Human Shoulder Posture Anthropometry System for Detecting Scoliosis Using Mediapipe Library

Abstract

The system proposed in this research is a posture detection system using real-time computer vision technology with system limitations aimed at detecting shoulder posture as part of anthropometric measurements, because if the shoulder posture is unbalanced and has a very significant height difference, it is called an indication of scoliosis. This research aims to facilitate the detection of scoliosis, especially in one of its symptoms, namely shoulder asymmetry with anthropometric measurements of the ‘Elbow-to-Elbow breadth’ position using the scoliomter method. In addition, common screening methods that can be used for scoliosis, especially in adolescents, include the Adams forward bend test, Cobb angle measurement, and Moire measurement. The anthropometric shoulder posture detection system includes the stages of preparation for detection using a webcam with T-position calibration, then MediaPipe Library processes 33 keypoints, OpenCV and Python to analyze body movements in real time, then this asymmetry is calculated using standard algorithms for pose prediction, vector projection and atan2 to obtain asymmetry angle information. The results of testing the shoulder detection system in the form of shoulder posture according to landmarks on one test subject and keypoint extraction on the user interface display in real time and provide information on the angle of asymmetry of the shoulder and hip in the front and rear facing positions. From testing 16 respondents, the shoulder tilt angle is obtained in the range of 7.42-19.84 degrees which will have a TRUE value if the angle is greater than 15 degrees. Information on the angle of more than 15 degrees can be used as a reference for scoliosis symptoms and further diagnosis by medical practitioners and through this detection system it will be easy to get information related to the results of shoulder posture detection accurately and in real time compared to using only a scoliometer.