Sang Hoon Chae1, Yushin Kim2, Kyoung-Soub Lee3, Hyung-Soon Park3,1. 1. Graduate School of Medical Science and Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, KR. 2. Major of Sports Health Rehabilitation, Cheongju University, Cheongju, KR. 3. Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon, KR.
Abstract
BACKGROUND: Recent advancement of wearable sensor technology has shown feasibility of remote physical therapy at home. Especially, current crisis of pandemic has revealed the need and opportunity of internet-based wearable technology in the future healthcare system. Previous researches have shown the feasibility of human activity recognition technology for monitoring rehabilitation activities at home environments; however, few comprehensive studies ranging from the development to the clinical evaluation exist. OBJECTIVE: This study aims to 1) develop a home-based rehabilitation (HBR) system which can recognize and record type and frequency of rehabilitation exercises conducted by the user by using smartwatch, smartphone application equipped with machine learning (ML) algorithm, and to 2) evaluate the efficacy of the home-based rehabilitation system through prospective comparative study with chronic stroke survivors. METHODS: The HBR system consisted of off-the-shelf smartwatch, smartphone, and custom developed applications. Convolution neural network (CNN) was used to train the ML algorithm for detecting the home exercises. To figure out the most accurate way for detecting the type of home exercise, we compared accuracy results with the dataset of personal/total data and accelerometer/gyroscope/accelerometer combined with gyroscope data. From March 2018 to February 2019, we conducted a clinical study with two groups of stroke survivors. Totally, 17 and 6 participants were enrolled for statistical analysis in HBR group and control group, respectively. To measure clinical outcomes, we assessed the Wolf motor function test (WMFT), Fugl-meyer assessment of upper extremity (FMA-UE), grip power test, Beck depression inventory and range of motion (ROM) of the shoulder joint at 0, 6, 12, and follow up assessment 6 weeks after retrieving the HBR system. RESULTS: The ML model created by personal data using the accelerometer combined with gyroscope data (99.9% [5590/5601]) was most accurate compared to accelerometer (98.1% [5496/5601]) or gyroscope (96.0% [5381/5601]). In comparative study, drop-out rates of control and HBR group were 4/10 (40%) and 5/22 (22%) at 12 weeks and 10/10 (100%) and 10/22 (45%) at 18 weeks, respectively. The HBR group (N=17) showed a significant improvement in mean WMFT score (39.7, vs 40.5, vs 42.5, overall P =.02), ROM of flexion (74.5, vs 93.9, P = .004) and internal rotation (50.4, vs 70.3, P = .001). The control group (N=6) showed a significant change only in shoulder internal rotation (50.8, vs 48.5, vs 57.3, P = .03). CONCLUSIONS: This research found that the homecare system using the commercial smartwatch and ML model can facilitate the participation of home training and improve the functional score of WMFT and shoulder ROM of flexion and internal rotation for the treatment of patients with chronic stroke. This strategy can possibly be one of cost-effective tools for homecare treatment of stroke survivors in the future. CLINICALTRIAL: Clinical Research information Service (CRIS) Registration number: KCT0004818. INTERNATIONAL REGISTERED REPORT: RR2-10.1109/ICORR.2019.8779475.
BACKGROUND: Recent advancement of wearable sensor technology has shown feasibility of remote physical therapy at home. Especially, current crisis of pandemic has revealed the need and opportunity of internet-based wearable technology in the future healthcare system. Previous researches have shown the feasibility of human activity recognition technology for monitoring rehabilitation activities at home environments; however, few comprehensive studies ranging from the development to the clinical evaluation exist. OBJECTIVE: This study aims to 1) develop a home-based rehabilitation (HBR) system which can recognize and record type and frequency of rehabilitation exercises conducted by the user by using smartwatch, smartphone application equipped with machine learning (ML) algorithm, and to 2) evaluate the efficacy of the home-based rehabilitation system through prospective comparative study with chronic stroke survivors. METHODS: The HBR system consisted of off-the-shelf smartwatch, smartphone, and custom developed applications. Convolution neural network (CNN) was used to train the ML algorithm for detecting the home exercises. To figure out the most accurate way for detecting the type of home exercise, we compared accuracy results with the dataset of personal/total data and accelerometer/gyroscope/accelerometer combined with gyroscope data. From March 2018 to February 2019, we conducted a clinical study with two groups of stroke survivors. Totally, 17 and 6 participants were enrolled for statistical analysis in HBR group and control group, respectively. To measure clinical outcomes, we assessed the Wolf motor function test (WMFT), Fugl-meyer assessment of upper extremity (FMA-UE), grip power test, Beck depression inventory and range of motion (ROM) of the shoulder joint at 0, 6, 12, and follow up assessment 6 weeks after retrieving the HBR system. RESULTS: The ML model created by personal data using the accelerometer combined with gyroscope data (99.9% [5590/5601]) was most accurate compared to accelerometer (98.1% [5496/5601]) or gyroscope (96.0% [5381/5601]). In comparative study, drop-out rates of control and HBR group were 4/10 (40%) and 5/22 (22%) at 12 weeks and 10/10 (100%) and 10/22 (45%) at 18 weeks, respectively. The HBR group (N=17) showed a significant improvement in mean WMFT score (39.7, vs 40.5, vs 42.5, overall P =.02), ROM of flexion (74.5, vs 93.9, P = .004) and internal rotation (50.4, vs 70.3, P = .001). The control group (N=6) showed a significant change only in shoulder internal rotation (50.8, vs 48.5, vs 57.3, P = .03). CONCLUSIONS: This research found that the homecare system using the commercial smartwatch and ML model can facilitate the participation of home training and improve the functional score of WMFT and shoulder ROM of flexion and internal rotation for the treatment of patients with chronic stroke. This strategy can possibly be one of cost-effective tools for homecare treatment of stroke survivors in the future. CLINICALTRIAL: Clinical Research information Service (CRIS) Registration number: KCT0004818. INTERNATIONAL REGISTERED REPORT: RR2-10.1109/ICORR.2019.8779475.