Efforts to enhance the specificity of CCTA have led to the development of computational fluid dynamics-based techniques such as computed tomography-derived fractional flow reserve (CT-FFR), aiming to simulate IFFR. CT-FFR, exemplified by HeartFlow CT-FFR, has shown promising results in improving specificity and positive predictive value while maintaining sensitivity. These advancements have influenced treatment strategies and cost-effectiveness by reducing the need for invasive tests. The PLATFORM trial, involving 584 patients with suspected coronary artery disease, compared the CT-FFR-guided strategy to a conservative non-invasive approach. Results indicated a significant reduction of invasive tests by up to 61% with the CT-FFR-guided strategy, without observing differences in clinical outcomes.5) Incorporating CT-FFR into the evaluation of coronary artery disease has demonstrated improved precision in identifying myocardial ischemia and other coronary artery conditions compared to CCTA alone.

Since then, numerous CT-FFR solutions have entered the market with more advanced technique including machine learning.6) However, despite their many advantages, widespread clinical implementation of CT-FFR still faces challenges, including the need for further enhancement of the diagnostic accuracy of CT-FFR and transmitting processed CT images to external interpretation centers for analysis, which can impede the efficiency and timeliness of clinical workflow. Therefore, while CT-FFR holds promise as the most useful non-invasive tool for diagnosing coronary artery disease in the near future, further advancements are necessary to overcome logistical obstacles and facilitate seamless integration into routine clinical practice.

In this issue of the Journal, Hwang et al.7) validated a newly developed on-site CT-FFR compared with invasive FFR in patients with coronary artery disease who underwent CCTA and invasive coronary angiography. The research aims to address the limitations of traditional CT-FFR computation methods by introducing a novel, simpler on-site approach. A total of 319 patients from five tertiary cardiovascular centers in Korea were included in the study. The authors found that the diagnostic accuracy of on-site CT-FFR was 80.6%, sensitivity of 88.1%, specificity of 75.6%, positive predictive value of 65.9% and negative predictive value of 88.9%, meeting the primary hypothesis for hemodynamically significant CAD prediction. They also found that a moderate correlation between CT-FFR and IFFR (r=0.637, p<0.001) and a discriminant ability of area under curve (AUC) 0.86 compared with anatomical obstructive stenosis on CCTA (AUC, 0.64). Notably, CT-FFR outperforms CCTA in predicting hemodynamically significant CAD, particularly in patients with mild to moderate stenosis. Compared to the existing on-site machine running solutions with sensitivities and specificities ranging from 75–85% and 65–85%, respectively, the current method yields comparable results.8)9)10) Moreover, with a high negative predictive value of 88.9% and only 3.9% exclusion rate of CCTA evaluation due to segmentation image issues, this approach proves promising for clinical utilization, providing hemodynamic insight on anatomic obstruction. Overall, the findings suggest that the automated on-site CT-FFR has the potential to enhance CAD diagnosis and treatment decision-making, offering a promising addition to non-invasive imaging modalities.

Presently, there is a lack of direct comparison clinical data between off-site and on-site CT-FFR techniques. However, the potential advantages of on-site analysis are apparent, particularly if it can ensure accuracy. In this context, the results of this study serve as a compelling evidence of how the adoption of on-site CT-FFR can effectively address these challenges, presenting improvements in the applications of CT-FFR in clinical practice.