Estimating Cervical Range of Motion Using Computer Vision

Surgery of the cervical spine can affect the range of motion both for adult and pediatric patients. The quantification of the cervical spine range of motion (CROM) both before and after surgery, in flexion/extension, rotation, and side bending can be difficult. Despite this difficulty an accurate measurement of cervical range of motion is important for both pre-operative surgical counseling and surgical planning. In clinical practice, currently there is no well-defined protocol to quantify this important parameter. In most cases, either an experienced surgeon estimates the range of motion based on their experience; or a CROM wearable device is used during the clinic visits or physical therapy. We propose a vision-based approach, where the cervical range of motion will be assessed in a quantitative manner by analyzing video as the subject performs flexion/extension, rotation, and side bending actions. The proposed solution is to build a software tool using the 6DRepNet neural network, which could be easily deployed in a clinical setting and provide an accurate quantification of CROM to assist clinical experts in making more informed decisions. We extensively evaluate the proposed method on patient data that was collected during this study. In addition, we compare the estimated values with computer vision-based tool with gyroscopic data for benchmarking.