Sparse Activations for Interpretable Disease Grading

Kerol R. Djoumessi Donteu, Indu Ilanchezian, Laura Kühlewein, Hanna Faber, Christian F. Baumgartner, Bubacarr Bah, Philipp Berens, Lisa M. Koch
Medical Imaging with Deep Learning, PMLR 227:1-17, 2024.

Abstract

Interpreting deep learning models typically relies on post-hoc saliency map techniques. However, these techniques often fail to serve as actionable feedback to clinicians, and they do not directly explain the decision mechanism. Here, we propose an inherently interpretable model that combines the feature extraction capabilities of deep neural networks with advantages of sparse linear models in interpretability. Our approach relies on straightforward but effective changes to a deep bag-of-local-features model (BagNet). These modifications lead to fine-grained and sparse class evidence maps which, by design, correctly reflect the model’s decision mechanism. Our model is particularly suited for tasks which rely on characterising regions of interests that are very small and distributed over the image. In this paper, we focus on the detection of Diabetic Retinopathy, which is characterised by the progressive presence of small retinal lesions on fundus images. We observed good classification accuracy despite our added sparseness constraint. In addition, our model precisely highlighted retinal lesions relevant for the disease grading task and excluded irrelevant regions from the decision mechanism. The results suggest our sparse BagNet model can be a useful tool for clinicians as it allows efficient inspection of the model predictions and facilitates clinicians’ and patients’ trust.

Cite this Paper

BibTeX
@InProceedings{pmlr-v227-donteu24a, title = {Sparse Activations for Interpretable Disease Grading}, author = {Donteu, Kerol R. Djoumessi and Ilanchezian, Indu and K\"uhlewein, Laura and Faber, Hanna and Baumgartner, Christian F. and Bah, Bubacarr and Berens, Philipp and Koch, Lisa M.}, booktitle = {Medical Imaging with Deep Learning}, pages = {1--17}, year = {2024}, editor = {Oguz, Ipek and Noble, Jack and Li, Xiaoxiao and Styner, Martin and Baumgartner, Christian and Rusu, Mirabela and Heinmann, Tobias and Kontos, Despina and Landman, Bennett and Dawant, Benoit}, volume = {227}, series = {Proceedings of Machine Learning Research}, month = {10--12 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v227/donteu24a/donteu24a.pdf}, url = {https://proceedings.mlr.press/v227/donteu24a.html}, abstract = {Interpreting deep learning models typically relies on post-hoc saliency map techniques. However, these techniques often fail to serve as actionable feedback to clinicians, and they do not directly explain the decision mechanism. Here, we propose an inherently interpretable model that combines the feature extraction capabilities of deep neural networks with advantages of sparse linear models in interpretability. Our approach relies on straightforward but effective changes to a deep bag-of-local-features model (BagNet). These modifications lead to fine-grained and sparse class evidence maps which, by design, correctly reflect the model’s decision mechanism. Our model is particularly suited for tasks which rely on characterising regions of interests that are very small and distributed over the image. In this paper, we focus on the detection of Diabetic Retinopathy, which is characterised by the progressive presence of small retinal lesions on fundus images. We observed good classification accuracy despite our added sparseness constraint. In addition, our model precisely highlighted retinal lesions relevant for the disease grading task and excluded irrelevant regions from the decision mechanism. The results suggest our sparse BagNet model can be a useful tool for clinicians as it allows efficient inspection of the model predictions and facilitates clinicians’ and patients’ trust.} }
Endnote
%0 Conference Paper %T Sparse Activations for Interpretable Disease Grading %A Kerol R. Djoumessi Donteu %A Indu Ilanchezian %A Laura Kühlewein %A Hanna Faber %A Christian F. Baumgartner %A Bubacarr Bah %A Philipp Berens %A Lisa M. Koch %B Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2024 %E Ipek Oguz %E Jack Noble %E Xiaoxiao Li %E Martin Styner %E Christian Baumgartner %E Mirabela Rusu %E Tobias Heinmann %E Despina Kontos %E Bennett Landman %E Benoit Dawant %F pmlr-v227-donteu24a %I PMLR %P 1--17 %U https://proceedings.mlr.press/v227/donteu24a.html %V 227 %X Interpreting deep learning models typically relies on post-hoc saliency map techniques. However, these techniques often fail to serve as actionable feedback to clinicians, and they do not directly explain the decision mechanism. Here, we propose an inherently interpretable model that combines the feature extraction capabilities of deep neural networks with advantages of sparse linear models in interpretability. Our approach relies on straightforward but effective changes to a deep bag-of-local-features model (BagNet). These modifications lead to fine-grained and sparse class evidence maps which, by design, correctly reflect the model’s decision mechanism. Our model is particularly suited for tasks which rely on characterising regions of interests that are very small and distributed over the image. In this paper, we focus on the detection of Diabetic Retinopathy, which is characterised by the progressive presence of small retinal lesions on fundus images. We observed good classification accuracy despite our added sparseness constraint. In addition, our model precisely highlighted retinal lesions relevant for the disease grading task and excluded irrelevant regions from the decision mechanism. The results suggest our sparse BagNet model can be a useful tool for clinicians as it allows efficient inspection of the model predictions and facilitates clinicians’ and patients’ trust.
APA
Donteu, K.R.D., Ilanchezian, I., Kühlewein, L., Faber, H., Baumgartner, C.F., Bah, B., Berens, P. & Koch, L.M.. (2024). Sparse Activations for Interpretable Disease Grading. Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 227:1-17 Available from https://proceedings.mlr.press/v227/donteu24a.html.

Related Material