Application of Deep Learning Algorithm to Monitor Upper Extremity Task Practice

Abstract

Upper extremity hemiplegia is a serious problem affecting the lives of many people post-stroke. Motor recovery requires high repetitions and quality of task-specific practice. Sufficient practice cannot be completed during therapy sessions, requiring patients to perform additional task practices at home on their own. Adherence to and quality of these home task practices are often limited, which is likely a factor reducing rehabilitation effectiveness post-stroke. However, home adherence is typically measured by self-reports that are known to be inconsistent with objective measurement. The objective of this study was to develop algorithms to enable the objective identification of task type and quality. Twenty neurotypical participants wore an IMU sensor on the wrist and performed four representative tasks in prescribed fashions that mimicked correct, compensatory, and incomplete movement qualities typically seen in stroke survivors. LSTM classifiers were trained to identify the task being performed and its movement quality. Our models achieved an accuracy of 90.8% for task identification and 84.9%, 81.1%, 58.4%, and 73.2% for movement quality classification for the four tasks for unseen participants. The results warrant further investigation to determine the classification performance for stroke survivors and if quantity and quality feedback from objective monitoring facilitates effective task practice at home, thereby improving motor recovery.

Wearable Sensor to Monitor Quality of Upper Limb Task Practice for Stroke Survivors at Home

“…Specifically, a previous study found an F1 score of 0.84 when detecting compensatory movement during a simple task of arm raise in neurotypical adults and stroke survivors using the random forest [ 25 ]. Another study found an F1 score of 0.89 when detecting compensatory and incomplete movements during functional tasks in neurotypical adults using long short-term memory [ 26 ]. The present study found an F1 score of 0.95 in stroke survivors, possibly due to a person-specific modeling approach used compared to other studies.…”

“…The 4 functional tasks used in this study are described in Table 3 . More details of these tasks along with task instructions and pictures can be found in the study of Li et al, 2023 [ 26 ]. These tasks were chosen as representative functional tasks prescribed in home exercise programs from the task practice manual [ 32 ] and vary in difficulty.…”

“…However, this task was somewhat limited in relevance for activities of daily living typically involving object manipulation using the hand. Another study distinguished the quality of movements for multiple representative functional task practices [ 26 ], but only tested healthy adults. Thus, the detection of subtle differences in stroke survivors’ way of performing functional tasks with versus without compensatory movement patterns using a wrist-worn IMU sensor has not been investigated.…”

“… Reference Participant Method Setting Finding Bochniewicz 2017 [ 27 ] 10 healthy persons, 10 stroke survivors RF Lab Classified functional vs. nonfunctional movement David 2021 [ 28 ] 5 healthy persons, 5 hemiparetic patients Thresholding Lab Classified functional vs. nonfunctional movement Gomez-Arrunategui 2022 [ 29 ] 12 stroke survivors RF, CNN Lab Detected reach time and number of reaching gestures during prescribed tasks. Lee 2018 [ 25 ] 9 healthy persons, 11 stroke survivors RF Lab Classified quality of arm raise Bhagat 2020 [ 30 ] 2 persons with spinal cord injury DTW, LSTM Lab Classified cylindrical vs. pincer grasp (to pick up a water bottle vs. pen) Li 2023 [ 26 ] 20 healthy persons LSTM Lab Classified functional task practice quality Lui 2019 [ 31 ] …”

Advanced Home-Based Shoulder Rehabilitation: A Systematic Review of Remote Monitoring Devices and Their Therapeutic Efficacy

“…The selection of the monitoring devices should prioritize usability and patient comfort. Wearable motion sensors, such as IMUs or M-IMUs, offer a non-intrusive solution for continuous monitoring of shoulder movements during rehabilitation exercises [16,[25][26][27][28][29][30]32]. These devices are small, lightweight, non-invasive electronic devices, enabling real-time feedback and data collection, empowering patients to track their progress and adherence to prescribed exercises.…”