A simplified coronary model for diagnosis of ischemia-causing coronary stenosis.

Background And Objective: The functional assessment of the severity of coronary stenosis from coronary computed tomography angiography (CCTA)-derived fractional flow reserve (FFR) has recently attracted interest. However, existing algorithms run at high computational cost. Therefore, this study proposes a fast calculation method of FFR for the diagnosis of ischemia-causing coronary stenosis.

Methods: We combined CCTA and machine learning to develop a simplified single-vessel coronary model for rapid calculation of FFR. First, a zero-dimensional model of single-vessel coronary was established based on CCTA, and microcirculation resistance was determined through the relationship between coronary pressure and flow. In addition, a coronary stenosis model based on machine learning was introduced to determine stenosis resistance. Computational FFR (cFFR) was then obtained by combining the zero-dimensional model and the stenosis model with inlet boundary conditions for resting (cFFR) and hyperemic (cFFR) aortic pressure, respectively. We retrospectively analyzed 75 patients who underwent clinically invasive FFR (iFFR), and verified the model accuracy by comparison of cFFR with iFFR.

Results: The average computing time of cFFR was less than 2 s. The correlations between cFFR and cFFR with iFFR were r = 0.89 (p < 0.001) and r = 0.90 (p < 0.001), respectively. Diagnostic accuracy, sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio, negative likelihood ratio for cFFR and cFFR were 90.7%, 95.0%, 89.1%, 76.0%, 98.0%, 8.7, 0.1 and 92.0%, 95.0%, 90.9%, 79.2%, 98.0%, 10.5, 0.1, respectively.

Conclusions: The proposed model enables rapid prediction of cFFR and exhibits high diagnostic performance in selected patient cohorts. The model thus provides an accurate and time-efficient computational tool to detect ischemia-causing stenosis and assist with clinical decision-making.