game parameters in the different difficulty levels

Game parameter Difficulty levelLevel minus3 Level minus2 Level minus1 Level 0 Level 1 Level 2 Level 3

Size of penguin All small 90 small 70 small 50 small 30 small 10 small All large10 large 30 large 50 large 70 large 90 large

Speed of devil Slow Slow Medium Medium Medium Fast FastFrequency of devil Infrequent Infrequent Normal Normal Normal Frequent FrequentLength of stabilization Short Short Normal Normal Normal Long LongObstacle wall No No No No No Yes YesAmount of coloring bucket 2 2 3 3 3 4 4

Width of coloring bucket All wide

Bottom 1wide 1narrow




Bottom 2wide 2narrow All narrow

Top 1 wide 1narrow

Top 2 wide 1narrow

Top 2 wide 1narrow

Top 1 wide 2narrow

Top Allnarrow

Exit platform No No No No Yes Yes Yes

(a) Level minus1 (b) Level minus2

(c) Level minus3

Figure 12 The easy levels in the penguin painting game

To determine the patientrsquos training progress it is neces-sary to obtain information of hisher training performanceand make a comparison over the last two training sessionsThe patientrsquos performance of each training session is calcu-lated as a function of the following performance metrics

(1) Task Completion TimeHowmuch time does the patient take to complete onetask (ie select and transport a penguin)How much is the slope of task completion times inone training session

(2) ScoreWhat score does the patient achieve in one gamesession

(3) ErrorHow many times does the patient make errors (iehitting the devil painting with the wrong color)

(4) PauseHowmany times does the patient make pause actions(ie motionless period between steps for longer than2 seconds)

(5) DistanceWhat is the distance traveled by the patient to com-plete one task (ie select and transport a penguin)Howmuch is the slope of the distance traveled in onetraining session


(a) Level 1 (b) Level 2

(c) Level 3

Figure 13 The difficult levels in the penguin painting game

51 User Study 1 Automatic Adjustment of Difficulty Levelin the Penguin Painting Game We have integrated theadaptation of automatic adjustment of difficulty levels inthe penguin painting game This results in seven difficultylevels which differ in the exercise parameters as described inSection 5 (see Table 1) We expect that supporting adaptivityin the adjustment of difficulty levels of the training exerciseswill not only deliver a personalized training to each MSpatient but also provide suitable challenge enable less bore-dom andminimize the therapistrsquos involvementTherefore wecarried out a user study to investigate the outcome of integrat-ing adaptive difficulty into the penguin painting game

511 Participants For the user studies in this research werecruited a group of participants which consists of ninepatients of the Rehabilitation and MS Centre in Overpelt(Belgium) who all suffer from upper limb dysfunction dueto MS and were willing to participate in both user studiesThe patients were 6 males and 3 females with an average ageof 60 years ranging from 47 to 72 years old The durationsince theMS diagnosis varies between 3 and 34 years with anaverage of 205 years Six of them used the left hand to operatethe HapticMaster in training while the other three used theirright hand Table 2 shows the personal information of eachMS patient participating in this research To have an overviewof the severity of their upper limb dysfunction we obtainedtheir clinical measures as shown in Table 3 upper limbstrength (motricity index [40]) upper limb functional capac-ity (action research arm test [41]) and arm motor functionscores (Brunnstrom Fugl-Meyer proximal and distal [42])

In this first user study only 8 patients of the group par-ticipated Patient 9 was unable to participate due to his healthcondition

Table 2 Personal information of MS patients in the user studies

Patient Gender Age (years) Diagnosisduration (years)

Training hand

1 Male 64 14 Left2 Female 58 3 Left3 Male 71 10 Right4 Female 47 14 Right5 Male 57 27 Left6 Male 55 27 Left7 Female 64 25 Right8 Male 56 30 Left9 Male 72 34 Left

Table 3 Clinical characteristics of MS patients in the user studies

Patient MI ARAT BFM-prox BFM-dist1 76 41 25 402 83 56 36 293 84 46 32 284 76 56 36 305 55 41 23 216 47 7 18 247 72 18 31 258 60 30 27 249 50 1 12 7MI motricity index (maximal score = 100) ARAT action research armtest (maximal score = 57) BFM-prox Brunnstrom Fugl-Meyer proximalscore (maximal score = 66) BFM-dist Brunnstrom Fugl-Meyer distal score(maximal score = 66)


(a) The patient trained at the initial level (b) The patientrsquos training wasadapted to a higher difficultylevel

Figure 14 The setup of user study automatic difficulty level adjustment in the penguin painting game (eg 2 exit platforms in the difficultlevel)

512 Procedure The user study consisted of seven sessionstwo elicitation sessions and five adaptive sessions all tookplace on the same day In the elicitation sessions participantswere asked to perform the penguin painting game in theinitial level (see Figures 11 and 14(a)) After two elicitationsessions the performance metrics of the participant werecalculated to determine the progress of hisher trainingBased on the information about the training progress overthese elicitation sessions it will be determined for the firstadaptive session whether or not the difficulty level shouldbe adapted Throughout the adaptive sessions participantswere offered an adaptive personalized training in terms of theadjustment of difficulty levels Three possibilities can happenover the course of five adaptive sessions stay at the same levelgo one level lower or go one level higher (see Figure 14(b))

The duration of each session is 3 minutes After eachadaptive session participants were asked to rate their sub-jective perception on enjoyment boredom challenge frus-tration and fun on a 5-point scale rating (eg 1 not at allto 5 very much) based on their experience of performingthe adaptive penguin painting game Averagely the userstudy lasted for about 30 minutes per participant Figure 14illustrates the setup of this user study

513 Results We have applied the adaptation of automaticdifficulty level adjustment in the penguin painting gamewhich provided an adaptive personalized training for eachparticipant Consequently the training trajectory was differ-ent for every participant during the five adaptive sessionsFor each session the participant can experience staying atthe same difficulty level going to a lower level or goingto a higher level depending on hisher individual perfor-mance Figure 15 shows the personalized training trajectoryfor each patient as a result of integrating the adaptivedifficulty level adjustment in the penguin painting gameAs can be observed no patient had the same trajectoryas the other patient due to the fact that every patientprogressed differently This finding confirms the need foradaptation Further we analyzed how adaptation influenced

the subjective perception on enjoyment boredom challengefrustration and fun across the sessions Due to the smallnumber of samples and observations in this user study weused the nonparametric methods for the statistical analysis

Based on the patientsrsquo subjective responses we calculatedthe average ratings of enjoyment boredom challenge frus-tration and fun for the three conditions of adaptation (go toa lower level stay at the same level and go to a higher level)as shown by Figure 16 Kruskal-Wallis test showed that nosignificant differences were found for Enjoyment Frustrationand Fun between the different conditions of adaptationsThisindicates that patients perceived the same level of enjoymentfrustration and fun eventhough the system introduced anautomatic adaptation of difficulty levels in the trainingexercises Patients rated a high level of enjoyment and fun(above 4) and a low level of frustration (below 2) in all theconditions

However there is a significant difference found for Bore-dom (H(2) = 15651 119875 lt 0001 2 for condition 1 133 forcondition 2 and 1 for condition 3) and Challenge (H(2) =24376 119875 lt 0001 2 for condition 1 289 for condition 2 and425 for condition 3) Mann-Whitney pairwise comparisontests showed that patients felt significantly less bored andmore challenged when the training was adapted to a higherlevel compared to when they had to adapt to a lower level orstayed at the same level (119875 lt 0001)

Furthermore we observed that some patients havenoticed the automatic adaptation to be related to theirtraining progress and they liked the diversity of difficultylevels We did not inform about the implemented adaptationto the patients before or during the user study A couple oftherapists appreciated the automatic adaptation as it providedthe patients with more variety in the training and alsogave them more freedom to train on their own withoutany interference from the therapist to manually adjust theexercise parameters This kind of adaptation could be usefulto determine an appropriate level to start training on a certainday according to the patients condition on that day thus lessdetermined by the previous training or the therapist


Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Figure 15 Adaptive personalized training trajectory in the penguin painting game


5

45

4

35

3

25

2

15

1

05

0

Enjoyment Boredom Challenge Frustration Fun

Mea

n ra

ting

Lower levelSame levelHigher level

Figure 16 Patientrsquos subjective rating with respect to adaptation

Figure 17 The social maze game initial level

6 Adaptive Difficulty in CollaborativeTraining The Social Maze Game

The social maze game is a collaborative training exercise thatwas designed to train several skill components such as liftingtransporting turning pushing and reachingThe goal of thisgame is to collect all symbols by picking up each symbol andbringing it to the collecting bin (see Section 33) To achievethis the patient should pair and closely collaborate withhisher training partner (eg a therapist family member orfellow patient) At the beginning of the training every pairstarts from an initial level as depicted in Figure 17

The same performance-evaluation and adaptation mech-anism with the penguin painting game is implemented Firstwe obtain an indication of the patientrsquos training progressby comparing the training performance of the last twotraining sessions When no training progress is indicatedno adaptation will be triggered and the pair will stay at thesame level When a training progress is indicated a higherdifficulty level will be automatically offered to the pair inthe next session When a training decline is indicated thesystem will automatically switch to a lower difficulty levelin the next session It is important to mention here thatalthough collaborative training exercises involve two personswe mainly focus the adjustment of difficulty level based on

the patientsrsquo progress in the training since they hold the corefunction in the rehabilitation Particularly in the sympatheticscenario we do not take into account the progress of a healthyperson as the training partner since hisher role is solelysupporting the rehabilitation training of the patient

We define eight difficulty levels ranging from very easy tovery difficult As shown in Figure 18 three different levels aredesigned as the easy levels in the social maze game with Levelminus3 (Figure 18(c)) being the easiest level Figure 19 illustratesthe four difficult levels in the social maze game with Level 4(Figure 19(d)) being the most difficult level

To adjust the difficulty levels we alter the following gameparameters in each level accordingly (see Table 4)

(i) viscosity of movement how high the viscosity is (verylowlownormalhighvery high)

(ii) speed of devil how fast the devil moves(slowmediumfast)

(iii) length of laser beam how long the laser beam lasts(shortnormallong)

(iv) friction of bomb the degree of friction that the bombhas (no frictionless frictionmuch friction)

(v) friction of wall the degree of friction that the wall has(no frictionless frictionmuch friction)

In the social maze game four performance metrics areemployed to determine the patientrsquos training progress asfollows

(1) SpeedHow fast can the patient complete one task (ie selectand transport a symbol)


(3) ErrorHow many times does the patient make errors (iehitting the devil hitting the laser beam)


61 User Study 2 Automatic Adjustment of Difficulty Level inthe Social Maze Game We have integrated the adaptation ofautomatic adjustment of difficulty levels in the social mazegame To investigate how patients perceive the outcome ofthis adaptation we carried out the second user study with ourgroup of participants as described in Section 511

611 Participants All nine patients of the group participatedin this user study More information on these participantscan be found in Table 2 for their personal information andTable 3 for their clinical characteristics In this user studywe applied the sympathetic scenario of social rehabilitationtraining where a patient collaborates with hisher therapist in



(c) Level minus3

Figure 18 The easy levels in the social maze game


(c) Level 3 (d) Level 4

Figure 19 The difficult levels in the social maze game

Table 4 Overview of the social maze game parameters in the different difficulty levels

Game parameter Difficulty levelLevel minus3 Level minus2 Level minus1 Level 0 Level 1 Level 2 Level 3 Level 4

Viscosity of movement Very low Low Low Normal Normal High High Very highSpeed of devil Slow Slow Slow Medium Medium Fast Fast FastLength of laser beam Short Short Short Normal Normal Long Long LongFriction of bomb No No No Less Less Much Much MuchFriction of wall No No No Less Less Much Much Much


Figure 20 The setup of user study automatic difficulty level adjustment in the social maze game

performing the social maze game Only one therapist whomevery patient has a good relationshipwith participated in thisuser studyTheNovint Falconwas used as the input device forthe therapist All patients used the HapticMaster as the inputdevice

612 Procedure Similar to the previous study describedin Section 512 the user study consisted of seven sessionstwo elicitation sessions and five adaptive sessions In theelicitation sessions pairs of participants (ie one patient andthe therapist) were asked to perform the social maze game inthe initial level (see Figure 17) After two elicitation sessionsthe performance metrics of the patient were calculated todetermine the progress of hisher training Depending on theobservation of the training progress the system will deter-mine how the difficulty level should be adjusted throughoutthe adaptive sessions There can be three possible results ofthe adaptation stay at the same level go one level lower or goone level higher

After each adaptive session patients were asked to ratetheir subjective perception on difficulty and enjoyment on a5-point scale rating (eg 1 not at all to 5 very much) basedon their experience of performing the social maze gameAveragely the user study lasted for about 30 minutes perparticipant Figure 20 illustrates the setup of this user study

613 Results We have applied the adaptation of automaticdifficulty level adjustment in the social maze game Sincethe adaptation resulted in a personalized training eachpatient had a unique training trajectory which differs fromeach other during the adaptive sessions Figure 21 shows thepersonalized training trajectory for each pair of participantsas a result of integrating adaptive difficulty in the social mazegame No pair of participants had exactly the same trajectoryas the other which confirms the need for adaptation basedon the patientrsquos individual performance We analyzed howadaptation influenced the subjective perception on difficultyand enjoyment across the sessions We used the nonpara-metric statistics because of the small number of samples andobservations in this user study

Based on the patientsrsquo subjective responses we calculatedthe average ratings of perceived difficulty and perceivedenjoyment for the three conditions of adaptation as shownby Figure 22 Kruskal-Wallis test showed that significantdifferences were found for perceived difficulty (H(2) = 13062119875 lt 0001 164 for condition 1 208 for condition 2 and296 for condition 3) and perceived enjoyment (H(2) = 9439119875 lt 005 391 for condition 1 367 for condition 2 and309 for condition 3) This indicates that patients perceivedthe difficulty and enjoyment differently across the conditionsof adaptation Mann-Whitney pairwise comparison testsshowed that patients perceived the training to be moredifficult and less enjoyable when the training was adapted toa higher level compared to when they had to adapt to a lowerlevel (119875 lt 005)

The finding which showed that as the levels got more dif-ficult the patients perceived the training to be less enjoyable issomewhat different than our previous finding where patientsperceived the samehigh level of enjoyment despite the changein difficulty levels We think that social effect plays a rolein this case In a collaborative training session the patientsmight have felt more social pressure to perform as good asthe training partner They might have feared that they willhinder the collaboration if they perform less well Howeveradditional investigation is needed to confirm this thoughtAnother possible explanation is that the differences amongthe difficulty levels were quite prominent This calls for afurther investigation to find a subtler difference of difficultythat maintains the patientrsquos enjoyment in training

With regard to social interaction we believe that thesocial maze game has become a social play medium betweenpatients and their therapist We have observed some rele-vant behaviors which indicated the development of socialinteraction during the collaborative rehabilitation trainingFigure 23 depicts two examples of behaviors shown by thepairs of participants throughout the training sessions Themost shown behavior during the social maze game was theact of discussing strategy (Figure 23(a)) It was pretty obviousthat the nature of the training exercise requires the twoparticipants to closely collaborate and discuss their strategyand necessary actions This behavior happened throughoutthe whole session sometimes followed by the act of looking


Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9

Figure 21 Adaptive personalized training trajectory in the social maze game


45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting


Perceived enjoymentPerceived difficulty


(a) Discussing strategy at the end of a session (b) Mimicking each other

Figure 23 Different behaviors observed during the social maze game

at each other At the beginning and during the sessionparticipants discussed what actions they should performand the best way to perform them At the end of thesession participants briefly reviewed their previous sessionand how they should perform better on the next sessionIn some participants we can also observe the behavior ofunconsciously mimicking each other during the strategydiscussion (Figure 23(b)) Throughout the sessions we canobserve that all participants showed other behaviors suchas smiling laughing and chuckling This mostly happenedwhen one of them made an error such as encountering thedevil or hitting the laser beam A couple of participantstended tomake jokes during the training session that resultedin both of them laughing at each other

7 Discussion

Our work has presented an investigation to integrate adaptiv-ity and personalization into individual rehabilitation trainingand collaborative rehabilitation training for MS patientsTypically rehabilitation training is individual where a MSpatient performs the training exercises alone under thesupervision of a therapist However rehabilitation trainingcan take place in a collaborative setting as well whereperforming the training exercises involves more than only asingle MS patient By changing the nature of rehabilitationtraining to be collaborative we can enhance the patientrsquos

motivation and also support social interaction between thepatient and hisher training partner (eg a family memberfriend or therapist)

We have discussed and implemented the adaptationof automatic difficulty level adjustment in both individualtraining (ie the penguin painting game) and collaborativetraining (ie the social maze game) Two user studies havebeen carried out to investigate the outcome of this adaptationOverall we can conclude that providing adaptive difficultyin the training exercises has delivered an adaptive person-alized training to each MS patient according to hisher ownindividual training progress Patients and the therapist haveappreciated the automatic adaptation of difficulty levels andconsidered it to provide more variety in the training and alsogive the patients more freedom to train on their own withoutany interference from the therapist to manually adjust theexercise parameters This kind of adaptation could be usefulto determine an appropriate level to start training on a certainday according to the patientrsquos condition on that day thus lessdetermined by the previous training or the therapist

Moreover we have observed the development of socialinteraction between patients and their training partner dur-ing the course of the collaborative rehabilitation trainingexercises While performing the social maze game collabo-ratively they have showed several particular behaviors suchas smiling laughing or chuckling looking at each otherand discussing strategy We realized that the adaptation in


the collaborative training exercise has been implementedwith a focus on the patientrsquos needs and characteristics despitethe existence of a training partner In the sympathetic socialscenario it may be interesting to investigate how or whatkinds of adaptation can facilitate the family member orthe therapist to achieve a more engaging and motivatingcollaboration with the patient For the empathetic socialscenario we also need to investigate how the different typesof adaptation can accommodate the needs and characteristicsof another patient as a training partner

It is not our focus to carry out an in-depth investigationon the adaptation algorithm used in these user studies Wewere more interested to observe the patientsrsquo response withrespect to the automatic adjustment of difficulty levels Werealize that amore accurate andwell-defined algorithm couldbe provided Therefore further investigation is needed tooptimize the adaptation algorithm which also matches thejudgment of therapists on the trigger and timing of adapta-tion Another next step is to extend our investigation of inte-grating adaptivity and personalization into MS rehabilitationtraining to other types of adaptation that may help patientsduring the course of their training for example automaticallyadjusting the assistance level based on the detected musclefatigue In some cases the muscle fatigue might developduring long training thus it might be necessary to providethe patients with some assistance to help them perform thetask and continue their training

8 Conclusion

This paper has discussed our investigation on the notionof integrating adaptivity and personalization into individualand collaborative rehabilitation training for MS patients Wehave strived for developing adaptive personalized traininggames by introducing the automatic adjustment of difficultylevels in the penguin painting game (individual training)and the social maze game (collaborative training) The userstudies showed that none of the MS patients experiencedthe same training trajectory as the other This confirms thatevery patient progresses differently which encourages theneed for adaptation Adaptive personalized training whichwas delivered to every MS patient according to hisher ownindividual training progress has shown to be beneficial andmuch appreciated The automatic adjustments of difficultylevels in the training games were also considered to providemore variety in the training and more freedom to trainindependently

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors acknowledge the INTERREG-IV program inparticular ldquorehabilitation robotics IIrdquo (Project no IVA-VLANED-114 Euregio Benelux) and ldquoI-TRAVLErdquo (Project

no IVA-VLANED-158 and the consortium partners (seehttpi-travletumblrcom) The authors greatly appreciateGeert Alders for his valuable insights Tom De Weyer for hisabundant help and Karel Robert for his graphical designTheauthors thank all participants for their contribution to theuser studies

References

[1] S Jezernik R Scharer G Colombo and M Morari ldquoAdaptiverobotic rehabilitation of locomotion a clinical study in spinallyinjured individualsrdquo Spinal Cord vol 41 no 12 pp 657ndash6662003

[2] L E Kahn W Z Rymer and D J Reinkensmeyer ldquoAdaptiveassistance for guided force training in chronic strokerdquo inProceedings of the 26th Annual International Conference of theIEEE Engineering in Medicine and Biology Society (EMBC rsquo04)pp 2722ndash2725 San Francisco Calif USA September 2004

[3] P Kan RHuq JHoey RGoetschalckx andAMihailidis ldquoThedevelopment of an adaptive upper-limb stroke rehabilitationrobotic systemrdquo Journal of NeuroEngineering andRehabilitationvol 8 no 1 article 33 2011

[4] S Jezernik G Colombo and M Morari ldquoAutomatic gait-pattern adaptation algorithms for rehabilitation with a 4-DOFrobotic orthosisrdquo IEEE Transactions on Robotics and Automa-tion vol 20 no 3 pp 574ndash582 2004

[5] M T Schultheis and A A Rizzo ldquoThe application of virtualreality technology in rehabilitationrdquo Rehabilitation Psychologyvol 46 no 3 pp 296ndash311 2001

[6] M K Holden ldquoVirtual environments for motor rehabilitationreviewrdquo Cyberpsychology amp Behavior vol 8 no 3 pp 187ndash2112005

[7] J S Webster P T McFarland L J Rapport B Morrill LA Roades and P S Abadee ldquoComputer-assisted training forimproving wheelchair mobility in unilateral neglect patientsrdquoArchives of Physical Medicine and Rehabilitation vol 82 no 6pp 769ndash775 2001

[8] D L Jaffe D A Brown C D Pierson-Carey E L Buckley andH L Lew ldquoStepping over obstacles to improve walking in indi-viduals with poststroke hemiplegiardquo Journal of RehabilitationResearch and Development vol 41 no 3 pp 283ndash292 2004

[9] J A Edmans J R F Gladman S Cobb et al ldquoValidity of avirtual environment for stroke rehabilitationrdquo Stroke vol 37 no11 pp 2770ndash2775 2006

[10] G Alankus A Lazar M May and C Kelleher ldquoTowardscustomizable games for stroke rehabilitationrdquo in Proceedingsof the 28th Annual CHI Conference on Human Factors inComputing Systems (CHI rsquo10) pp 2113ndash2122 Atlanta Ga USAApril 2010

[11] G Saposnik R Teasell MMamdani et al ldquoEffectiveness of vir-tual reality usingwii gaming technology in stroke rehabilitationa pilot randomized clinical trial and proof of principlerdquo Strokevol 41 no 7 pp 1477ndash1484 2010

[12] M Ma M McNeill D Charles et al ldquoAdaptive virtual realitygames for rehabilitation of motor disordersrdquo in UniversalAccess in Human-Computer Interaction Ambient Interaction CStephanidis Ed vol 4555 of Lecture Notes in Computer Sciencepp 681ndash690 Springer 2007

[13] M S Cameirao S Bermudez i Badia E D Oller and P F MJ Verschure ldquoUsing a multi-task adaptive VR system for upperlimb rehabilitation in the acute phase of strokerdquo in Proceedings


of the International Conference on Virtual Rehabilitation (ICVRrsquo08) pp 2ndash7 Vancouver Canada 2008

[14] O Barzilay and A Wolf ldquoAn adaptive virtual biofeedbacksystem for neuromuscular rehabilitationrdquo in Proceedings of the11th IUPESMWorld Congress onMedical Physics and BiomedicalEngineering pp 1291ndash1294Munich Germany September 2009

[15] M Duff Y Chen S Attygalle et al ldquoAn adaptive mixed realitytraining system for stroke rehabilitationrdquo IEEE Transactions onNeural Systems and Rehabilitation Engineering vol 18 no 5 pp531ndash541 2010

[16] M Baran N Lehrer D Siwiak et al ldquoDesign of a home-based adaptive mixed reality rehabilitation system for strokesurvivorsrdquo in Proceedings of the 33rd Annual International Con-ference of the IEEE Engineering in Medicine and Biology Society(EMBS rsquo11) pp 7602ndash7605 BostonMass USA September 2011

[17] U Dalgas E Stenager and T Ingemann-Hansen ldquoMulti-ple sclerosis and physical exercise recommendations for theapplication of resistance- endurance- and combined trainingrdquoMultiple Sclerosis vol 14 no 1 pp 35ndash53 2008

[18] R W Motl and J L Gosney ldquoEffect of exercise training onquality of life in multiple sclerosis a meta-analysisrdquo MultipleSclerosis vol 14 no 1 pp 129ndash135 2008

[19] G Kwakkel R CWagenaar J W R Twisk G J Lankhorst andJ C Koetsier ldquoIntensity of leg and arm training after primarymiddle-cerebral-artery stroke a randomised trialrdquo The Lancetvol 354 no 9174 pp 191ndash196 1999

[20] G Kwakkel B J Kollen and H I Krebs ldquoEffects of robot-assisted therapy on upper limb recovery after stroke a system-atic reviewrdquo Neurorehabilitation and Neural Repair vol 22 no2 pp 111ndash121 2008

[21] J H Burridge and A-M Hughes ldquoPotential for new technolo-gies in clinical practicerdquo Current Opinion in Neurology vol 23no 6 pp 671ndash677 2010

[22] K Coninx C Raymaekers J D Boeck et al ldquoUsing thephantom device for rehabilitation of the arm in MS patientsa case studyrdquo in Proceedings of the 6th International ConferenceonMethods and Techniques in Behavioral Research (MeasuringBehavior rsquo08) pp 148ndash149 Maastricht The Netherlands 2008

[23] P Feys G Alders D Gijbels et al ldquoArm training in multiplesclerosis using phantom clinical relevance of robotic outcomemeasuresrdquo in Proceedings of the 11th IEEE International Confer-ence on Rehabilitation Robotics (ICORR rsquo09) pp 576ndash581 KyotoJapan June 2009

[24] I-TRAVLE ldquoRevalidatierobotica II an Interreg IV projectrdquo2011 httpwwwi-travleeu

[25] T de Weyer S Notelaers K Coninx et al ldquoWatering theflowers virtual haptic environments for training of forearmrotation in persons with central nervous deficitsrdquo in Proceedingsof the 4th International Conference on Pervasive TechnologiesRelated to Assistive Environments (PETRA rsquo11) pp 1ndash6 CreteGreece 2011

[26] S Notelaers T de Weyer C Raymaekers K Coninx HBastiaens and I Lamers ldquoData management for multimodalrehabilitation gamesrdquo in Proceedings of the 21st InternationalWorkshop on Database and Expert Systems Applications (DEXArsquo10) pp 137ndash141 Bilbao Spain September 2010

[27] W M van den Hoogen I Lamers K Baeten et al ldquoEffects ofshading and droplines on object localization in virtual rehabili-tation for patients with neurological conditionsrdquo in Proceedingsof the International Conference on Virtual Rehabilitatio (ICVRrsquo11) pp 1ndash6 Zurich Switzerland June 2011

[28] W van den Hoogen P Feys I Lamers et al ldquoVisualizing thethird dimension in virtual training environments for neuro-logically impaired persons beneficial or disruptiverdquo Journal ofNeuroEngineering and Rehabilitation vol 9 article 73 2012

[29] HWoldag andHHummelsheim ldquoEvidence-based physiother-apeutic concepts for improving arm andhand function in strokepatients a reviewrdquo Journal of Neurology vol 249 no 5 pp 518ndash528 2002

[30] A A A Timmermans R P J Geers J A Franck et alldquoT-TOAT a method of task-oriented arm training for strokepatients suitable for implementation of exercises in rehabilita-tion technologyrdquo in Proceedings of the IEEE 11th InternationalConference on Rehabilitation Robotics (ICORR rsquo09) pp 98ndash102Kyoto Japan June 2009

[31] L Vanacken S Notelaers C Raymaekers et al ldquoGame-basedcollaborative training for arm rehabilitation of MS patients aproof-of-concept gamerdquo in Proceedings of GameDays pp 65ndash75 2010

[32] W van denHoogenW IJsselsteijn andY deKort ldquoYesWii canUsing digital games as a rehabilitation platform after strokemdashthe role of social supportrdquo in Proceedings of the InternationalConference on Virtual Rehabilitation (ICVR rsquo09) p 195 HaifaIsrael July 2009

[33] J Decety and T Chaminade ldquoNeural correlates of feelingsympathyrdquo Neuropsychologia vol 41 no 2 pp 127ndash138 2003

[34] J Decety and P L Jackson ldquoThe functional architectureof human empathyrdquo Behavioral and Cognitive NeuroscienceReviews vol 3 no 2 pp 71ndash100 2004

[35] T de Weyer K Robert J Hariandja G Alders and K ConinxldquoThe social maze a collaborative game to motivate ms patientsfor upper limb trainingrdquo in Proceedings of the 11th internationalconference on Entertainment Computing pp 476ndash479 SpringerBerlin Germany 2012

[36] S Notelaers T de Weyer J Octavia K Coninx H Bastiaensand P Feys ldquoIndividualized robot-assisted training for ms- andstroke patients in i-travlerdquo BIOWeb of Conferences vol 1 article00069 2011

[37] J R Octavia K Coninx and P Feys ldquoAs i am not youaccommodating user diversity through adaptive rehabilitationtraining for multiple sclerosis patientsrdquo in Proceedings of the24th Australian Computer-Human Interaction Conference pp424ndash432 ACM New York NY USA 2012

[38] M Csikszentmihalyi Flow The Psychology of Optimal Experi-ence Harper amp Row 1990

[39] M L OrsquoToole ldquoOverreaching and overtraining in enduranceathletesrdquo in Overtraining in Sport R B Kreider A C Fry andM L OrsquoToole Eds pp 3ndash18 Human Kinetics 1998

[40] D T Wade ldquoMeasuring arm impairment and disability afterstrokerdquo Disability and Rehabilitation vol 11 no 2 pp 89ndash921989

[41] W J G de Weerdt ldquoMeasuring recovery of arm-hand functionin stroke patients a comparison of the Brunnstrom-Fugl-Meyertest and the action research arm testrdquo Physiotherapy Canadavol 37 no 2 pp 65ndash70 1985

[42] P W Duncan M Propst and S G Nelson ldquoReliability ofthe Fugl-Meyer assessment of sensorimotor recovery followingcerebrovascular accidentrdquo Physical Therapy vol 63 no 10 pp1606ndash1610 1983

Submit your manuscripts athttpwwwhindawicom

Neurology Research International

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Alzheimerrsquos DiseaseHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


BioMed Research International


Research and TreatmentSchizophrenia

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Neural Plasticity


Parkinsonrsquos Disease


Research and TreatmentAutism

Sleep DisordersHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Neuroscience Journal

Epilepsy Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Psychiatry Journal


Computational and Mathematical Methods in Medicine

Depression Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Brain ScienceInternational Journal of

StrokeResearch and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Neurodegenerative Diseases


Journal of

Cardiovascular Psychiatry and NeurologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


it can improve as a result of the treatment and rehabilitationeffortsThis change of condition should be taken into accountto ensure providing the right level of rehabilitation trainingto the patient at the right time Therefore due to thediversity among MS patients it would be unwise to offer thesame rehabilitation training to every patient This situationdemands a suited personalized rehabilitation training tomeet every patientrsquos needs

To acquire a good result of rehabilitation it is necessaryto maintain patient motivation Generally rehabilitationinvolves the same training exercises that should be performedrepetitively and for a long period of time Using games asthe training platform has been considered to maintain andenhance patientsrsquo motivation during rehabilitation especiallycollaborative games in which social interaction is incorpo-rated However the usage of game-like training exercise is notthe ultimate solution Some patients may feel less motivatedwhen finding the game to be too easy or too difficultor when reaching a certain point in the training wherethey become bored with the game Therefore rehabilitationtraining should be set at an appropriate level of challenge ordifficulty to maintain the motivation of patients This raisesthe need of adaptivity in the rehabilitation training to ensurethe effectiveness of the rehabilitation

Our work focuses on the idea of integrating adaptivityand personalization into the rehabilitation training for MSpatients and investigating to what extent the adaptive person-alized training may contribute to a successful neurologicalrehabilitation This paper firstly discusses the survey ofrelated work concerning adaptation in rehabilitation train-ing followed by a description of our rehabilitation systemdeveloped to support personalized rehabilitation trainingfor MS patients We further elaborate on the investigationof integrating the adaptivity of automatic difficulty leveladjustment in the rehabilitation training for MS patientsincluding the user studies carried out with a group of MSpatients to acquire their feedback

2 Related Work

In this section we first provide a brief overview of researchwhich has studied adaptation in rehabilitation training Wethen discuss several works which have investigated the usageof virtual environments as the platform for rehabilitationtraining Lastly we describe a number of studies whichcombined both and attempted to integrate adaptation invirtual environments for rehabilitation training

21 Adaptation in Rehabilitation Training The integration ofadaptation in rehabilitation training has been the focus ofseveral studies [1ndash4] These studies have mainly investigatedthe integration of adaptivity in robot-assisted rehabilitationtraining which aimed at providing a personalized trainingto the patients according to their individual characteristicsneeds and abilities Moreover the adaptivity is intendedto facilitate an automated training system to minimize thetherapistrsquos effort in manually adjusting the rehabilitationtraining

Jezernik et al [1 4] studied the adaptation in rehabilita-tion training of locomotion for stroke and spinal cord injuredpatients An automated treadmill training system was intro-duced using a robotic rehabilitation device to increase thetraining duration and reduce the physiotherapistsrsquo effort Aclinical study on six spinal cord injured patients showed thatthe treadmill training with adaptive gait patterns increasesthe motivation of the patient and gives himher the feelingthat they are controlling themachine rather than themachinecontrolling them Kahn et al [2] described the integrationof adaptive assistance into guided force training as part ofthe upper extremity rehabilitation for chronic stroke patientsAn adaptive algorithm was developed to individually tailorthe amount of assistance provided in completing the trainingtask This algorithm has been evaluated with one patientin a two-month training program which showed significantimprovements in the patientrsquos arm function reflected by theperformance increase of functional activities of daily livingsuch as tucking a shirt and stabilizing a pillow Kan et al [3]presented an adaptive upper-limb rehabilitation robotic sys-tem for stroke patients which accounts for the specific needsand abilities of different patients Using the decision theo-retic model the system autonomously facilitates upper-limbreaching rehabilitation by tailoring the exercise parametersand estimating the patientrsquos fatigue based on the observationin hisher compensation or control ofmovementsThe systemperformance was evaluated by comparing the decisionsmadeby the system with those of a human therapist Overall thetherapist agreed with the system decisions approximately65 of the time and also thought the system decisions werebelievable and could envision this system being used in botha clinical and home setting

22 Virtual Environments for Rehabilitation Training An-other emerging technology that has been widely applied inrehabilitation is the virtual environment technology Vir-tual environments provide a variety of potential benefitsfor many aspects of rehabilitation training Schultheis andRizzo [5] discussed a number of advantages for the use ofvirtual environments in rehabilitation One key benefit isthat virtual environments are rather naturalistic or ldquoreal-liferdquo environments which may allow the users or patientsto be immersed within the environment and ldquoforgetrdquo thatthey are in a rehabilitation training session Within virtualenvironments patients are also facilitated to perform therehabilitation training tasks using 3D interaction which givesa close resemblance to the actual movements in the realworld Schultheis and Rizzo [5] also pointed out that virtualenvironments facilitate the design of individualized trainingenvironments where therapists can better tailor the trainingexercises based on an individualrsquos abilities and needs andthey can also easily apply gradual increments of difficulty andchallenge Another benefit is that virtual environments allowthe introduction of gaming factors into the rehabilitationscenario to enhance motivation of patients

Holden [6] provided a thorough overview of the useof virtual environments in the field of motor rehabilitationIn the context of motor rehabilitation several studies have


























(a) Basic training
































































Low

Low


Hig

hD

ifficu

lty le

vel

Overtraining

Optimal tra

ining zone

Undertraining






























Top 1 wide 1narrow

Top 2 wide 1narrow

Top 2 wide 1narrow

Top 1 wide 2narrow

Top Allnarrow



(c) Level minus3










(c) Level 3







Training hand















Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8



5

45

4

35

3

25

2

15

1

05

0


Mea

n ra

ting
























(c) Level minus3


















Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of



























(a) Basic training
































































Low

Low


Hig

hD

ifficu

lty le

vel

Overtraining

Optimal tra

ining zone

Undertraining






























Top 1 wide 1narrow

Top 2 wide 1narrow

Top 2 wide 1narrow

Top 1 wide 2narrow

Top Allnarrow



(c) Level minus3










(c) Level 3







Training hand















Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8



5

45

4

35

3

25

2

15

1

05

0


Mea

n ra

ting
























(c) Level minus3


















Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of






























































Low

Low


Hig

hD

ifficu

lty le

vel

Overtraining

Optimal tra

ining zone

Undertraining






























Top 1 wide 1narrow

Top 2 wide 1narrow

Top 2 wide 1narrow

Top 1 wide 2narrow

Top Allnarrow



(c) Level minus3










(c) Level 3







Training hand















Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8



5

45

4

35

3

25

2

15

1

05

0


Mea

n ra

ting
























(c) Level minus3


















Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of













Top 1 wide 1narrow

Top 2 wide 1narrow

Top 2 wide 1narrow

Top 1 wide 2narrow

Top Allnarrow



(c) Level minus3










(c) Level 3







Training hand















Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8



5

45

4

35

3

25

2

15

1

05

0


Mea

n ra

ting
























(c) Level minus3


















Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of




(c) Level 3







Training hand















Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8



5

45

4

35

3

25

2

15

1

05

0


Mea

n ra

ting
























(c) Level minus3


















Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of













Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8



5

45

4

35

3

25

2

15

1

05

0


Mea

n ra

ting
























(c) Level minus3


















Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of



5

45

4

35

3

25

2

15

1

05

0


Mea

n ra

ting
























(c) Level minus3


















Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of




(c) Level minus3


















Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of












Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 1

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 3

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 5

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 7

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 2

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 4

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 6

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 8

Diffi

culty

leve

l

Adaptive session

3

4

2

1

0

minus1

minus2

minus3

Start 1 2 3 4 5

Patient 9



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of



45

4

35

3

25

2

15

1

05

0

Mea

n ra

ting







7 Discussion








8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of





8 Conclusion




Acknowledgments



References

























































Neural Plasticity










Psychiatry Journal









Journal of













































Neural Plasticity










Psychiatry Journal









Journal of


clinical study adaptive personalized training games for...

Documents