Promoting continuous rehabilitation with AI-based motion capture and biomechanical analysis: objective feedback as a...

Exercise therapy is effective for various diseases, but one of the problems in rehabilitation is "continuity" and preventing quitting. This study assumes that objectively tracking changes in motor skills and providing feedback may contribute to the continuity of rehabilitation. Upper limb joint moments were estimated by our program using coordinates obtained through AI-based markerless motion capture. This study aimed to visualize and provide feedback on participants' assessments of their motor function to promote an understanding of changes in their motor function and to improve their motivation to participate in the exercise. Then, the promoting effect of the feedback is examined with a questionnaire.

Participants were late-stage elderly (n = 31) patients of a hospital's rehabilitation facilities. The four disease groups included Cerebrovascular Disease, Parkinson's Disease, Diabetes mellitus, and walking problems with Hallux Valgus. The experiment was done with a non-random assignment design. They were divided into two groups: with feedback (intervention group) and without feedback (control group), by balancing the number of participants for each disease. The intervention period consisted of 8 weeks of exercise, with a feedback report provided after 4 weeks and a follow-up after an additional 4 weeks. The original questionnaire (MEQ: Motivation in Exercise Questionnaire), comprising three questions, was used to assess motivation for exercising, self-understanding of improvements in athletic ability, and intention of continuing exercising. MEQ was administered at the start of the program and after each feedback. Changes in questionnaire scores were examined by multiple regression analysis, with explanatory variables of the kind of disease and with or without feedback.

The t-test was performed on each partial regression coefficient, and Parkinsonism participants had significantly negative values compared to those in other disease groups (t(26) = -2.43, p = 0.022, 95%CI [-1.85, -0.15]). For participants other than those with Parkinson's disease, changes in questionnaire scores were analyzed using a paired t-test and Wilcoxon signed-rank test. There was a tendency for scores to improve in the feedback group regarding willingness and to continue to participate in exercising. Q1 "Did participating in the exercise motivate you to participate in the exercise?" (one-sided, difference = 0.72, t(10) = 1.39, p = 0.09, 95%CI [-0.22, inf], Cohen's d = 0.50; 95%CI [-0.36, 1.34]). Q3 regarding motivation to continue exercising ("Do you want to continue exercising?", (one-sided, mean difference = 0.64, t(10) = 1.47, p = 0.09, d = 0.43; 95%CI [-0.42, 1.27]). Q2 "Were you able to understand changes in your athletic ability?" has a smaller difference (one-sided, mean difference = 0.55, t(10) = 1.20, p = 0.13, 95%CI [-0.28, inf], Wilcoxon Signed rank test: one-sided, V = 25.5, p = 0.15, d = 0.38; 95%CI [-0.47, 1.22]).

As indicated by the changes in the scores of the questionnaire survey in this study, the participants' motivation to participate improved after the feedback was given; however, these positive effects had disappeared in the measurement one month later. Thus, continuous feedback more than once per month seems to be required to keep motivation to exercise.