Closing the cardiovascular gender gap via AI: the EMABASK breast screening model, Basque Country

Background

Cardiovascular disease is the leading female mortality cause but remains underdiagnosed due to gender bias in risk algorithms. With >80% participation, the Basque Breast Cancer Screening Programme offers a high-reach platform to detect subclinical atherosclerosis via Breast Arterial Calcifications (BAC), a sex-specific biomarker present in 10–12% of mammograms but systematically overlooked.

Objectives

To design and evaluate EMABASK, an AI-assisted BAC detection model aiming to close the cardiovascular gender gap via population-level risk stratification at zero patient burden, facilitating its future incorporation into routine clinical practice.

Methods

Pilot implementation study combining: (1) semi-quantitative BAC visual scoring in the Basque screening cohort; (2) multidisciplinary referral circuit design (Radiology–Primary Care–Cardiology) with high-specificity AI thresholds; (3) Health Technology Assessment (HTA) framework to evaluate cost-effectiveness using NRI indices and Basque demographic data.

Results

Analysis shows 65.35% of detected BAC cases occur in women aged 60–69 (overall cohort prevalence: 6.4%). Among participants, 2.18% present a high-risk profile (high BAC burden plus low mammographic density), the priority cohort for proactive follow-up. Projections indicate 11,000–18,000 women identified annually (∼3,000 high-risk). HTA modelling confirms near-zero marginal detection cost per patient, supporting scalability.

Conclusions

EMABASK provides a scalable, gender-responsive AI framework for cardiovascular screening. High adherence (>80%) and the standardised Basque mammographic network (320,000 women) enable population-wide quantification. Synergising with CARAMEL (Horizon Europe), this model offers an internationally replicable blueprint for evidence-based HTA, closing the gender gap via digital screening transformation.

Key messages

• EMABASK identifies ∼3,000 high-risk women per year currently invisible to cardiovascular systems — using only their routine mammogram. Zero extra cost, zero extra appointments, zero gender bias.

• Cost-neutrality requires 22 MACE/yr at AI scale. Using strict dual criteria (high-grade BAC + low density), our conservative pilot alone projects 300+ events annually—fifteen times above threshold.