Vertebrae segmentation techniques for spinal medical images

Abstract

Accurate localization and segmentation of the spine from medical images plays an important role in CAD (Computer Aided Diagnoses) as it used for many clinical tasks to diagnose many diseases like degenerative disc disease, kyphosis, scoliosis and spondylolisthesis. Besides it can be used as an input to the classification process. Although the bone structures in medical images have high contrast but the vertebrae identification is considered a challenging task due to many difficulties like the unclear boundaries of the vertebrae, the abnormal spine curves and complex structure of the vertebra. In order to achieve an accurate, efficient and automated spine segmentation and detection from medical images, there are several techniques. This paper analyze and review the different vertebrae segmentation techniques. The spine segmentation task divides into three main stages: initial spine skeleton detection, vertebrae segmentation of the spine while the third stage is designed to enhance the results of the vertebrae segmentation. This paper presents different algorithms for each stage as each algorithm is supported by one or more paper which explain the algorithm.