3D Inception U-Net for Aorta Segmentation using Computed Tomography Cardiac Angiography

3D Inception U-Net for Aorta Segmentation using Computed Tomography Cardiac Angiography
Title:
3D Inception U-Net for Aorta Segmentation using Computed Tomography Cardiac Angiography
Other Titles:
IEEE-EMBS International Conference on Biomedical and Health Informatics 2019 (BHI-2019)
Publication Date:
19 May 2019
Citation:
S. R. Ravichandran et al., "3D Inception U-Net for Aorta Segmentation using Computed Tomography Cardiac Angiography," 2019 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), Chicago, IL, USA, 2019, pp. 1-4. doi: 10.1109/BHI.2019.8834582
Abstract:
Computed Tomography Coronary Angiography (CTCA) is an effective imaging technique used for diagnosis and surgical planning. Segmentation of the aorta from the CTCA can be used clinically for interpretation, aortic valve measurement for intervention and identification of coronary structures. The process of segmentation done manually is tedious and time consuming. In this paper we propose an approach to automatic aorta segmentation from cardiac CTCA scans using deep learning. A dataset of 20 pairs of CT and mask is used, with the manually segmented masks as ground truth. The dataset is split into a training set of 10 pairs, validation set of 3 pairs and test set of 7 pairs. The proposed framework uses two different U-Net models trained for location and refined segmentation tasks. The U-Net architecture and multiple variants of the U-Net are adopted, and their performances are compared. The paper also explores different types of data preprocessing and augmentation that can be performed to improve the performance of the deep learning model. The results show that 3D Inception U-Net performs the best in localization and segmentation tasks. The Localization test DSC is 0.77 and Segmentation test DSC is 0.81. Qualitative testing shows that predicted masks are quite similar to the original.
License type:
PublisherCopyrights
Funding Info:
Description:
(c) 2019 IEEE.
ISSN:
2641-3604
2641-3590
Files uploaded:

File Size Format Action
bhi-final-inception-unet.pdf 656.28 KB PDF Open