Zenodo
How Well Do Foundation Models Generalise in Chest X-Ray Analysis?
By Brian Daniel, Newham Josh and Finlay Ross-Davie.
Abstract
Self-supervised vision foundation models have achieved state-of-the-art performance on chest X-ray analysis, yet their ability to generalise beyond pre-training objectives remains underexplored. This work evaluates EVA-X across two dimensions of generalisation: task generalisation, through sex classification and lung age regression on frontal X-rays, and domain generalisation, through pathology classification on lateral chest X-rays unseen during pre-training. Three model sizes (6M-86M parameters) are benchmarked against a CNN baseline.
Keywords
- Medical Imaging Techniques and Applications
- Foundation (evidence)
- Work (physics)
- Measure (data warehouse)
- Feature (linguistics)
- Fuzzy logic
- Open access
Source
Published in Zenodo (CERN European Organization for Nuclear Research). The abstract is shown as published and remains the work of its authors; read the full article at the source.
Prognia lists this article for reference and education. It is not medical advice. Always read the original and follow current guidelines and local protocols before acting on it.
