Novel deep learning CCTA-FFR for detecting functionally significant coronary stenosis: Comparison with iFR

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Noninvasive Cardiology

Background:

• Deep learning-based fractional flow reserve (CT-FFR) from coronary CT angiography (CCTA) allows noninvasive assessment of lesion-specific ischemia.
• Onsite CT-FFR systems offer near-real-time physiologic evaluation, potentially reducing invasive procedures.

Purpose of the Study:

• To evaluate the diagnostic performance of a novel onsite deep learning CT-FFR algorithm.
• To compare the CT-FFR algorithm against invasive instantaneous wave-free ratio (iFR).

Main Methods:

• Retrospective analysis of 44 patients (44 lesions) with CCTA and invasive iFR.
• CT-FFR values generated using an onsite deep learning algorithm (cFFR v6) distal to stenoses.
• Diagnostic performance metrics calculated and compared with CCTA stenosis thresholds and iFR.

Main Results:

• CT-FFR demonstrated high accuracy (81.8%) with an AUC of 0.79, showing good agreement with iFR.
• The algorithm maintained favorable performance in moderate (AUC 0.73) and severe (AUC 0.84) stenoses.
• CT-FFR significantly improved sensitivity, specificity, and AUC compared to CCTA alone.

Conclusions:

• The onsite deep learning CT-FFR algorithm shows good diagnostic agreement with invasive iFR.
• It provides incremental diagnostic value over CCTA alone and is feasible for rapid workstation integration.
• Larger multicenter studies are needed to confirm clinical utility and role.