AI can pick up signs of heart disease in mammograms, new research suggests

New research suggests AI applied to mammograms can pick up signs of heart disease, implying breast cancer screening could also predict risk of heart attack and stroke. The item is shared by Manchester Evening News and elicited mixed public reactions on social media; findings are preliminary and unlikely to have immediate market impact.

Analysis

AI that extracts cardiovascular risk signals from mammograms turns a high-frequency, insurance-covered imaging touchpoint into a continuous cardiovascular screening feed — a potential $ multi-hundred-million per-year TAM for vendors who can productize risk scores and obtain reimbursement. Expect two adoption pathways: (1) software-as-a-service sold to mammography OEMs and imaging centers as an add-on (fast: pilots in 3–12 months, commercial rollouts 12–24 months), and (2) payor-sponsored population screening programs (slow: 18–36 months until guidelines/reimbursement align).

Second-order demand flows favor imaging OEMs and cloud/AI infra suppliers rather than pure cardiology-procedural device makers: more actionable risk flags will increase statins, remote monitoring and outpatient cardiology consults while potentially compressing acute cath lab volume over multi-year horizons. Data ownership and consent create monetization leverage — health systems or payors that stitch mammography AI outputs into claims/EHRs can triage interventions and capture downstream savings, pressuring vendors who only supply one-off analytics without integration.

Key tail risks are classifier bias and litigation: mammography populations skew by sex/age and predictive value for cardiovascular events will vary by ethnicity, raising false-positive cascades and malpractice exposure that could stall adoption. Regulatory gating (FDA/CE) and CPT code creation are binary catalysts — positive coverage decisions in 12–36 months could re-rate OEMs and cloud partners, while adverse bias findings or privacy enforcement could abruptly derail commercial rollouts.

Key Decisions for Investors

• Buy Hologic (HOLX) stock or buy HOLX 12–18 month calls: HOLX controls mammography hardware/service channels and can bundle AI SaaS; upside if 10–20% share gains in software ARR over 2 years; downside is limited to device replacement cycles — set a 20% trailing stop.
• Long NVDA (NVIDIA) or buy 6–12 month NVDA calls to play inference/cloud demand: inference workloads from medical imaging are an underappreciated driver of datacenter utilization; reward skewed if adoption hits pilots in 6–12 months, risk is macro-driven compute pullback.
• Pair trade (12–36 months): long GE HealthCare (GEHC) + HOLX, short Boston Scientific (BSX) sized 1:1 by notional — expresses a shift toward prevention/diagnostics over cath-lab volume; stop-loss 15% on either leg if procedure volumes stay resilient.
• Event-driven watchlist: deploy small asymmetric long positions in public AI-imaging vendors or health IT integrators upon FDA/CMS positive guidance (12–36 month horizon). If a CPT reimbursement or payor pilot is announced, scale to full size; if regulatory/bias concerns emerge, cut to zero within 30 days.