Improved function performance in individuals with transradial limb difference using cloud based health system and deep learning

Rehabilitation Institute Of Chicago D/B/A Shirley Ryan Abilitylab, Chicago IL

Investigators

Linked publications, trials & patents

Abstract

PROJECT SUMMARY Transradial amputation is the most common type of major upper limb absence and causes substantial disability. On average, over 1,700 Americans receive a new upper limb amputation (wrist disarticulation level or higher) each year. Over 75% of these individuals are between the ages of 18 and 64 and therefore desire to continue being active. Prosthetic devices are accepted treatment options for upper limb amputation to restore the functional capabilities of the lost arm. However, despite impressive improvements in the last decade, current prosthetic device capabilities still fall far below the functionality of the natural arm and hand, something that contributes to high device rejection rates. Based on our prior study upon which this proposed renewal builds, we identified two important causes for this lack of functionality: signal detection losses caused by physical issues such as intermittent electrode contact and broken lead wires, and user issues such as not making appropriate and distinct muscle contractions when recalibrating the prosthetic device. This study will address these issues through the application of novel technology: deep learning combined with data augmentation to automatically compensate for EMG signal deterioration/loss. This innovative technology will be applied and tested in the following aim. A linear discriminate analysis pattern recognition prosthetic controller will be compared to a prosthetic controller supplemented with deep learning to see which performs better when noise is introduced, as determined by functional and virtual outcomes testing. The primary outcome will be the total number of blocks moved before participants are unable to continue when noise is added to the system. Secondary outcomes include standardized clinical outcome measures such as the Southampton Hand Assessment Procedure, Clothespin Relocation Test, and Jebsen-Taylor Hand Function Test. This work will (i) formulate an effective deep learning method of compensating for control system signal detection issues. By mitigating signal detection issues, this work will provide significant improvements in prosthesis control/function, laying the groundwork for more intuitively controlled prosthetics for individuals with upper limb absence.