CT morphological index provides incremental value to machine learning based CT-FFR for predicting hemodynamically...

Area of Science:

• Cardiovascular Imaging
• Interventional Cardiology
• Artificial Intelligence in Medicine

Background:

• Fractional flow reserve (FFR) is the gold standard for assessing the functional significance of coronary artery stenosis.
• Non-invasive methods are sought to reduce the need for invasive coronary angiography (ICA).
• Coronary computed tomographic angiography (CCTA) provides anatomical information, but differentiating functionally significant lesions remains challenging.

Purpose of the Study:

• To evaluate the diagnostic performance of the Duke Jeopardy score (DJS) to minimal lumen diameter (MLD) ratio derived from CCTA and machine learning-based CT-FFR.
• To compare these parameters against invasive FFR for differentiating functionally significant from insignificant coronary lesions.

Main Methods:

• Retrospective analysis of 129 patients with 166 coronary lesions who underwent both CCTA and invasive FFR within two weeks.
• Calculation of CT-FFR, DJS/MLD ratio, and other lesion parameters from CCTA.
• Lesions with invasive FFR ≤0.8 were defined as functionally significant.

Main Results:

• The DJS/MLD ratio and CT-FFR showed high diagnostic performance (AUC=0.83 and AUC=0.85, respectively) for identifying significant stenosis.
• Combining CT-FFR with the DJS/MLD ratio improved diagnostic accuracy to 83.7%.
• Several anatomical parameters, including lesion length and stenosis severity, were significantly associated with functionally significant lesions.

Conclusions:

• The combination of machine learning-based CT-FFR and the DJS/MLD ratio provides accurate non-invasive discrimination of flow-limiting coronary lesions.
• This integrated approach enhances the non-invasive assessment of coronary artery disease, potentially reducing the need for invasive procedures.