Evaluating Federated Vision-Language and Vision-Only Models for Privacy-Aware Breast Cancer Imaging

A federated learning study posted to arXiv (2610.36379v1) found that five vision-language and vision-only models reached ROC-AUC of 0.940-0.990 on normal-versus-sick breast imaging classification, but that acquisition year alone scored 0.974 on the same test partition, suggesting confounding rather than genuine pathology recognition. Benign-versus-malignant discrimination stayed uncertain at ROC-AUC 0.504-0.637, with all confidence intervals including 0.50, while removing the CLIP ViT-B/16 text pathway preserved accuracy and halved communication cost. Client-level differential privacy cut ROC-AUC by 0.375, which the authors attribute to perturbing the full model state rather than only trainable parameters.

arXiv:2610.36379v1 Announce Type: new Abstract: Artificial intelligence AI shows promise for breast imaging, but limited labeled data, institutional data fragmentation, and privacy concerns hinder centralized training. We investigate a privacy-aware federated learning framework combining pretrained vision-language models VLMs with client-level differential privacy. We compare general-purpose CLIP models, biomedical BiomedCLIP, and an ImageNet-pretrained ViT-B/16 baseline on two public breast imaging datasets, with and without differential privacy. We also evaluate a CLIP ViT-B/16 ablation that removes the text pathway. Evaluation includes patient-level ROC-AUC with bootstrap 95% confidence intervals, baseline comparisons, threshold sensitivity, inter-model agreement, and communication cost. Under federated learning, all five models achieve high normal-versus-sick discrimination ROC-AUC: 0.940-0.990 . However, acquisition year alone achieves 0.974 on the same test partition, raising concerns about confounding rather than genuine pathology recognition. Benign-versus-malignant discrimination remains uncertain ROC-AUC: 0.504-0.637 , with all confidence intervals including 0.50. Removing the text pathway preserves classification accuracy while halving communication cost. Differential privacy reduces ROC-AUC by 0.375, attributable to perturbing the full model state rather than only trainable parameters. These findings highlight the importance of evaluating confounding, predictive reliability, privacy implementation, and deployment efficiency when developing federated AI systems for breast imaging.