Classification of Luminal Subtypes in Full Mammogram Images Using Transfer Learning

What this paper does

Molecular subtype is determined from tissue, so the question here is whether the mammogram itself carries any usable signal for it. A ResNet-18 was transferred from a breast abnormality classification task and fine-tuned on the public CMMD dataset for luminal versus non-luminal classification, using image-level labels only — no tumour outlines. It reached mean AUC 0.6688 and mean F1 0.6693 on the test set, a statistically significant improvement over the baseline (p < 0.0001). That is well above chance and well below anything clinically usable, which is the honest reading of the result.

How

A ResNet-18 pretrained on a breast abnormality classification task is fine-tuned for luminal versus non-luminal classification on full-field CMMD mammograms using image-level labels only.

Reported results

• Mean AUC 0.6688 and mean F1 0.6693 on the CMMD test set.
• The improvement over the baseline was statistically significant at p < 0.0001.
• Uses image-level labels only, with no pixel-level tumour annotation.

Relevance

A useful reference point for how much molecular information a mammogram plausibly carries under weak supervision — and a reminder that a significant improvement can still be far from a clinically useful one.

What it does not show

A preprint with no peer review, a single public dataset, and performance around AUC 0.67 that is not clinically actionable.

Citation