Comparative validation and clinical utility of an artificial intelligence-based CT-SYNTAX score in complex coronary artery disease.

Authors

Abstract

This study aimed to validate the agreement and clinical utility of an artificial intelligence-based CT-SYNTAX score (AI-CT-SS) against the invasive angiography-based reference standard (ICA-SS) and manual CT-SS in patients with complex coronary artery disease (CAD). This retrospective study was conducted at The First Affiliated Hospital of Soochow University. Patients with complex CAD who underwent both CCTA and ICA within 30 days between June 2016 and October 2024 were included. SYNTAX scores were calculated using ICA (ICA-SS), manual CCTA interpretation (manual CT-SS), and AI (AI-CT-SS). Agreement was evaluated using Cohen's kappa, Bland-Altman plots, and paired t-tests. Among 411 patients (mean age 62.5 ± 12.6 years; 75.4% male), 201 (48.9%) had left main and/or three-vessel disease (LM/3VD). AI-CT-SS showed a small mean bias (0.22; 95% limits of agreement - 19.19 to 19.63) and moderate correlation (r = 0.56) versus ICA-SS, comparable to manual CT-SS. For categorical risk classification, AI-CT-SS achieved low-moderate agreement with ICA-SS (κ = 0.37; 69.2% agreement), higher than manual CT-SS (κ = 0.26; 63.1%). In the LM/3VD subgroup, AI-CT-SS required significantly less processing time (49.2 ± 13.9 s) compared with manual CT-SS and ICA-SS (both P < 0.001). Treatment recommendation agreement between AI-CT-SS and ICA-SS was 80.6% (κ = 0.27), slightly higher than manual CT-SS (76.6%; κ = 0.21). AI-CT-SS showed moderate agreement with ICA-SS and improved scoring efficiency, suggesting potential utility as a supportive adjunct for evaluating anatomical complexity in complex CAD. Future prospective studies are needed to validate its clinical applicability.

Tags

Topics