Background: Spine magnetic resonance image (MRI) plays a very important role\nin the diagnosis of various spinal diseases, such as disc degeneration, scoliosis, and\nosteoporosis. Accurate localization and segmentation of the intervertebral disc (IVD)\nin spine MRI can help accelerate the diagnosis time and assist in the treatment by\nproviding quantitative parameters. In this paper, a method based on Gabor filter bank\nis proposed for IVD localization and segmentation.\nMethods: First, the structural features of IVDs are extracted using a Gabor filter bank.\nSecond, the Gabor features of spine are calculated and spinal curves are detected.\nThird, the Gabor feature images (GFI) of IVDs are calculated and adjusted according\nto the spinal curves. Fourth, the IVDs are localized by clustering analysis with GFI.\nFinally, an optimum grayscale-based algorithm with self-adaptive threshold, combined\nwith the localization results and Gabor features of the spine, is performed for IVDs\nsegmentation.\nResults: The proposed method is verified by an MRI dataset consisting of 278 IVDs\nfrom 37 patients. The accuracy of localization is 98.23 % and the dice similarity index for\nsegmentation evaluation is 0.9237.\nConclusions: The proposed Gabor filter based method is effective for IVD localization\nand segmentation. It would be useful in computer-aided diagnosis of IVD diseases and\ncomputer-assisted spine surgery.
Loading....