Know Your Space: Inlier and Outlier Construction for Calibrating Medical OOD Detectors

Vivek Narayanaswamy, Yamen Mubarka, Rushil Anirudh, Deepta Rajan, Andreas Spanias, Jayaraman J. Thiagarajan
Medical Imaging with Deep Learning, PMLR 227:190-211, 2024.

Abstract

We focus on the problem of producing well-calibrated out-of-distribution (OOD) detectors, in order to enable safe deployment of medical image classifiers. Motivated by the difficulty of curating suitable calibration datasets, synthetic augmentations have become highly prevalent for inlier/outlier specification. While there have been rapid advances in data augmentation techniques, this paper makes a striking finding that the space in which the inliers and outliers are synthesized, in addition to the type of augmentation, plays a critical role in calibrating OOD detectors. Using the popular energy-based OOD detection framework, we find that the optimal protocol is to synthesize latent-space inliers along with diverse pixel-space outliers. Based on empirical studies with multiple medical imaging benchmarks, we demonstrate that our approach consistently leads to superior OOD detection ($15% - 35%$ in AUROC) over the state-of-the-art in a variety of open-set recognition settings.

Cite this Paper

BibTeX
@InProceedings{pmlr-v227-narayanaswamy24a, title = {Know Your Space: Inlier and Outlier Construction for Calibrating Medical OOD Detectors}, author = {Narayanaswamy, Vivek and Mubarka, Yamen and Anirudh, Rushil and Rajan, Deepta and Spanias, Andreas and Thiagarajan, Jayaraman J.}, booktitle = {Medical Imaging with Deep Learning}, pages = {190--211}, 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/narayanaswamy24a/narayanaswamy24a.pdf}, url = {https://proceedings.mlr.press/v227/narayanaswamy24a.html}, abstract = {We focus on the problem of producing well-calibrated out-of-distribution (OOD) detectors, in order to enable safe deployment of medical image classifiers. Motivated by the difficulty of curating suitable calibration datasets, synthetic augmentations have become highly prevalent for inlier/outlier specification. While there have been rapid advances in data augmentation techniques, this paper makes a striking finding that the space in which the inliers and outliers are synthesized, in addition to the type of augmentation, plays a critical role in calibrating OOD detectors. Using the popular energy-based OOD detection framework, we find that the optimal protocol is to synthesize latent-space inliers along with diverse pixel-space outliers. Based on empirical studies with multiple medical imaging benchmarks, we demonstrate that our approach consistently leads to superior OOD detection ($15% - 35%$ in AUROC) over the state-of-the-art in a variety of open-set recognition settings.} }
Endnote
%0 Conference Paper %T Know Your Space: Inlier and Outlier Construction for Calibrating Medical OOD Detectors %A Vivek Narayanaswamy %A Yamen Mubarka %A Rushil Anirudh %A Deepta Rajan %A Andreas Spanias %A Jayaraman J. Thiagarajan %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-narayanaswamy24a %I PMLR %P 190--211 %U https://proceedings.mlr.press/v227/narayanaswamy24a.html %V 227 %X We focus on the problem of producing well-calibrated out-of-distribution (OOD) detectors, in order to enable safe deployment of medical image classifiers. Motivated by the difficulty of curating suitable calibration datasets, synthetic augmentations have become highly prevalent for inlier/outlier specification. While there have been rapid advances in data augmentation techniques, this paper makes a striking finding that the space in which the inliers and outliers are synthesized, in addition to the type of augmentation, plays a critical role in calibrating OOD detectors. Using the popular energy-based OOD detection framework, we find that the optimal protocol is to synthesize latent-space inliers along with diverse pixel-space outliers. Based on empirical studies with multiple medical imaging benchmarks, we demonstrate that our approach consistently leads to superior OOD detection ($15% - 35%$ in AUROC) over the state-of-the-art in a variety of open-set recognition settings.
APA
Narayanaswamy, V., Mubarka, Y., Anirudh, R., Rajan, D., Spanias, A. & Thiagarajan, J.J.. (2024). Know Your Space: Inlier and Outlier Construction for Calibrating Medical OOD Detectors. Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 227:190-211 Available from https://proceedings.mlr.press/v227/narayanaswamy24a.html.

Related Material