Energy loss minimization-based side branch flow model for FFR calculation based on intracoronary images

Fractional flow reserve (FFR) based on intracoronary images, referred to as iFFR, is considered an important functional assessment index for diagnosing coronary stenosis. However, intracoronary images lack side branch information, making accurate side branch flow compensation highly challenging. Existing methods using bifurcation fractal law or supplementary side branch geometry fail to achieve fast and accurate side branch flow compensation.

To address this challenge, we proposed an Energy Loss Minimization-based Side Branch Flow Model (ESBF) model to directly estimate side branch blood flow. Using the principle of minimum energy loss, we created a dataset that correlated vessel geometric features with flow ratios and then used a fully connected neural network to establish the relationship. By extracting vessel geometric features from a single frame of Coronary Angiography images, the model can estimate each side branch flow to calculate the iFFR.

Our study included 116 vessels from 89 patients. Using invasive FFR as a reference, the Pearson correlation coefficient of iFFR based on the ESBF model (iFFR_esbf) was 0.93. iFFR_esbf achieved a high diagnostic performance in identifying the hemodynamic significance of coronary stenosis, with an AUC of 0.96, sensitivity of 90.2%, specificity of 92.3%, and overall accuracy of 91.3%.

The proposed model can accurately perform side branch flow compensation in the analysis of iFFR_esbf. iFFR_esbf analysis demonstrates high feasibility and good diagnostic performance in evaluating the functional significance of coronary stenosis.