PostureCheck: A vision-based compensatory-posture-detection tool to enhance performance of the BURT upper-extremity stroke-therapy device

SummaryAbstractThis Small Business Innovation ResearchSBIRPhase I project proposes the development of an imageprocessing based toolnamed PostureCheckaimed at automatically detecting when patients perform undesirable compensatory movements during robot assisted upper limb rehabilitation exercisesThe system will be based on a standard video camerae gGoProthat will be used to capture the movements of the subjectThe automatic detection of undesirable compensatory movements is especially important when patients use a rehabilitation robotic system with minimum supervisioni ewhen a single therapist oversees the therapeutic sessions of multiple patients simultaneouslyIn this contextPostureCheckmay be capable of tracking robot assisted rehabilitation exercises and enable feedback modalities to discourage the performance of undesirable compensatory movementsOur long term goal is to integrate PostureCheckwith the Barrett Upper extremity Robotic TrainerBURTwhich we developed with special emphasis on stroke rehabilitationThe combination of PostureCheckwith the BURTdevice would be ideally suited for deployment inRobotic Gymswhere a single therapist oversees the therapeutic sessions of several patients simultaneouslythus allowing rehabilitation centers to offer high dosagehigh intensity interventions despite the limited number of therapists currently available in the USTo demonstrate the feasibility of the proposed conceptwe will develop PostureCheckto detect the most common compensatory movements automaticallyTo achieve this goalwe will rely on recently developed artificial intelligenceAImethods referred to as Deep LearningThese methods have recently broken records in the human posture analysisjoint skeleton detectionand recognition of human activities using a single inexpensive cameraThe proposed video based PostureChecktool will be the first system to exploit the capabilities of hybrid Deep Neural Networksfor real time detection of compensatory movements during robotassisted rehabilitationThe proposed SBIR Phase I activities are organized in three aimsIn Aimfeedback from rehabilitation experts at Spaulding Rehabilitation Hospital will be used to collect video data and to label compensatory movements observed during the performance of robot assisted rehabilitation exercises by using the BURTsystemIn AimDeep Learning techniques will be used to develop a robust detection of undesirable compensatory movements during the performance of robot assisted rehabilitation exercisesFinallyin Aimthe algorithms developed in Aimwill be optimizedSpecificallywe will test implementations that are suitable to generate real time feedbackComputationally efficient implementations of the algorithms will enablein future studiesthe development of new modalities of control of the rehabilitation robot with the objective of discouraging undesirable compensatory movements Project NarrativeEach yearnearlyAmericans suffer from a type of stroke that particularly weakens one side of the bodyDuring upper limb rehabilitationappropriate feedback from a therapist to discourage stroke survivors from performing undesirable compensatory movements results in better motor recovery andeventuallyimproved functionThis proposal aims to develop a novel video based toolnamed PostureCheckto detect undesirable compensatory movements and enable automatic corrective feedback to the patient during robotassisted upper limb therapy