1 ABSTRACT Title: Comparing physiotherapist's functional test assessments with OpenPose kinematic data Objectives: Assess wether is OpenPose (OP) system able to detect parameters of poor performance in the Single Leg Squat Test and Step Down Test. Presence of these parameters was indicated by consensus process among 3 physiotherapists. Methods: 48 healthy volunteers participated on the research. Each volunteer demonstrated 3 repetitions of Single Leg Squat Test and Step Down Test on each leg. Repetitions were recorded on video and at the same time evaluated by 3 physiotherapists (PT) using 15 binary parameters. Consensus process indicating presence of a parameter involved agreement of at least 2 of the raters while the inter-rater reliability coeficient for all 3 raters had to be κ ≥ 0,41. Changes in angles and relative distances of observed body segments were calculated using OpenPose keypoint detection model. Therefore each execution of the functional test was described by OP data and PT data. Statistical analysis was performed to assess wether was OpenPose able to detect significant differences between poor and good execution of a functional test in the 15 rated parameters. Results: OP was able to reliably assess following parameters: Loss of balance, Excessive upper extremity movement and Lateral...

Title: Detection of Signs of Fatigue in Functional Tests in Healthy Population by OpenPose Objectives: The aim of this study is if the OpenPose system can detect changes in range of motion and unwanted movements of other segments in repetitive a full arm abduction test with weight and a single leg stance based on comparation of changes in angles and relative distances between first and last seconds of the movement. Methods: This pilot study was dedicated to healthy people between 18 and 65 years. A video recording was made within an intervention while subjects were doing a full arm abduction test with weight and a single leg stance until the maximum endurance. Videorecording was processed by the OpenPose software which detects human movements base on algorithm without using a physical markers. Key point combinations (markers) were defined to obtain information about changes in movement and posture (angles, relative distances). Basic statistical measurements were done and compared between first and last seconds of the movement. Temporospatial data were collected to reach the results. Results: Almost all differences in key point combinations to detect arm abduction movement were evaluated as statistically significant. Changes in single leg stance were usually evaluated as not statistically...