Artificial Intelligence–Enabled Quantitative Coronary Plaque and Hemodynamic Analysis for Predicting Acute Coronary Syndrome
Authors' conclusion
AI-enabled plaque and hemodynamic quantification enhanced the predictability for ACS culprit lesions over the conventional coronary CTA analysis.
Study context
- Population
- Acute coronary syndrome (ACS) patients with coronary computed tomography angiography (CTA) 1 month to 3 years before the ACS event; culprit and nonculprit lesions adjudicated by invasive angiography2,088 nonculprit and 363 culprit lesions
- Participants
- 351 patients
- Comparison
- Reference model + AI-QCPHA features vs reference model
- Outcome
- Predictability of risk models for ACS culprit lesions
Methodological detailsdiagnostic/prognostic model development and validation, derivation and validation cohorts
- Diagnostic/prognostic model development and validation
- Derivation and validation cohorts
Main finding
As reported in the abstract
Prediction of ACS culprit lesions (validation cohort, n = 108)
| Group | AUC |
|---|---|
| Reference model | 0.78 |
| Reference model + AI-QCPHA | 0.84 |
AUC
Reference model0.78
Reference model + AI-QCPHA0.84
- Reference model + AI-QCPHA vs Reference model: P < 0.001 — 0.84 vs 0.78
Observational data: associations, not effects.
Further findings
Best AI-QCPHA features (derivation cohort, n = 243)
- FFR across the lesion, plaque burden, total plaque volume, low-attenuation plaque volume, averaged percent total myocardial blood flow
Additive value across timepoints from coronary CTA
- Additive value of AI-QCPHA consistent across different timepoints from coronary CTA
Interval from coronary CTA to ACS event
- Median interval, days (Q1-Q3: 95-645)
- 375
- Presented with MI
- 223 (63.5%)
Original publication
Artificial Intelligence–Enabled Quantitative Coronary Plaque and Hemodynamic Analysis for Predicting Acute Coronary Syndrome. JACC Cardiovasc Imaging. Published 15 May 2024. doi:10.
Summarised from the published abstract; every number is checked against it