Intelligent Haptic Robotic Device for Upper-Limb Stroke Patients
Keywords: Haptic, stroke, automated, rehabilitation.
In collaboration with: Quanser and Toronto Rehabilitation Institute.
Overview of Research
Hemiparetic stroke patients with an affected upper-limb (i.e. arm) usually have great difficulties performing many activities of daily living (ADL), such as pushing on a chair to stand up or getting into bed, due to a lack of coordination and weakness in muscle control. Therapists guide patients through a reaching task exercise to rehabilitate arm control and stability to improve a stroke patient’s functional independence and quality of life.
In the early stages of therapy, patients often have very little control of their limb and require extensive practice and assistance to regain motor skills. Consequently, therapists must spend significant amounts of time guiding patients through the exercise, which can be difficult for both patient and therapist. Discussions with experienced therapists have identified early stage exercises for upper-limb stroke patients as an area of rehabilitation that is in need of efficient supplementary tools.
The Upper-limb Stroke Rehabilitation prototype is an intelligent haptic robotic system designed to conduct the early stages of physiotherapy for upper-limb stroke patients. The goal is to simulate targeted linear-reaching task therapy, using elbow position monitoring and assistance to rehabilitate reaching motor skills. The device will be able to autonomously adjust to meet the specific abilities of different patients, as well as increasing the difficulty of the exercise as patient performance improves. Feedback will be provided to both the patient (in the form of graphic motivation and games) and therapist (in the form of patient performance statistics).
The block diagram in Figure 1 depicts the overall intelligent haptic-robotic system, which consists of a hardware device, an artificial intelligence controller, and a virtual reality user interface.
Figure 1: Block diagram of AUSR system - red dashed box denotes
hardware/software boundary (click on figure to enlarge)
The first robotic prototype has been completed and is shown in Figure 2. The non-restraining, haptic platform provides resistance and directional guidance for the user during the exercise. It has two active and two passive degrees of freedom, and allows the reaching exercise of either upper-limb to be performed in three-dimensional space. Built-in encoders and unobtrusive sensors provide data to indicate hand position and abnormal upper-limb posture during the rehabilitation exercise, respectively. Other data, such as velocity and force, can be calculated to monitor proper technique and posture. This array of information forms the input parameters for the controller.
Figure 2: Upper-limb stroke rehabilitation prototype
The control system is based on a partially observable Markov decision process (POMDP), a versatile decision-theoretic modelling technique, and is responsible for guiding the patient through the reaching exercise. It can autonomously adjust the exercise parameters (e.g. resistance, target distance) to the abilities of each patient and increase the difficulty of the exercise as the patient improves. The goal of the controller is to help the patient achieve their maximum reaching distance at the greatest level of resistance, while maintaining control. Figure 3 depicts the current POMDP controller, modelled as a dynamic Bayesian network (DBN).
Figure 3: Diagram of POMDP controller (click on figure to enlarge)
User trials for this system are schedule for summer 2008 in Toronto, Canada, and are expected to be completed in the fall.
Videos
The following video demonstrates the task the AUSR system is being designed to replicate (right-click and "Save as..."). If you would like a copy of the video in another format than .wmv, please e-mail Jen.
Funding Sources
Communications and Information Technology Ontario (CITO)
Research Team
Paul Lam, M.A.Sc. (IBBME, University of Toronto)
Patricia Kan, M.A.Sc. candidate (IBBME, University of Toronto)
Alex Mihailidis, Ph.D., P.Eng. (University of Toronto)
Stephen Hill, Ph.D. (Toronto Rehabilitation Institute)
Jacob Apkarian Ph.D. (Director R&D, Quanser)
Debbie Hébert, M.Sc. (Occupational Science and Occupational Therapy, University of Toronto)
Craig Boutilier, Ph. D. (Chair, Computer Science, University of Toronto)
Jesse Hoey, Ph. D. (Computer Science, University of Toronto)
Geoff Fernie, Ph.D. P.Eng. (Toronto Rehabilitation Institute)
