Mammography-Based Artificial Intelligence for Breast Arterial Calcification Detection, Quantification, and Cardiovascular Risk Assessment: A Systematic Review
Journal of Clinical Medicine · 17.11.2026 · 15 (22) · 9003
DOI: 10.3390/jcm15229003 ↗Kurz gefasst
Systematische Übersicht in der Zeitschrift Journal of Clinical Medicine, online erschienen am 17.11.2026. Verfasst von Romuald Ferré und 2 weiteren Autor:innen. Die Arbeit nennt 33 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Background/Objectives: Artificial intelligence (AI) may automate breast arterial calcification (BAC) assessment and extract cardiovascular information from mammograms. In this study, we review technical BAC assessment, computed tomography (CT)-derived coronary artery calcification (CAC) surrogate assessment, and cardiovascular associations and prediction.
Methods
MEDLINE and Embase were searched through 29 August 2026, supplemented by citation searching. Reports were synthesized narratively by target task, reference standard, and predictor type. QUADAS-2 and PROBAST informed risk-of-bias assessment.
Results
Database screening yielded 17 eligible reports, and citation searching added three, resulting in 20 reports from 19 studies. These included 12 technical BAC reports, two CAC-surrogate reports, and six cardiovascular reports comprising five BAC-specific studies and one whole-mammogram study. BAC-detection areas under the receiver operating characteristic curve (AUCs) reached 0.95–0.96; segmentation Jaccard coefficients were approximately 0.58–0.61, and one small internal validation reported a Dice of 0.89. Mammography-only CAC AUCs ranged from 0.761 to 0.850 across thresholds; an AI-derived BAC score yielded AUC 0.64 for marked CAC. In an external health system cohort, severe versus absent BAC was associated with major adverse cardiovascular events after adjustment for Predicting Risk of cardiovascular disease EVENTs (PREVENT) (hazard ratio 2.80; 95% confidence interval 2.36–3.32). Separately, adding categorical BAC to PREVENT increased the external cohort C-index from 0.62 to 0.64.
Conclusions
AI enables automated BAC assessment, and BAC-derived measures show prognostic associations with limited evidence of incremental predictive performance. Whole-mammogram models provide complementary evidence that cannot establish BAC-specific prognostic value. These tasks require separate interpretation. Standardized measurement, external validation, calibration, and prospective evidence of clinical benefit remain necessary.
Zitieren
APA
Ferré, R., Benefield, T., & Kuzmiak, C. M. (2026). Mammography-Based Artificial Intelligence for Breast Arterial Calcification Detection, Quantification, and Cardiovascular Risk Assessment: A Systematic Review. Journal of Clinical Medicine, 15(22), 9003. https://doi.org/10.3390/jcm15229003
Vancouver
Ferré R, Benefield T, Kuzmiak CM. Mammography-Based Artificial Intelligence for Breast Arterial Calcification Detection, Quantification, and Cardiovascular Risk Assessment: A Systematic Review. Journal of Clinical Medicine. 2026;15(22):9003. doi:10.3390/jcm15229003
BibTeX
@article{Ferr2026,
author = {Ferré, Romuald and Benefield, Thad and Kuzmiak, Cherie M.},
title = {Mammography-Based Artificial Intelligence for Breast Arterial Calcification Detection, Quantification, and Cardiovascular Risk Assessment: A Systematic Review},
journal = {Journal of Clinical Medicine},
year = {2026},
volume = {15},
number = {22},
pages = {9003},
doi = {10.3390/jcm15229003}
}
EUVIMED