Early Feasibility of Automated Artificial Intelligence Angiography Based Fractional Flow Reserve Estimation.
Despite the evidence of improved patients’ outcome, fractional flow reserve (FFR) is underused in current everyday practice. We aimed to evaluate the feasibility of a novel automated artificial intelligence angiography-based FFR software (AutocathFFR) as a decision supporting tool for interventional cardiologists. AutocathFFR was performed on angiographic images of patients who underwent coronary angiography with pressure wire FFR measurement. Sensitivity and specificity for detection of FFR cut-off of 0.8 were calculated. Thirty-one patients were included in the current study, with a mean age of 64±10 years, 80% were males, 32% patients had diabetes, 39% had prior percutaneous coronary intervention (PCI). The left anterior descending artery was the target vessel in 80% of patients. Automatic lesion detection was successful in all of the lesions with FFR value of ≤ 0.8. The sensitivity of AutocathFFR for predicting a wire based FFR≤0.8 was 88% and the specificity for FFR>0.8 was 93%, with a positive predictive value of 94% and negative predictive value of 87%, indicating an accuracy level of 90% and area under the curve of 0.91. AutocathFFR has excellent accuracy in prediction of wire based FFR and is a promising technology that may facilitate appropriate decision and treatment choices for coronary artery disease patients.