Bone surface extraction and dynamic tracking from ultrasound images by semantic segmentation

In this study, our aim is to extract bone surface of the tibial from ultrasound images by using semantic segmentation and track the feature point of the extraction during knee flexion and extension automatically for dynamic tracking of the tibial movement in the horizontal plane. Automatic dynamic tracking is performed by approximating the bone surface extraction line to a cubic function curve and tracking the inflection point. The effectiveness of the proposed method is verified by analyzing the lower leg during knee flexion and extension by three healthy men in their 20s using ultrasound diagnostic device and confirming the physiological movement of screw-home movement(SHM) during knee flexion and extension movement. From the experimental results, SHM is confirmed during knee flexion and extension movements in each subject. Therefore, the proposed method of this study is effective in the dynamic tracking of the tibial bone surface.