Method for determining flat-valgus deformity of the foot based on image analysis

This article examines the problem of flat-valgus deformity of the feet and proposes a new approach to its primary diagnosis based on the analysis of images of the posterior part of the feet. A sample of ~550 images was collected, including both patients with pathology and healthy people. The images are obtained from the medical center, open sources and volunteers. The images have been preprocessed (filtering, standardization, segmentation) and augmented to increase the diversity of the data. The developed methods for calculating the pronation angle by analyzing the image of the posterior part of the feet are presented. The main method is the combined method. The combined method consists of two methods previously developed by the authors for determining the pronation angle based on an image. The first method is proportional, based on the identification of key anatomical points by analyzing the contour of the leg. The second method uses the YOLO deep learning model to automatically identify anatomical points. The results of the study showed that the combined method demonstrates a relative error of 14% for the left leg and 16% for the right. Clinical trials are needed to confirm the effectiveness of the method in real conditions.