DETECTION AND QUANTIFICATION OF BREAST ARTERIAL CALCIFICATION AS A BIOMARKER OF CARDIOVASCULAR RISK IN WOMEN

Cardiovascular risk is underestimated in women: traditional risk scores do not consider sex-specific risk factors. As a matter of fact, cardiovascular disease (CVD) is the leading cause of mortality in the female population, and women are often undertreated and present worse cardiovascular outcomes than men. In a recently published call to action, the American Heart Association urges a cross-disciplinary approach to reduce the gender gap in cardiovascular health. In this context, breast arterial calcifications (BAC), an expression of Monckeberg’s sclerosis involving the tunica media of medium-caliber breast arteries, present as a promising sex-specific cardiovascular risk biomarker. BAC are observed in around 13% of mammograms and are associated with hypertension and increasing age. Interestingly, BAC have not shown any association with other notable cardiovascular risk factors, such as smoking, underlying their distinct pathogenesis from atherosclerotic plaques. Nevertheless, women with BAC have an increased risk of cardiovascular adverse events, such as acute myocardial infarction, ischemic stroke, CVD death and a moderate-severe BAC burden is associated with coronary artery disease, indicating how the role of BAC in this setting may complement to that of traditional risk factors. Still, at present the clinical application of BAC assessment is hampered by the lack of a standardised, robust, fast, and reliable quantification method, an issue that could potentially be solved by artificial intelligence-based methods. As such, more widespread assessment of BAC could help improve cardiovascular risk stratification in women supporting the decision-making toward appropriate lifestyle changes and other personalised preventive strategies, leveraging mammographic screening practice. This thesis will thus provide an overview of the current attitude of European breast radiologists towards BAC assessment in mammography, present a score for BAC quantification that is quick to use and reproducible, and implement a deep learning-based tool for BAC detection and quantification.