A GPU Based Diffusion Method for Whole-Heart and Great Vessel Segmentation

详细信息    查看全文

• 关键词：Segmentation ; Diffusion ; Random walks ; Interactive segmentation ; Semi ; automatic segmentation ; Multi ; GPU
• 刊名：Lecture Notes in Computer Science
• 出版年：2017
• 出版时间：2017
• 年：2017
• 卷：10129
• 期：1
• 页码：121-128
• 丛书名：Reconstruction, Segmentation, and Analysis of Medical Images
• ISBN：978-3-319-52280-7
• 卷排序：10129

文摘

Segmenting the blood pool and myocardium from a 3D cardiovascular magnetic resonance (CMR) image allows to create a patient-specific heart model for surgical planning in children with complex congenital heart disease (CHD). Implementation of semi-automatic or automatic segmentation algorithms is challenging because of a high anatomical variability of the heart defects, low contrast, and intensity variations in the images. Therefore, manual segmentation is the gold standard but it is labor-intensive. In this paper we report the set-up and results of a highly scalable semi-automatic diffusion algorithm for image segmentation. The method extrapolates the information from a small number of expert manually labeled reference slices to the remaining volume. While results of most semi-automatic algorithms strongly depend on well-chosen but usually unknown parameters this approach is parameter-free. Validation is performed on twenty 3D CMR images.