App for motion analysis: pilot study

Abstract

Objective: Alternative and low-cost measures may be important for analyzing human movement. The objective of this study was to verify the agreement of human movement analysis of a monitoring app that uses artificial intelligence compared to three-dimensional movement analysis. Methods: Observational cross-sectional case report study in which a healthy volunteer performed arm flexion, elbow flexion, trunk flexion, lateral trunk bending, and sitting and standing. Images of the volunteer were simultaneously captured by a three-dimensional movement analysis system based on infrared cameras and the Linkfit app of two mobile devices (smartphones). The body angles estimated by the Linkfit app were compared with the corresponding angles measured by the three-dimensional movement analysis system. The Granger causality test was used to compare the pairs of angles for each parallel data series. Results: The use of smartphone cameras and deep learning techniques for motion detection had an 84% degree of agreement compared to measurements generated by the three-dimensional movement analysis performed in the laboratory. Conclusion: The use of smartphone cameras and deep learning techniques is promising for conducting studies for body movement detection compared to the gold standard measures of movement analysis. This technology may become an alternative for movement analysis. Future studies should consider a more significant number of volunteers and model movements to strengthen the results obtained in this study.