Pushing a Frozen CXR Foundation Model: A LoRA Partial-Fine-Tuning Study on NIH ChestX-ray14 with a Model-Conditional Label-Flip Sensitivity Analysis
BAI, T.-C.; YEH, S.-C.
Show abstract
Foundation models for chest X-ray interpretation make it possible to adapt specialised visual representations with relatively small trainable modules. We report a retrospective study of Low-Rank Adaptation (LoRA) of Rad-DINO Vision Transformer Base with 14x14 patches (ViT-B/14) for 14-class multi-label classification on the National Institutes of Health (NIH) ChestX-ray14 dataset. The official test labels were accessed during earlier model development and configuration comparisons; consequently, every official-test result in this manuscript is explicitly descriptive and non-confirmatory. We used a patient-disjoint 90/10 split of the official trainval pool (77,988 training and 8,536 validation images) and retained the released 25,596-image test partition. The historically selected all-linear LoRA configuration with safe augmentation and g=37 produced a descriptive test macro AUROC of 0.8462 versus the frozen baseline of 0.8295. Comparisons of target modules, patch-token grids, and a Rad-DINO-specific local query head are reported as retrospective comparisons rather than unbiased model-selection evidence. A confident-learning diagnostic flagged 17,653 of 86,524 trainval images (20.4%); this is a model-based flag rate, not a ground-truth label-error rate. A separate counterfactual relabeling sensitivity analysis, which uses the same model to identify and rescore disagreements, changed the descriptive AUROC to approximately 0.9445 after 6,509 policy-defined flips. This value is not achieved model performance and is not a radiologist-audited label-quality ceiling. We provide a validation-only threshold and artifact protocol for future locked evaluation, but a genuinely untouched holdout and new locked selection are required for a confirmatory headline. The existing Zenodo record contains the 25 publication figures only.
Matching journals
The top 8 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Generative AI Enables Medical Image Segmentation in Ultra Low-Data Regimes 92%
- Segmenting functional tissue units across human organs using community-driven development of generalizable machine learning algorithms 92%
- Weakly supervised classification of rare aortic valve malformations using unlabeled cardiac MRI sequences 92%
Similar papers in this journal
- Dual Adversarial Deconfounding Autoencoder for joint batch-effects removal from multi-center and multi-scanner radiomics data 93%
- Toward Understanding COVID-19 Pneumonia: A Deep-learning-based Approach for Severity Analysis and Monitoring the Disease 92%
- Tracking And Predicting COVID-19 Radiological Trajectory Using Deep Learning On Chest X-Rays: Initial Accuracy Testing 92%
Similar papers in this journal
- BioFuse: An Embedding Fusion Framework for Biomedical Foundation Models 93%
- Deep learning models for COVID-19 chest x-ray classification: Preventing shortcut learning using feature disentanglement 92%
- Enhancing Semantic Segmentation in Chest X-Ray Images through Image Preprocessing: ps-KDE for Pixel-wise Substitution by Kernel Density Estimation 90%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.