Automatic segmentation of intervertebral disk from MR images of the spine based on graph cut method

摘要

In this study a fully automatic segmentation of intervertebral disk method based on graph cut (GC) algorithm is presented. In graph cut algorithm, a graph is constructed from image pixels and two additional nodes named source (s) and sink (t). GC algorithm requires some seed points for initialization. Presented method extracts several features from training images and uses them to automate GC algorithm. Post-processing operation like morphological and content-based refinement is applied to segmented image by GC to get desire ROI. The result of intervertebral disk segmentation with this method shows 95% accuracy.