The goal of this project is to develop a framework for post-stroke rehabilitation using body-mounted inertial measurement units in conjunction with hands-off robotics to assist in the recovery and rehabilitation of stroke-affected patients.
Stroke is a major health problem in the United States, with up to 500,000 new or recurrent cases each year. With similar statistics in other developing countries, there are many numerous people around the world suffering from stroke or stroke-related ailments. An increasing percentage of elderly in the population, and longer life expectancies are resulting in increasing numbers of people living at risk for, or recovering from stroke. The resulting physical impairment of stroke can cause a number of problems. Beyond the direct effects of the physical impairment, such as pain, spasticity, and limited motion, there are indirect effects such as loss of independence. Often, in patients suffering from hemiplegia, the burden of the non-functioning limb can shifts to the functioning limb; this poses problems for health and fitness, such as early onset of arthritis in the functioning limbs.
It has been shown that preventing permanent loss of physical functionality for stroke-affected limbs requires rehabilitation to begin immediately after the stroke. Unfortunately, a lack of resources often limits the stroke-affected patient's access to the necessary health professionals and ultimately results in a continued decline in health and functionality. An in-home hands-off rehabilitation robot is uniquely suited to accommodate such patients. In the post-subacute period, when a patient is in his or her own home recovering from the stroke, a robot can assist with and augment the treatment regimen prescribed by the health professional. We plan to use the robot in conjunction with body-mounted motion capture to monitor physical status and augment the treatment regimen.
The overall system consists of small inertial measurement units (IMUs), worn on specific body parts. The IMUs capture linear acceleration and angular rate of change, and are used to reconstruct the patient's biomechanical model. This information is transmitted to the robot in real time (the information can subsequently be transmitted by the robot to a local laptop or PC, where it is either stored in a database or made available over the internet to the relevant health professional). With this framework, the system has two modes of operation;Assessment -
In assessment mode, the robot leads the stroke-affected patient through a set of pre-determined functional motor tasks, and evaluates the patient's performance. The robot's judgement will be driven by knowledge of motion primitives as well as activity recognition techniques. This will allow the robot to determine the 'quality' of the user motion, based on parameters such as motion velocity, timing, directness of path, etc. In rehabilitation mode, we add the requirement of real-time feedback. If necessary, the robot reminds the patient to begin his or her exercise. Subsequently, the robot evaluates the user motion in order to provide positive feedback during exercise. This feedback consists of giving constructive feedback with regard to the 'correctness' of motion, as well as encouragement to continue with the activity.
With this work, we expect that the increased frequency of evaluation will help the health professional to tailor the patient's regimen. Further, the ability to observe the user undergoing normal, 'daily' activities such as folding clothing and reaching for objects on a shelf with give additional data to be used to determine how well the rehabilitation regimen is working.
Inertial measurement components
This work is supported by USC Women in Science and Engineering (WiSE) Postdoctoral Fellowship and the National Science Foundation grants titled "HRI: Personalized Assistive Human-Robot Interaction: Validation in Socially-Assisted Post-Stroke Rehabilitation" -
Award Number 0713697, and "CRI: IAD - Computing Research Infrastructure for Human-Robot Interaction and Socially Assistive Robotics" - Award Number: 0709296. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.