Faculty, Staff and Student Publications
Publication Date
10-1-2023
Journal
UNSURE2025 Uncertainty for Safe Utilization of Machine Learning in Medical Imaging
DOI
10.1007/978-3-031-44336-7_15
PMID
40443712
PMCID
PMC12120689
PubMedCentral® Posted Date
5-29-2025
PubMedCentral® Full Text Version
Author MSS
Abstract
Clinically-deployed deep learning-based segmentation models are known to fail on data outside of their training distributions. While clinicians review the segmentations, these models do tend to perform well in most instances, which could exacerbate automation bias. Therefore, it is critical to detect out-of-distribution images at inference to warn the clinicians that the model likely failed. This work applies the Mahalanobis distance post hoc to the bottleneck features of a Swin UNETR model that segments the liver on T1-weighted magnetic resonance imaging. By reducing the dimensions of the bottleneck features with principal component analysis, images the model failed on were detected with high performance and minimal computational load. Specifically, the proposed technique achieved 92% area under the receiver operating characteristic curve and 94% area under the precision-recall curve and can run in seconds on a central processing unit.
Keywords
Out-of-distribution detection, Swin UNETR, Mahalanobis distance, Principal component analysis
Published Open-Access
yes
Recommended Citation
Woodland, McKell; Patel, Nihil; Taie, Mais Al; et al., "Dimensionality Reduction for Improving Out-of-Distribution Detection in Medical Image Segmentation" (2023). Faculty, Staff and Student Publications. 4260.
https://digitalcommons.library.tmc.edu/uthgsbs_docs/4260
Included in
Bioinformatics Commons, Biomedical Informatics Commons, Genetic Phenomena Commons, Medical Genetics Commons, Oncology Commons