Work

Subtask performance of Timed Up and Go test in maintenance hemodialysis patients based on computer vision: a comparison with healthy controls

Year: 2022

Type: preprint

Abstract: Background Subtask durations of the Timed Up and Go (TUG) test may provide detailed information about mobility impairment in patients undergoing maintenance hemodialysis (MHD). The aims of this study were to: (1) compare the TUG subtask durations between MHD patients and healthy controls, (2) verify the validity of subtask durations based on computer vision (CV). Methods The 3-meter TUG test was performed by MHD patients and healthy controls, and videos were recorded meanwhile. Additionally, 2-minute walking test (2MWT), five repetition sit-to-stand test (5STS) and 360-degree turn test (360TT) were performed by MHD patients and objective parameters of these tests were recorded by wearable inertial measurement units. The TUG test was divided into five subtasks: standing up (St1), walking forward (St2), turning around (St3), walking backward (St4) and turning to sitting down (St5). Five subtask durations were extracted from videos by CV algorithm and human observation independently. Results A total of 35 MHD patients and 35 healthy controls were recruited. MHD patients had significantly longer durations of St1, St3, and St5 compared with healthy controls (Cohen’s d =1.112, 1.378, 0.801 respectively; P<0.01 for all comparisons). The CV based subtask durations shown good agreements with ground truths extracted by human observations (Intraclass correlation coefficient＞0.8). Strong correlations were found between St2 duration and gait speed of 2MWT (r=-0.815), St3 duration and turning speed of 360TT (r=-0.632), St1 duration and standing time of 5STS (r=0.715), St5 duration and sitting time of 5STS (r=0.704) (all P <0.001). Conclusion The worse TUG test performance in MHD patients mainly manifested as the slowness of turning around, standing up and sitting down. The CV based TUG test was a highly efficient tool for multi-dimensional motor evaluation.

Source: Research Square

DOI: 10.21203/rs.3.rs-2356359/v1