Diagnostic Performance of a Machine Learning-Based CT-Derived FFR in Detecting Flow-Limiting Stenosis

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• Non-invasive quantification of fractional flow reserve (FFRCT) using advanced AI software and CT scanners shows potential for detecting coronary ischemia.
• Previous studies have explored FFRCT, but performance with earlier CT generations requires further evaluation.

Purpose of the Study:

• To evaluate the diagnostic performance of FFRCT in detecting significant coronary artery disease (CAD).
• To compare FFRCT against invasive FFR (iFFR) using previous generation CT scanners (128 and 256-detector rows).

Main Methods:

• Retrospective study including 93 patients and 152 vessels undergoing coronary CT angiography and invasive FFR.
• Images acquired using Siemens Somatom Definition Flash (256-detector rows) and AS+ (128-detector rows) CT scanners.
• FFRCT and minimal lumen area (MLA) analyzed with cFFR software; obstructive CAD defined as CTA lumen reduction ≥50%, flow-limiting stenosis as iFFR ≤0.8.

Main Results:

• Good agreement between FFRCT and iFFR (bias: -0.02).
• FFRCT performance was significantly superior to visual stenosis classification (AUC 0.93 vs. 0.61) and MLA (AUC 0.93 vs. 0.75).
• Optimal FFRCT cut-off of 0.85 demonstrated 87% sensitivity and 86% specificity, reducing false positives.

Conclusions:

• Machine learning-based FFRCT using older CT scanners demonstrates good diagnostic performance for CAD detection.
• FFRCT can effectively reduce the need for invasive procedures.
• The study validates FFRCT as a reliable non-invasive tool for assessing coronary artery disease.