Self-supervised Out-of-Distribution detection for medical imaging
Keywords
Loading...
Authors
Issue Date
2023-05-23
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Out-of-distribution (OOD) detection is an important aspect of deep learning-based medical
imaging approaches for ensuring the safety and accuracy of diagnostic tools. In this paper,
we investigate the effectiveness of three self-supervised learning techniques for OOD detection
in both a labeled RadboudXR dataset and a clinical dataset with OOD data but no
labels. Specifically, we explore two predictive self-supervised techniques and one contrastive
self-supervised technique and evaluate their ability to detect OOD samples. Furthermore,
we evaluate the performance of the state-of-the-art vision transformer model on medical
data both as a standalone method and as the backbone of a self-supervised task. Our results
indicate that the contrastive self-supervised method Bootstrap-Your-Own-latent (BYOL)
and vision transformer model were not effective in detecting OOD samples. However, the
predictive methods performed well on both 2D and 3D data, and demonstrated scalability
in difficulty. These findings suggest the potential utility of self-supervised learning techniques
for OOD detection in medical imaging. When determining an OOD cut-off value
for clinical usage there are, however, problems with separation between datasets. These
challenges suggest that further research is needed before these techniques can be adopted
for clinical usage.
Keywords: Out-of-Distribution detection, medical imaging
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen