# Dataset-Origin Signatures and Shortcut Learning in Screening Mammography AI: A Cross-Dataset Case Study

arXiv: https://arxiv.org/abs/2607.15416 · HTML: https://mlsanity.com/paper/2607.15416/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 8 of its 39 references that are in the map.
- First posted on arXiv 16 Jul 2026.

## Builds on (most cited first, 8 of 8)

- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [Confounding variables can degrade generalization performance of radiological deep learning models](https://mlsanity.com/paper/1807.00431/) (2018)
- [Domain Adaptation for Medical Image Analysis: A Survey](https://mlsanity.com/paper/2102.09508/) (2021)
- [VinDr-Mammo: A large-scale benchmark dataset for computer-aided diagnosis in full-field digital mammography](https://mlsanity.com/paper/2203.11205/) (2022)
- [A Decade's Battle on Dataset Bias: Are We There Yet?](https://mlsanity.com/paper/2403.08632/) (2024)
- [OPTIMAM Mammography Image Database: a large scale resource of mammography images and clinical data](https://mlsanity.com/paper/2004.04742/) (2020)
- [Mammo-CLIP: A Vision Language Foundation Model to Enhance Data Efficiency and Robustness in Mammography](https://mlsanity.com/paper/2405.12255/) (2024)
- [Full Field Digital Mammography Dataset from a Population Screening Program](https://mlsanity.com/paper/2411.02710/) (2024)

## Built on by (most cited first, 1 of 1)

- [TRUST: Threshold-Recalibrated Uncertainty-Safe Training for Certified Dismissal in Breast Cancer Screening](https://mlsanity.com/paper/2609.00300/) (2026)

## Abstract (arXiv)

Reliable AI for screening mammography requires training data representative of the low cancer prevalence and subtle abnormalities found in screening populations. We examined whether supplementing such data with biopsy-confirmed cases from abnormal-enriched external datasets improves performance. Using the Newfoundland and Labrador Breast Screening Dataset (NLBSD) alongside CBIS-DDSM and CMMD, we evaluated an EfficientNet-B5 encoder initialized with Mammo-CLIP weights as a frozen linear probe under consistent preprocessing and patient-level splits. The NLBSD-only model achieved an AUC-ROC of 0.737 (95% CI [0.686, 0.785]). Adding external positive cases reduced performance in every configuration (AUC-ROC = 0.620--0.644; DeLong test, Holm-corrected p < 0.05), with degradation increasing as additional sources were introduced. Domain-matched evaluation produced modest gains only when the training and test domains coincided, and no configuration surpassed the NLBSD-only model. As a diagnostic, we reframed the task as predicting each examination's dataset of origin. The datasets were separated almost perfectly despite identical preprocessing, indicating that dataset-specific characteristics strongly influence the learned representation. These findings show that na"ively pooling abnormal-enriched mammography datasets can introduce domain shift that outweighs the benefit of additional positive cases. Differences in acquisition, intensity mapping, and dataset construction persist after normalization, motivating domain-aware strategies for combining heterogeneous mammography datasets.