Health AI Central
Share:
Peer-Reviewed Publication
Med Image Anal2025;105103668.October 1, 2025Journal Article

Robust image representations with counterfactual contrastive learning.

Mélanie Roschewitz1, Fabio De Sousa Ribeiro2, Tian Xia2, Galvin Khara3, Ben Glocker4
1Imperial College London, Department of Computing, London, UK. Electronic address: m.roschewitz21@imperial.ac.uk.
2Imperial College London, Department of Computing, London, UK.
3Kheiron Medical Technologies, London, UK.
4Imperial College London, Department of Computing, London, UK; Kheiron Medical Technologies, London, UK.
PubMed DOI

Abstract

Contrastive pretraining can substantially increase model generalisation and downstream performance. However, the quality of the learned representations is highly dependent on the data augmentation strategy applied to generate positive pairs. Positive contrastive pairs should preserve semantic meaning while discarding unwanted variations related to the data acquisition domain. Traditional contrasti…

Create a free account to keep reading

Free members get 10 full research views every month across publications, clinical trials, FDA clearances, adverse events, and NIH grants. No credit card required.

Create Free AccountSign In

Want unlimited research access? See Pro plans

Data Accuracy Notice: Research intelligence on Health AI Central is aggregated from public sources (PubMed, ClinicalTrials.gov, FDA, NIH, CMS, and others) and refreshed nightly. Classifications and derived metrics are produced by automated methods described in our Methodology. We recommend verifying critical data points against the primary sources before making decisions.