Khan · The international journal of cardiovascular imaging 2026 · Retrospective cross-sectional comparative study · n=461

Using AI-Quantitative CT to evaluate the relationship between coronary artery calcium and segment involvement scores in quantifying coronary plaque burden.

Cited 0 times in the scientific literature.

Level 4 - case-series / case-control

Retrospective cross-sectional comparative imaging study

PubMed 41351723 · doi:10.1007/s10554-025-03569-6 · record verified 2026-08-27

What was done

Researchers retrospectively analyzed 461 coronary computed tomography angiography (CCTA) studies using an AI-based quantitative CT platform (Cleerly). Coronary artery calcium (CAC) scores were extracted using natural language processing, and segment involvement scores (SIS) were determined by AI quantification. Unilateral CAD-RADS 2.0 plaque burden scores (PCAC and PSIS) were calculated and evaluated for concordance using Cohen's kappa, Wilcoxon signed-rank, and McNemar's tests, alongside subgroup analyses by symptom status and in patients with a CAC of zero.

What was found

Median participant age was 67 years, and 38.7% were female. PCAC and PSIS were concordant in 23% of cases and discordant in 77%. The higher overall CAD-RADS 2.0 P-score was determined by SIS in 75.5% of cases and by CAC in 1.5%. Agreement was modest (kappa = 0.40, p < 0.01), with SIS systematically assigning higher categories (p < 0.01). Among 97 patients with CAC = 0 (median age 62, 65% female), 95% had non-calcified plaque detected on AI-QCT, and only 4 patients had both CAC and SIS equal to zero. Plaque score distributions did not consistently differ by symptom status (p > 0.05 overall).

Why it matters

Using CAC alone to assess coronary disease risk systematically underestimates plaque burden in CAD-RADS 2.0, particularly in individuals with zero calcium who nonetheless harbor substantial non-calcified plaque. Combining both calcium and segment-based AI quantification provides a more complete assessment of atherosclerotic burden.

Limits

The study is retrospective and relies on a single proprietary AI platform (Cleerly). Clinical outcomes, event rates, and downstream management impacts were not evaluated, leaving the prognostic superiority of SIS-driven reclassification unconfirmed in this dataset.

Cited by