Automatic Spine Curvature Estimation from X-ray Images

Automatic Spine Curvature Estimation from X-ray Images

Team: Bidur Khanal, Lavsen Dahal, Saurav Jha and Bishesh Khanal

Idiopathic scoliosis is one of the most common spinal deformities that can be potentially lethal if not intervened early enough. Cobb angle that provides a measurement for the spinal deformation curvature is a clinical standard for diagnosing scoliosis and an important metric for treatment planning. However, cobb angle measurement suffers from inter-operator variability and is tedious to measure for surgeons or radiologists.

We are developing an automated method for measuring cobb angle directly from X-ray images of spin as part of the MICCAI 2019 challenge on Accurate Automated Spinal Curvature Estimation. The challenge dateset consists of 609 spinal anterior-posterior x-ray images for training. Each vertebra was located by four landmarks with respect to four corners by two professional doctors in London Health Sciences Center. The Cobb angles were calculated using these landmarks. The evaluation metrics will be Cobb angle accuracy.

Project Category: Computer Vision, Machine Learning, Medical Imaging