← All publications

conference · 2025 · Published (e-poster)

AI-Assisted Detection of Malignant Breast Cancer in High Breast Density Cohorts: A Retrospective Comparative Study

Adarsh Bhandary Panambur · Anh Hoang Dinh · L. Chung · L. T. Le · Andreas Maier · Alejandro Rodriguez Ruiz · M. Andrews · B. Schmitt · Siming Bayer

DOI ↗Poster ↗
01 / SUMMARY

What this paper does

This follow-up to the ECR 2024 study focuses on the subgroup where mammography is weakest: dense breasts with a suspicious finding. A retrospective 12-month Hanoi Medical University Hospital cohort of 443 patients (BI-RADS 4, 5 and selected 3) was assessed against biopsy-confirmed outcomes — 136 negative cases, 292 with unilateral lesions of which 138 were malignant, and 15 bilateral cases with 3 malignant. Inter-rater agreement between readers was ICC 0.718–0.780 (p < 0.001). The commercial AI independently identified 97% of positive patients correctly, while classifying 21% of biopsied patients as low risk.

02 / METHOD

How

Retrospective comparative analysis of biopsy-confirmed outcomes, inter-reader BI-RADS agreement measured by intraclass correlation, and commercial AI risk classification on a cohort enriched for high breast density.

03 / FINDINGS

Reported results

  • Cohort of 443 patients (BI-RADS 4, 5 and selected 3) over a 12-month period at Hanoi Medical University Hospital.
  • 136 negative cases; 292 patients with unilateral lesions of which 138 were malignant; 15 bilateral cases of which 3 were malignant.
  • Inter-rater agreement ICC 0.718 to 0.780 (p < 0.001).
  • The AI independently identified 97% of positive patients correctly.
  • The AI classified 21% of patients who underwent biopsy as low risk.
04 / WHY IT MATTERS

Relevance

The 97% figure and the 21% figure have to be read together. Reporting only the first would be the kind of selective quotation that makes clinical AI claims untrustworthy.

05 / LIMITATION

What it does not show

Retrospective, single-centre, enriched cohort published as a conference e-poster. An enriched cohort inflates prevalence relative to a screening population, so these numbers do not transfer directly to screening.

06 / REFERENCE

Citation

A. Bhandary Panambur, A. Hoang Dinh, L. Chung, L. T. Le, A. Maier, A. Rodriguez Ruiz, M. Andrews, B. Schmitt, S. Bayer. AI-Assisted Detection of Malignant Breast Cancer in High Breast Density Cohorts: A Retrospective Comparative Study. ECR 2025, EPOS C-21046. doi:10.26044/ecr2025/C-21046

View original abstract

Retrospective comparative study of AI-assisted malignant breast cancer detection in a cohort enriched for high breast density, using biopsy-confirmed outcomes, inter-reader BI-RADS agreement, and commercial AI risk classification at Hanoi Medical University Hospital.

Related work

Connected publications

BE-WISE architecture: pre-contrast, post-contrast and subtraction MRI volumes stacked into a (32, 3, 224, 224) tensor, encoded by a shared Swin-Tiny slice encoder into (32, 768) embeddings, then split into a slice-level lesion predictor producing a per-slice probability curve and a breast-level classifier producing normal, benign or malignant

Bildverarbeitung für die Medizin (BVM) 2026, 10–17

BE-WISE: Breast MRI Evaluation with Weakly-Informed Slice-Level Explanation

Published

Breast MRI models usually need lesions outlined voxel by voxel, which does not scale. BE-WISE asks how far one annotated slice per exam gets you. Pre-contrast, post-contrast and subtraction volumes are stacked and encoded slice by slice with a shared Swin-Tiny transformer. One branch predicts a per-slice lesion probability against a Gaussian target centred on the radiologist's slice; the other pools the same embeddings by attention into a normal / benign / malignant decision. On the multicenter ODELIA dataset this reached test AUC 0.8683 and an ODELIA score of 0.7098, against 0.7968 and 0.5695 for a medical slice transformer baseline.

Explainable AIMedical ImagingDetection / Localization
1 min read → Paper ↗
BE-WISER architecture: pre-contrast, post-contrast and subtraction inputs go through a SwinV2-Tiny slice encoder into slice embeddings, which split into a breast-level classifier module with ABMIL attention, feature aggregation, global head and global logits, and a weak slice-level lesion localization branch with a slice head, slice logits and per-slice class probabilities over normal, benign and malignant

Research Square preprint (in review)

BE-WISER: Class-Aware Weak Slice-Level Supervision for Breast MRI

Preprint · in review

BE-WISE could say which slice contained something; BE-WISER says what that something is. Each slice inside the Gaussian region around the radiologist's marked slice inherits the exam's class — benign or malignant — while the rest are labelled normal, so the slice head becomes three-class instead of binary. On the latest multicenter ODELIA release (926 / 230 / 292 examinations from six institutions) this reached breast-level AUC 0.864 and an ODELIA score of 0.699, against 0.807 and 0.591 for a breast-level baseline. Localization also improved: Hit@1 rose from 0.554 to 0.632 and the mean slice distance fell from 2.838 to 2.505 slices.

Explainable AIDetection / LocalizationMedical Imaging
1 min read → Paper ↗
conference 2026 Vision-Language Models · Foundation Models

European Congress of Radiology (ECR) 2026, scientific exhibit C-15780

Lesion-Aware AI for Mammography: Multi-Dataset Pretraining with ROI-Guided Contrastive Learning and Clinical Image Retrieval

Accepted

Global contrastive pretraining lets a mammography model align on breast density or view rather than on the lesion, because those cues are enough to match an image to its report. This work adds region-of-interest guidance to multi-dataset contrastive pretraining, so localized pathology has to contribute to the representation, and then tests the embeddings on clinical image retrieval — finding visually and diagnostically similar prior cases. Accepted as a scientific exhibit at ECR 2026; quantitative results are not yet publicly available.

Vision-Language ModelsFoundation ModelsMedical Imaging
Details → DOI ↗