Explainable HCC Diagnosis on Dynamic Contrast-Enhanced MRI with a Li-RADS Concept Bottleneck

Killian Monnin, Patrick Jeltsch, Lucia Fernandes-Mendes, Vasco Cazzagon, Murat Yüce, Vivek Yadav, Mario Jreige, Marianna Gulizia, Montserrat Fraga Christinet, Raphaël Girardet, Clarisse Dromain, Bachir Taouli, Naïk Vietti-Violi, Jonas Richiardi
Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, PMLR 315:3283-3313, 2026.

Abstract

We propose an explainable end-to-end framework for hepatocellular carcinoma (HCC) diagnosis on dynamic contrast-enhanced (DCE) liver MRI. Our method embeds Liver Imaging Reporting and Data System (Li-RADS)–inspired concepts into the network via a multi-head concept bottleneck. A 2.5D EfficientNet backbone processes lesion-centred multiphase MRI crops, and a 4-head architecture jointly predicts continuous soft labels for non-rim arterial phase hyperenhancement (APHE), portal venous/delayed washout and capsule, lesion morphology, and a LR-5 score (definite HCC vs non-HCC) based on the Li-RADS guidelines. Soft labels are derived automatically from intra-lesional, peri-lesional and parenchymal intensity patterns, and the network is trained with uncertainty-weighted losses to balance concept prediction, contrast regression and HCC classification. On our cohort, the Li-RADS–inspired bottleneck substantially improves NormGrad explanation accuracy, geometric stability and intensity robustness while maintaining PR AUC comparable to a single-head baseline, highlighting an interpretable alternative to a black-box HCC classifier.

Cite this Paper

BibTeX
@InProceedings{pmlr-v315-monnin26a, title = {Explainable HCC Diagnosis on Dynamic Contrast-Enhanced MRI with a Li-RADS Concept Bottleneck}, author = {Monnin, Killian and Jeltsch, Patrick and Fernandes-Mendes, Lucia and Cazzagon, Vasco and Y{\"u}ce, Murat and Yadav, Vivek and Jreige, Mario and Gulizia, Marianna and Fraga Christinet, Montserrat and Girardet, Rapha{\"e}l and Dromain, Clarisse and Taouli, Bachir and Vietti-Violi, Na{\"i}k and Richiardi, Jonas}, booktitle = {Proceedings of The 9th International Conference on Medical Imaging with Deep Learning}, pages = {3283--3313}, year = {2026}, editor = {Huo, Yuankai and Gao, Mingchen and Kuo, Chang-Fu and Jin, Yueming and Deng, Ruining}, volume = {315}, series = {Proceedings of Machine Learning Research}, month = {08--10 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v315/main/assets/monnin26a/monnin26a.pdf}, url = {https://proceedings.mlr.press/v315/monnin26a.html}, abstract = {We propose an explainable end-to-end framework for hepatocellular carcinoma (HCC) diagnosis on dynamic contrast-enhanced (DCE) liver MRI. Our method embeds Liver Imaging Reporting and Data System (Li-RADS)–inspired concepts into the network via a multi-head concept bottleneck. A 2.5D EfficientNet backbone processes lesion-centred multiphase MRI crops, and a 4-head architecture jointly predicts continuous soft labels for non-rim arterial phase hyperenhancement (APHE), portal venous/delayed washout and capsule, lesion morphology, and a LR-5 score (definite HCC vs non-HCC) based on the Li-RADS guidelines. Soft labels are derived automatically from intra-lesional, peri-lesional and parenchymal intensity patterns, and the network is trained with uncertainty-weighted losses to balance concept prediction, contrast regression and HCC classification. On our cohort, the Li-RADS–inspired bottleneck substantially improves NormGrad explanation accuracy, geometric stability and intensity robustness while maintaining PR AUC comparable to a single-head baseline, highlighting an interpretable alternative to a black-box HCC classifier.} }
Endnote
%0 Conference Paper %T Explainable HCC Diagnosis on Dynamic Contrast-Enhanced MRI with a Li-RADS Concept Bottleneck %A Killian Monnin %A Patrick Jeltsch %A Lucia Fernandes-Mendes %A Vasco Cazzagon %A Murat Yüce %A Vivek Yadav %A Mario Jreige %A Marianna Gulizia %A Montserrat Fraga Christinet %A Raphaël Girardet %A Clarisse Dromain %A Bachir Taouli %A Naïk Vietti-Violi %A Jonas Richiardi %B Proceedings of The 9th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2026 %E Yuankai Huo %E Mingchen Gao %E Chang-Fu Kuo %E Yueming Jin %E Ruining Deng %F pmlr-v315-monnin26a %I PMLR %P 3283--3313 %U https://proceedings.mlr.press/v315/monnin26a.html %V 315 %X We propose an explainable end-to-end framework for hepatocellular carcinoma (HCC) diagnosis on dynamic contrast-enhanced (DCE) liver MRI. Our method embeds Liver Imaging Reporting and Data System (Li-RADS)–inspired concepts into the network via a multi-head concept bottleneck. A 2.5D EfficientNet backbone processes lesion-centred multiphase MRI crops, and a 4-head architecture jointly predicts continuous soft labels for non-rim arterial phase hyperenhancement (APHE), portal venous/delayed washout and capsule, lesion morphology, and a LR-5 score (definite HCC vs non-HCC) based on the Li-RADS guidelines. Soft labels are derived automatically from intra-lesional, peri-lesional and parenchymal intensity patterns, and the network is trained with uncertainty-weighted losses to balance concept prediction, contrast regression and HCC classification. On our cohort, the Li-RADS–inspired bottleneck substantially improves NormGrad explanation accuracy, geometric stability and intensity robustness while maintaining PR AUC comparable to a single-head baseline, highlighting an interpretable alternative to a black-box HCC classifier.
APA
Monnin, K., Jeltsch, P., Fernandes-Mendes, L., Cazzagon, V., Yüce, M., Yadav, V., Jreige, M., Gulizia, M., Fraga Christinet, M., Girardet, R., Dromain, C., Taouli, B., Vietti-Violi, N. & Richiardi, J.. (2026). Explainable HCC Diagnosis on Dynamic Contrast-Enhanced MRI with a Li-RADS Concept Bottleneck. Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 315:3283-3313 Available from https://proceedings.mlr.press/v315/monnin26a.html.

Related Material