AI analysis of CT scans finds 42.9% of patients have coronary inflammation

Commenters spent far less time on the headline prevalence number than on whether the AI-derived CT score measures anything clinically useful. The sharpest criticism was that the preprint appears to tie the new inflammation measure mainly to BMI, while failing to show a relationship with coronary calcium or, more importantly, future cardiac events. Several readers argued that without an outcome-based validation set, the claim that routine labs and risk calculators miss these patients does not establish that the scan is correctly finding dangerous disease. One reader also noted that the paper does report a sex association, pushing back on the claim that BMI is its sole correlate. There was some recognition of the medical rationale for looking beyond calcium scoring: calcium scans can miss noncalcified, rupture-prone plaque, so an imaging-based AI signal could theoretically catch risk that a zero calcium score overlooks. But that promise was outweighed by suspicion that the study is a premium health-service advertisement dressed as AI research. The $17,500 program, AI-agent-heavy marketing, and unclear relationship between membership and study access made readers especially wary. Debate over hs-CRP and related blood tests reinforced the central problem: even if this model detects a different local signal, commenters wanted proof that it predicts outcomes better or changes treatment decisions.