LUMINA: A Multi-Vendor Mammography Benchmark with Energy Harmonization Protocol

2026-03-15

AI Summary

This study addresses the limitations of existing public mammography datasets—particularly their insufficient scale, lack of comprehensive clinical labels, and limited diversity in imaging equipment vendors—which hinder the robustness of AI models. To overcome these challenges, the authors construct a new benchmark dataset featuring multi-vendor and multi-energy-mode mammograms, with systematic annotations for both vendor and energy information. They propose a foreground-preserving, pixel-level energy normalization method that aligns image appearance across varying acquisition conditions while retaining lesion morphology. Comprehensive multi-task evaluations on this benchmark demonstrate that two-view models significantly outperform single-view counterparts: EfficientNet-B0 achieves 93.54% AUC in benign-malignant diagnosis, and Swin-T attains 89.43% macro-AUC in breast density classification. The proposed normalization strategy further enhances both model performance and lesion localization focus.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Publicly available full-field digital mammography (FFDM) datasets remain limited in size, clinical labels, and vendor diversity, which hinders the training of robust models. We present LUMINA, a curated, multi-vendor FFDM dataset that explicitly encodes acquisition energy and vendor metadata to expose clinically relevant appearance shifts that current benchmarks overlook. This innovative resource comprises 1824 images from 468 patients (960 benign, 864 malignant) with pathology-confirmed outcomes, BI-RADS assessments, and breast-density annotations. LUMINA spans six acquisition systems and both high- and low-energy styles, exposing vendor- and energy-driven appearance shifts. To reduce cross-vendor/energy drift while preserving lesion morphology, we introduce a foreground-only, pixel-space alignment (''energy harmonization'') that aligns each image to a low-energy reference style, leaving the zero-valued background unchanged. By benchmarking modern CNN and transformer baselines on three clinically meaningful tasks -- diagnosis (benign vs. malignant), BI-RADS risk grouping, and density -- we unify single-vs-two-view evaluation and show that two-view models consistently outperform single-view; in our benchmark, EfficientNet-B0 attains AUC 93.54% for diagnosis, and Swin-T yields the best macro-AUC 89.43% for density. Harmonization improves AUC/ACC across backbones and yields more focal Grad-CAM localization around suspicious regions. Being a richly annotated resource, LUMINA thus provides (a) a vendor-diverse, energy-labeled benchmark and (b) a model-agnostic harmonization protocol that together catalyze reliable, deployable mammography AI.

Problem

• mammography
• vendor diversity
• energy harmonization
• appearance shift
• benchmark

Innovation

• energy harmonization
• multi-vendor mammography
• foreground-only alignment
• FFDM benchmark
• domain shift mitigation