What question did this study set out to answer?

Evaluate the performance of AI-QCTISCHEMIA in diagnosing coronary ischemia in diabetic patients.

February 2, 2026

Improved detection of coronary ischemia in patients with diabetes mellitus using a novel artificial intelligence-based ischemia algorithm derived from coronary ct angiography (AI-QCTISCHEMIA)

Key Points

Evaluate the performance of AI-QCTISCHEMIA in diagnosing coronary ischemia in diabetic patients.
Post-hoc analysis of CREDENCE validation cohort with 305 patients and 868 vessels.
Patients underwent various imaging assessments including CCTA, SPECT/PET, FFR-CT, and invasive coronary angiography (ICA).
Diagnostic performance assessed using vessel-level area under the receiver operating characteristic curve (AUC).
Missing values were treated as positive using an intention-to-diagnose approach.
Diabetic patients had higher total plaque burden compared to non-diabetics (20.1% vs. 16.6%, p<0.0001).
AI-QCTISCHEMIA showed superior AUC of 0.842 for ischemia detection compared to FFR-CT (AUC: 0.739) and MPI (AUC: 0.635).
Diagnostic accuracy was equivalent for diabetic and non-diabetic groups.

Abstract

Abstract Background Diabetic patients have a higher risk of coronary artery disease (CAD), greater atherosclerotic burden, faster disease progression, and increased mortality compared to those without DM, making accurate evaluation crucial yet challenging. AI-enhanced coronary computed tomography angiography (CCTA) provides detailed quantitative assessment of coronary plaque and vascular features, offering a reliable estimate of coronary ischemia (AI-QCTISCHEMIA). Purpose To evaluate the diagnostic performance of AI-QCTISCHEMIA in detecting coronary ischemia in patients with diabetes. Methods This post-hoc analysis of the CREDENCE validation cohort included 305 patients (868 vessels) with suspected CAD who underwent CCTA, myocardial perfusion imaging (MPI; SPECT/PET), FFR-CT, and invasive coronary angiography (ICA) with FFR as reference. The diagnostic performance of AI-QCTISCHEMIA in predicting myocardial ischemia was evaluated using vessel-level area under the receiver operating characteristic curve (AUC) analysis, stratified by diabetic status. Missing values were imputed as positive in accordance with an intention-to-diagnose approach. Results Diabetic patients were older (66. 3 ± 9. 4 vs. 63. 6 ± 9. 9 years, p = 0. 025) and had a higher prevalence of hypertension (75. 8%, p = 0. 007), with no differences in sex, BMI, or smoking. On a per-vessel basis, DM patients showed significantly greater total plaque burden (PAV: 20. 1% vs. 16. 6%, p0. 0001), including both calcified (PAVCP: 7. 4% vs. 5. 2%, p=0. 0001) and non-calcified (PAVNCP: 12. 5% vs. 11. 2%, p=0. 014) components. In diabetic patients, AI-QCTISCHEMIA demonstrated superior diagnostic performance (AUC: 0. 842) compared with FFR-CT (AUC: 0. 739, p=0. 003) and MPI (AUC: 0. 635, p 0. 0001). Notably, diagnostic accuracy of AI-QCTISCHEMIA was similar between diabetic and non-diabetic subgroups. Conclusion AI-QCTISCHEMIA exhibits significantly higher diagnostic accuracy for ischemia detection in diabetic patients compared to MPI and FFR-CT. Table 1. Baseline characteristics Table 2. Diagnostic Accuracy

Ask AI

Helpful

Bookmark

View Full Paper