AutoSpine-Net: Spine Detection Using Convolutional Neural Networks for Cobb Angle Classification in Adolescent Idiopathic Scoliosis

Abstract

The Cobb angle measurement in adolescent idiopathic scoliosis is prone to inter- and intra-observer variations. This paper proposes a deep learning architecture for detection of spine vertebrae from X-ray images to automatically evaluate the Cobb angle, and to assess for the presence of scoliosis and severity of the curvature. The public AASCE MICCAI 2019 anterior–posterior X-ray image dataset was used for training and testing of the proposed convolutional neural network architecture. Seventeen vertebrae were detected from the input image to obtain 68 landmark features of the spine. The obtained landmarks were processed to measure the Cobb angle and to assess whether scoliosis was present. The severity of the curvature was further classified into mild, moderate and severe, if scoliosis was present. The results showed that the proposed algorithm has a classification accuracy of approximately 0.9 (90%). This architecture may be used as a tool to augment Cobb angle measurement in X-ray images of patients with adolescent idiopathic scoliosis in a real-world clinical setting.