Author(s) :
P. Lavanya
Article Name :
Volumetric MR Image of Spine Parsing Using a Deep Learning Approach With Semantic Image Representation
Abstract :
The automatic recognition of spine structures in images is significant for computer-aided diagnosis of spinal issues. Identification of the globe spine and structural data of local vertebra, including pose, spine shape and vertebra location is the goal of automatic vertebra recognition. Due to the significant visual changes in various high geometric noises and image modalities views the in spine format, vertebral recognition is difficult. Here, to overcome the aforementioned issues, the two-stage frame work called SpinePraseNet is proposed and it is intended for accomplishing the computerized volumetric MR images for spine parsing. The SpineParseNet comprises of a 2D residual U-Net (ResUNet) for 2D segmentation refinement and a 3D Graph Convolutional Segmentation Network (3D GCSN) for 3D coarse segmentation. And this network is employed to transform the input images into a graph representation, where each functioning effectively to a different spinal structure. By constructing the changing network representation to an image data format, the proposed region UN pooling module enables the 3D GCSN to more efficiently accomplish accurate coarse segmentation. The proposed technique provides better outcomes for disease detection and treatment. The PYTHON software is utilized to construct the project.
No. of Downloads :
8