CAD-RADS Scoring using Deep Learning and Task-Specific Centerline Labeling

Felix Denzinger, Michael Wels, Oliver Taubmann, Mehmet A. Gülsün, Max Schöbinger, Florian André, Sebastian J. Buss, Johannes Görich, Michael Sühling, Andreas Maier
Proceedings of The 5th International Conference on Medical Imaging with Deep Learning, PMLR 172:315-324, 2022.

Abstract

With coronary artery disease (CAD) persisting to be one of the leading causes of death worldwide, interest in supporting physicians with algorithms to speed up and improve diagnosis is high. In clinical practice, the severeness of CAD is often assessed with a coronary CT angiography (CCTA) scan and manually graded with the CAD-Reporting and Data System (CAD-RADS) score. The clinical questions this score assesses are whether patients have CAD or not (rule-out) and whether they have severe CAD or not (hold-out). In this work, we reach new state-of-the-art performance for automatic CAD-RADS scoring. We propose using severity-based label encoding, test time augmentation (TTA) and model ensembling for a task-specific deep learning architecture. Furthermore, we introduce a novel task- and model-specific, heuristic coronary segment labeling, which subdivides coronary trees into consistent parts across patients. It is fast, robust, and easy to implement. We were able to raise the previously reported area under the receiver operating characteristic curve (AUC) from 0.914 to \textbf{0.942} in the rule-out and from 0.921 to \textbf{0.950} in the hold-out task respectively.

Cite this Paper


BibTeX
@InProceedings{pmlr-v172-denzinger22a, title = {{CAD-RADS Scoring using Deep Learning and Task-Specific Centerline Labeling}}, author = {Denzinger, Felix and Wels, Michael and Taubmann, Oliver and G{\"u}ls{\"u}n, Mehmet A. and Sch{\"o}binger, Max and Andr{\'e}, Florian and Buss, Sebastian J. and G{\"o}rich, Johannes and S{\"u}hling, Michael and Maier, Andreas}, booktitle = {Proceedings of The 5th International Conference on Medical Imaging with Deep Learning}, pages = {315--324}, year = {2022}, editor = {Konukoglu, Ender and Menze, Bjoern and Venkataraman, Archana and Baumgartner, Christian and Dou, Qi and Albarqouni, Shadi}, volume = {172}, series = {Proceedings of Machine Learning Research}, month = {06--08 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v172/denzinger22a/denzinger22a.pdf}, url = {https://proceedings.mlr.press/v172/denzinger22a.html}, abstract = {With coronary artery disease (CAD) persisting to be one of the leading causes of death worldwide, interest in supporting physicians with algorithms to speed up and improve diagnosis is high. In clinical practice, the severeness of CAD is often assessed with a coronary CT angiography (CCTA) scan and manually graded with the CAD-Reporting and Data System (CAD-RADS) score. The clinical questions this score assesses are whether patients have CAD or not (rule-out) and whether they have severe CAD or not (hold-out). In this work, we reach new state-of-the-art performance for automatic CAD-RADS scoring. We propose using severity-based label encoding, test time augmentation (TTA) and model ensembling for a task-specific deep learning architecture. Furthermore, we introduce a novel task- and model-specific, heuristic coronary segment labeling, which subdivides coronary trees into consistent parts across patients. It is fast, robust, and easy to implement. We were able to raise the previously reported area under the receiver operating characteristic curve (AUC) from 0.914 to \textbf{0.942} in the rule-out and from 0.921 to \textbf{0.950} in the hold-out task respectively.} }
Endnote
%0 Conference Paper %T CAD-RADS Scoring using Deep Learning and Task-Specific Centerline Labeling %A Felix Denzinger %A Michael Wels %A Oliver Taubmann %A Mehmet A. Gülsün %A Max Schöbinger %A Florian André %A Sebastian J. Buss %A Johannes Görich %A Michael Sühling %A Andreas Maier %B Proceedings of The 5th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2022 %E Ender Konukoglu %E Bjoern Menze %E Archana Venkataraman %E Christian Baumgartner %E Qi Dou %E Shadi Albarqouni %F pmlr-v172-denzinger22a %I PMLR %P 315--324 %U https://proceedings.mlr.press/v172/denzinger22a.html %V 172 %X With coronary artery disease (CAD) persisting to be one of the leading causes of death worldwide, interest in supporting physicians with algorithms to speed up and improve diagnosis is high. In clinical practice, the severeness of CAD is often assessed with a coronary CT angiography (CCTA) scan and manually graded with the CAD-Reporting and Data System (CAD-RADS) score. The clinical questions this score assesses are whether patients have CAD or not (rule-out) and whether they have severe CAD or not (hold-out). In this work, we reach new state-of-the-art performance for automatic CAD-RADS scoring. We propose using severity-based label encoding, test time augmentation (TTA) and model ensembling for a task-specific deep learning architecture. Furthermore, we introduce a novel task- and model-specific, heuristic coronary segment labeling, which subdivides coronary trees into consistent parts across patients. It is fast, robust, and easy to implement. We were able to raise the previously reported area under the receiver operating characteristic curve (AUC) from 0.914 to \textbf{0.942} in the rule-out and from 0.921 to \textbf{0.950} in the hold-out task respectively.
APA
Denzinger, F., Wels, M., Taubmann, O., Gülsün, M.A., Schöbinger, M., André, F., Buss, S.J., Görich, J., Sühling, M. & Maier, A.. (2022). CAD-RADS Scoring using Deep Learning and Task-Specific Centerline Labeling. Proceedings of The 5th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 172:315-324 Available from https://proceedings.mlr.press/v172/denzinger22a.html.

Related Material