Khan, A; Alwazzan, O; Benning, M; Slabaugh, G; (2023) Sequential Segmentation of the Left Atrium and Atrial Scars Using a Multi-scale Weight Sharing Network and Boundary-Based Processing. In: Zhuang, X and Li, L and Wang, S and Wu, F, (eds.) Left Atrial and Scar Quantification and Segmentation: First Challenge, LAScarQS 2022, Held in Conjunction with MICCAI 2022, Singapore, September 18, 2022, Proceedings. (pp. pp. 69-82). Springer: Cham, Switzerland.

Preview

Text Benning_Sequential_segmentation_Khan_et_al_accepted.pdf - Accepted Version Download (8MB) | Preview

Abstract

Left atrial (LA) segmentation and quantification of atrial scars have opened a path to automating Atrial Fibrillation (AF) diagnosis. This paper proposes a two-stage approach for sequential segmentation of the LA cavity and scars. Our Multi-scale Weight Sharing (MSWS) Network extracts features at multiple scales and is used for LA cavity segmentation. We also propose a Boundary2Patches method which performs segmentation of scars around the detected LA cavity boundary. The MSWS network learns a better representation of features through sharing weights across scales, and the Boundary2Patches method focuses on smaller scars constrained in the region around the LA cavity wall. On the challenge cohort (validation set), our method achieves an average Dice score of 0.938 and 0.558 for the LA cavity and scars segmentation of task 1, and a Dice score of 0.846 for LA cavity segmentation of task 2. The pre-trained models, source code, and implementation details are available at https://github.com/kabbas570/LAScarQS2022.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as XMLJSONCSV

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item