Projects per year
Abstract
Aortic dissection (AD) is a condition of the main artery of the human body, resulting in the formation of a new flow channel, or false lumen (FL). The disease is usually diagnosed with a computed tomography angiography (CTA) scan during the acute phase. A better understanding of the causes of AD requires knowledge of aortic geometry prior to the event, which is available only in very rare circumstances. In this work, we propose an approach to reconstruct the aorta before the formation of a dissection by performing 3D inpainting with a two-stage generative adversarial network (GAN). In the first stage of our two-stage GAN, a network is trained on the 3D edge information of the healthy aorta to reconstruct the aortic wall. The second stage infers the image information of the aorta to reconstruct the entire dataset. We train our two-stage GAN with 3D patches from 55 non-dissected aortic datasets and evaluate it on 20 more non-dissected datasets, demonstrating that our proposed 3D architecture outperforms its 2D counterpart. To obtain pre-dissection aortae, we mask the entire FL in AD datasets. Finally, we provide qualitative feedback from a renown expert on the obtained pre-dissection cases
Original language | English |
---|---|
Title of host publication | Thoracic Image Analysis - Second International Workshop, TIA 2020, Held in Conjunction with MICCAI 2020, Proceedings |
Subtitle of host publication | Second International Workshop, TIA 2020, Held in Conjunction with MICCAI 2020, Lima, Peru, October 8, 2020, Proceedings |
Editors | Jens Petersen, Raúl San José Estépar, Alexander Schmidt-Richberg, Sarah Gerard, Bianca Lassen-Schmidt, Colin Jacobs, Reinhard Beichel, Kensaku Mori |
Place of Publication | Cham |
Publisher | Springer |
Pages | 130-140 |
Number of pages | 11 |
ISBN (Electronic) | 978-3-030-62469-9 |
ISBN (Print) | 978-3-030-62468-2 |
DOIs | |
Publication status | E-pub ahead of print - Nov 2020 |
Event | 2nd International Workshop on Thoracic Image Analysis - Virtuell, Peru Duration: 8 Oct 2020 → 8 Oct 2020 |
Publication series
Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
---|---|
Volume | 12502 LNCS |
ISSN (Print) | 0302-9743 |
ISSN (Electronic) | 1611-3349 |
Conference
Conference | 2nd International Workshop on Thoracic Image Analysis |
---|---|
Abbreviated title | TIA 2020 |
Country/Territory | Peru |
City | Virtuell |
Period | 8/10/20 → 8/10/20 |
Keywords
- Aortic dissection
- Deep learning
- Edge reconstruction
- Generative adversarial networks
- Inpainting
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)
Fingerprint
Dive into the research topics of 'Semi-supervised Virtual Regression of Aortic Dissections Using 3D Generative Inpainting'. Together they form a unique fingerprint.Projects
- 1 Finished
-
Aortic Dissection
Egger, J., Pepe, A., Schmalstieg, D., Schussnig, R., von der Linden, W., Melito, G. M., Holzapfel, G., Ramalho Queiroz Pacheco, D., Jafarinia, A., Brenn, G., Ranftl, S., Müller, T. S., Gupta, I., Steinbach, O., Fries, T., Badeli, V., Hochrainer, T., Schanz, M., Rolf-Pissarczyk, M., Biro, O. & Ellermann, K.
1/01/18 → 31/12/20
Project: Research project
Activities
- 1 Talk at conference or symposium
-
Semi-supervised Virtual Regression of Aortic Dissections Using 3D Generative Inpainting
Antonio Pepe (Speaker)
8 Oct 2020Activity: Talk or presentation › Talk at conference or symposium › Science to science