Improving Explainability of Disentangled Representations using Multipath-Attribution Mappings

Lukas Klein, João B. S. Carvalho, Mennatallah El-Assady, Paolo Penna, Joachim M. Buhmann, Paul F. Jaeger
Proceedings of The 5th International Conference on Medical Imaging with Deep Learning, PMLR 172:689-712, 2022.

Abstract

Explainable AI aims to render model behavior understandable by humans, which can be seen as an intermediate step in extracting causal relations from correlative patterns. Due to the high risk of possible fatal decisions in image-based clinical diagnostics, it is necessary to integrate explainable AI into these safety-critical systems. Current explanatory methods typically assign attribution scores to pixel regions in the input image, indicating their importance for a model’s decision. However, they fall short when explaining why a visual feature is used. We propose a framework that utilizes interpretable disentangled representations for downstream-task prediction. Through visualizing the disentangled representations, we enable experts to investigate possible causation effects by leveraging their domain knowledge. Additionally, we deploy a multi-path attribution mapping for enriching and validating explanations. We demonstrate the effectiveness of our approach on a synthetic benchmark suite and two medical datasets. We show that the framework not only acts as a catalyst for causal relation extraction but also enhances model robustness by enabling shortcut detection without the need for testing under distribution shifts. Code available at https://github.com/IML-DKFZ/m-pax_lib.

Cite this Paper

BibTeX
@InProceedings{pmlr-v172-klein22a, title = {Improving Explainability of Disentangled Representations using Multipath-Attribution Mappings}, author = {Klein, Lukas and Carvalho, Jo{\~a}o B. S. and El-Assady, Mennatallah and Penna, Paolo and Buhmann, Joachim M. and Jaeger, Paul F.}, booktitle = {Proceedings of The 5th International Conference on Medical Imaging with Deep Learning}, pages = {689--712}, 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/klein22a/klein22a.pdf}, url = {https://proceedings.mlr.press/v172/klein22a.html}, abstract = {Explainable AI aims to render model behavior understandable by humans, which can be seen as an intermediate step in extracting causal relations from correlative patterns. Due to the high risk of possible fatal decisions in image-based clinical diagnostics, it is necessary to integrate explainable AI into these safety-critical systems. Current explanatory methods typically assign attribution scores to pixel regions in the input image, indicating their importance for a model’s decision. However, they fall short when explaining why a visual feature is used. We propose a framework that utilizes interpretable disentangled representations for downstream-task prediction. Through visualizing the disentangled representations, we enable experts to investigate possible causation effects by leveraging their domain knowledge. Additionally, we deploy a multi-path attribution mapping for enriching and validating explanations. We demonstrate the effectiveness of our approach on a synthetic benchmark suite and two medical datasets. We show that the framework not only acts as a catalyst for causal relation extraction but also enhances model robustness by enabling shortcut detection without the need for testing under distribution shifts. Code available at https://github.com/IML-DKFZ/m-pax_lib.} }
Endnote
%0 Conference Paper %T Improving Explainability of Disentangled Representations using Multipath-Attribution Mappings %A Lukas Klein %A João B. S. Carvalho %A Mennatallah El-Assady %A Paolo Penna %A Joachim M. Buhmann %A Paul F. Jaeger %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-klein22a %I PMLR %P 689--712 %U https://proceedings.mlr.press/v172/klein22a.html %V 172 %X Explainable AI aims to render model behavior understandable by humans, which can be seen as an intermediate step in extracting causal relations from correlative patterns. Due to the high risk of possible fatal decisions in image-based clinical diagnostics, it is necessary to integrate explainable AI into these safety-critical systems. Current explanatory methods typically assign attribution scores to pixel regions in the input image, indicating their importance for a model’s decision. However, they fall short when explaining why a visual feature is used. We propose a framework that utilizes interpretable disentangled representations for downstream-task prediction. Through visualizing the disentangled representations, we enable experts to investigate possible causation effects by leveraging their domain knowledge. Additionally, we deploy a multi-path attribution mapping for enriching and validating explanations. We demonstrate the effectiveness of our approach on a synthetic benchmark suite and two medical datasets. We show that the framework not only acts as a catalyst for causal relation extraction but also enhances model robustness by enabling shortcut detection without the need for testing under distribution shifts. Code available at https://github.com/IML-DKFZ/m-pax_lib.
APA
Klein, L., Carvalho, J.B.S., El-Assady, M., Penna, P., Buhmann, J.M. & Jaeger, P.F.. (2022). Improving Explainability of Disentangled Representations using Multipath-Attribution Mappings. Proceedings of The 5th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 172:689-712 Available from https://proceedings.mlr.press/v172/klein22a.html.

Related Material