2021/02/02 by Tarun Kanti Ghosh, Md. Kamrul Hasan, Ghosh, Tarun Kanti +9 · 1 citation
Computer Science · Medicine · #Advanced Neural Network Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Medical Image Segmentation Techniques #Medical Imaging Techniques and Applications #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2102.01822
openalex publication_date 2021/02/02 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Accurate and robust whole heart substructure segmentation is crucial in\ndeveloping clinical applications, such as computer-aided diagnosis and\ncomputer-aided surgery. However, segmentation of different heart substructures\nis challenging because of inadequate edge or boundary information, the\ncomplexity of the background and texture, and the diversity in different\nsubstructures' sizes and shapes. This article proposes a framework for\nmulti-class whole heart segmentation employing non-rigid registration-based\nprobabilistic atlas incorporating the Bayesian framework. We also propose a\nnon-rigid registration pipeline utilizing a multi-resolution strategy for\nobtaining the highest attainable mutual information between the moving and\nfixed images. We further incorporate non-rigid registration into the\nexpectation-maximization algorithm and implement different deep convolutional\nneural network-based encoder-decoder networks for ablation studies. All the\nextensive experiments are conducted utilizing the publicly available dataset\nfor the whole heart segmentation containing 20 MRI and 20 CT cardiac images.\nThe proposed approach exhibits an encouraging achievement, yielding a mean\nvolume overlapping error of 14.5 % for CT scans exceeding the state-of-the-art\nresults by a margin of 1.3 % in terms of the same metric. As the proposed\napproach provides better-results to delineate the different substructures of\nthe heart, it can be a medical diagnostic aiding tool for helping experts with\nquicker and more accurate results.\n