Automating Cobb Angle Measurement for Adolescent Idiopathic Scoliosis Using Instance Segmentation

Scoliosis is a three-dimensional deformity of the spine, often diagnosed in childhood. It affects 2-3% of the population, representing seven million people in North America. Currently, the gold standard for assessing scoliosis is done manually by measuring Cobb angles. This manual process is time-consuming and unreliable as it is affected by inter- and intra-observer variance. To eliminate these inaccuracies, machine learning (ML) methods can be used to automate the Cobb angle measurement process. This paper proposes to address the Cobb angle measurement task using an instance segmentation model. The proposed method first segments the vertebrae in an X-ray image using YOLACT model, then tracks the decisive landmarks using minimum bounding boxes. Lastly, the extracted landmarks are used to calculate the corresponding Cobb angles. The proposed method achieved a Symmetric Mean Absolute Percentage Error (SMAPE) score of 10.76%, outperforming the results presented in previous research. Additionally, more than 94% of the estimated Cobb angles had an error of less than ten degrees. The proposed method demonstrates reliability in both vertebra localization and Cobb angle measurement.