Artificial Intelligence-Enabled Quantitative Coronary Plaque and Hemodynamic Analysis for Predicting Acute Coronary Syndrome
Bon-Kwon Koo,Seokhun Yang,Jae Wook Jung,Jinlong Zhang,Keehwan Lee,Doyeon Hwang,Kyu-Sun Lee,Joon-Hyung Doh,Chang-Wook Nam,Tae Hyun Kim,Eun-Seok Shin,Eun Ju Chun,Su-Yeon Choi,Hyun Kuk Kim,Young Joon Hong,Hun-Jun Park,Song-Yi Kim,Mirza Husic,Jess Lambrechtsen,Jesper M Jensen,Bjarne L Nørgaard,Daniele Andreini,Pal Maurovich-Horvat,Bela Merkely,Martin Penicka,Bernard de Bruyne,Abdul Ihdayhid,Brian Ko,Georgios Tzimas,Jonathon Leipsic,Javier Sanz,Mark G Rabbat,Farhan Katchi,Moneal Shah,Nobuhiro Tanaka,Ryo Nakazato,Taku Asano,Mitsuyasu Terashima,Hiroaki Takashima,Tetsuya Amano,Yoshihiro Sobue,Hitoshi Matsuo,Hiromasa Otake,Takashi Kubo,Masahiro Takahata,Takashi Akasaka,Teruhito Kido,Teruhito Mochizuki,Hiroyoshi Yokoi,Taichi Okonogi,Tomohiro Kawasaki,Koichi Nakao,Tomohiro Sakamoto,Taishi Yonetsu,Tsunekazu Kakuta,Yohei Yamauchi,Jeroen J Bax,Leslee J Shaw,Peter H Stone,Jagat Narula
DOI: https://doi.org/10.1016/j.jcmg.2024.03.015
Abstract:Background: A lesion-level risk prediction for acute coronary syndrome (ACS) needs better characterization. Objectives: This study sought to investigate the additive value of artificial intelligence-enabled quantitative coronary plaque and hemodynamic analysis (AI-QCPHA). Methods: Among ACS patients who underwent coronary computed tomography angiography (CTA) from 1 month to 3 years before the ACS event, culprit and nonculprit lesions on coronary CTA were adjudicated based on invasive coronary angiography. The primary endpoint was the predictability of the risk models for ACS culprit lesions. The reference model included the Coronary Artery Disease Reporting and Data System, a standardized classification for stenosis severity, and high-risk plaque, defined as lesions with ≥2 adverse plaque characteristics. The new prediction model was the reference model plus AI-QCPHA features, selected by hierarchical clustering and information gain in the derivation cohort. The model performance was assessed in the validation cohort. Results: Among 351 patients (age: 65.9 ± 11.7 years) with 2,088 nonculprit and 363 culprit lesions, the median interval from coronary CTA to ACS event was 375 days (Q1-Q3: 95-645 days), and 223 patients (63.5%) presented with myocardial infarction. In the derivation cohort (n = 243), the best AI-QCPHA features were fractional flow reserve across the lesion, plaque burden, total plaque volume, low-attenuation plaque volume, and averaged percent total myocardial blood flow. The addition of AI-QCPHA features showed higher predictability than the reference model in the validation cohort (n = 108) (AUC: 0.84 vs 0.78; P < 0.001). The additive value of AI-QCPHA features was consistent across different timepoints from coronary CTA. Conclusions: AI-enabled plaque and hemodynamic quantification enhanced the predictability for ACS culprit lesions over the conventional coronary CTA analysis. (Exploring the Mechanism of Plaque Rupture in Acute Coronary Syndrome Using Coronary Computed Tomography Angiography and Computational Fluid Dynamics II [EMERALD-II]; NCT03591328).