Conformal confidence sets for biomedical image segmentation

Samuel Davenport

Keywords: Deep learning, neural networks, uncertainty quantification, confidence sets

TL;DR: Conformal uncertainty quantification for the output of black-box image segmentation models

Abstract: We develop confidence sets which provide spatial uncertainty guarantees for the output of a black-box machine learning model designed for image segmentation. To do so we adapt conformal inference to the imaging setting, obtaining thresholds on a calibration dataset based on the distribution of the maximum of the transformed logit scores within and outside of the ground truth masks. We prove that these confidence sets, when applied to new predictions of the model, are guaranteed to contain the true unknown segmented mask with desired probability. We show that learning appropriate score transformations on an independent learning dataset before performing calibration is crucial for optimizing performance. We illustrate and validate our approach on polyps colonscopy, brain imaging and teeth datasets. To do so we obtain the logit scores from deep neural networks trained for polyps, brain mask and tooth segmentation segmentation. We show that using distance and other transformations of the logit scores allows us to provide tight inner and outer confidence sets for the true masks whilst controlling the false coverage rate.

Primary Area: applications to computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 7562