Research map / Label-efficient and robust learning / Recognition and long-tail
LUMINA: A Multi-Vendor Mammography Benchmark with Energy Harmonization Protocol
Hongyi Pan, Gorkem Durak, Halil Ertugrul Aktas, Andrea M. Bejar, Baver Tutun, Emre Uysal, Ezgi Bulbul, Mehmet Fatih Dogan, Berrin Erok, Berna Akkus Yildirim, Sukru Mehmet Erturk, Ulas Bagci
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add foreground-only histogram harmonization (ablation) | best two-view model · LUMINA diagnosis 224 | AUC (%): 91.42 -> 92.99 | Improved |
| add foreground-only histogram harmonization (ablation) | best two-view model · LUMINA diagnosis 512 | AUC (%): 91.27 -> 93.54 | Improved |
| add foreground-only histogram harmonization (ablation) | EfficientNet-B0 · LUMINA BI-RADS two-class 512 | AUC (%): 89.21 -> 92.75 | Improved |
| add foreground-only histogram harmonization (ablation) | Swin-T · LUMINA density 512 | AUC (%): 88.92 -> 89.14 | No clear effect |
| use two views instead of single view (ablation) | EfficientNet-B0 · LUMINA diagnosis 224 | AUC (%): 90.66 -> 92.99 | Improved |
| use two views instead of single view (ablation) | DenseNet-121 · LUMINA diagnosis 512 | AUC (%): 90.52 -> 88.33 | Hurt |
| use two views instead of single view (ablation) | ResNet-50 · LUMINA diagnosis 224 | AUC (%): 87.42 -> 88.71; accuracy 79.49 -> 76.70 | Mixed |
| raise input resolution from 224 to 512 (ablation) | EfficientNet-B0 single-view · LUMINA diagnosis | AUC (%): 90.66 -> 92.13 | Improved |
| raise input resolution from 224 to 512 (ablation) | Swin-T · LUMINA density | macro-AUC (%): 89.43 -> 89.14 | Hurt |
| share backbone between views instead of independent (ablation) | EfficientNet-B0 two-view · LUMINA diagnosis 224 | AUC (%): 93.54 -> 92.99; params 8.34M -> 4.34M | No clear effect |
| share backbone between views instead of independent (ablation) | EfficientNet-B0 two-view · LUMINA diagnosis 512 | AUC (%): 92.71 -> 93.54 | No clear effect |
| harmonize energy (ablation) | models · LUMINA energy subsets | AUC (%): low-energy gain larger than high-energy | Improved |
Closest papers by meaning
- NaFV-Net: An Adversarial Four-view Network for Mammogram Classification (AAAI 2025 · similarity 0.83)
- Knowledge-based in silico models and dataset for the comparative evaluation of mammography AI for a range of breast characteristics, lesion conspicuities and doses (NeurIPS 2023 · similarity 0.83)
- Act Like a Radiologist: Towards Reliable Multi-View Correspondence Reasoning for Mammogram Mass Detection (TPAMI 2021 · similarity 0.83)
- CSAW-M: An Ordinal Classification Dataset for Benchmarking Mammographic Masking of Cancer (NeurIPS 2021 · similarity 0.83)
- Check and Link: Pairwise Lesion Correspondence Guides Mammogram Mass Detection (ECCV 2022 · similarity 0.82)
- TumorChain: Interleaved Multimodal Chain-of-Thought Reasoning for Traceable Clinical Tumor Analysis (ICLR 2026 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
