Objectives: This study aims to investigate the differences in plaque characteristics and fat attenuation index (FAI) between in patients who received revascularization versus those who did not receive revascularization and examine whether the machine learning (ML)-based model constructed by plaque characteristics and FAI can predict revascularization.
Methods: This study was a post hoc analysis of a prospective single-centre registry of sequential patients undergoing coronary computed tomography angiography, referred from inpatient and emergency department settings (n = 261, 63 years ± 8; 188 men). The primary outcome was revascularization by percutaneous coronary revascularization. The computed tomography angiography (CTA) images were analysed by experienced radiologists using a dedicated workstation in a blinded fashion. The ML-based model was automatically computed.
Results: The study cohort consisted of 261 subjects. Revascularization was performed in 105 subjects. Patients receiving revascularization had higher FAI value (67.35 ± 5.49 vs -80.10 ± 7.75 Hu, P < .001) as well as higher plaque length, calcified, lipid, and fibrous plaque burden and volume. When FAI was incorporated into an ML risk model based on plaque characteristics to predict revascularization, the area under the curve increased from 0.84 (95% CI, 0.68-0.99) to 0.95 (95% CI, 0.88-1.00).
Conclusions: ML algorithms based on FAI and characteristics could help improve the prediction of future revascularization and identify patients likely to receive revascularization.
Advances in knowledge: Pre-procedural FAI could help guide revascularization in symptomatic coronary artery disease patients.
Keywords: coronary angiography computed tomography; coronary artery disease; fat attenuation index; fractional flow reserve.
© The Author(s) 2024. Published by Oxford University Press on behalf of the British Institute of Radiology. All rights reserved. For permissions, please email: journals.permissions@oup.com.