Broad Learning-based Intervertebral Discs Localization and Segmentation

Abstract

Magnetic resonance image (MRI) can detect a variety of conditions of the spine, including problems with the bones, soft tissues, and disks. Accurate localization and segmentation of intervertebral disc is crucial for the assessment of spine disease diagnosis. In this paper, we present a broad learning-based intervertebral discs localization and segmentation algorithm. Firstly, intervertebral discs localization system based on broad learning is designed. Secondly, based on the localization result, the shape prior of intervertebral discs is established. Finally, a Chan-Vese model method with shape prior is proposed to obtain the segmentation result of intervertebral discs. Several baselines, with different segmentation algorithms, were used to demonstrate the effectiveness of the proposed architecture.