Diagnostic Performance of a Novel Optical Coherence Tomography-Derived Fractional Flow Reserve: A Prospective Multicenter Study.

UNLABELLED: Optical coherence tomography-derived fractional flow reserve (OCT-FFR) offers a combined assessment of anatomical and physiological lesions. However, its diagnostic accuracy needs improvement. In this prospective, multicenter, single-arm study, 210 patients across four centers with indications for percutaneous coronary intervention underwent simultaneous FFR and OCT-FFR assessments. OCT-FFR was computed using a novel multitask learning algorithm that integrated lumen segmentation and bifurcation detection, enabling fluid dynamics-based modeling of pressure gradients across coronary lesions. The primary outcome was the diagnostic accuracy of OCT-FFR. The per-protocol analysis included 200 lesions in 200 patients. The mean stenosis diameter was 67.43 ± 9.01%, and mean FFR was 0.85 ± 0.08. Using FFR ≤ 0.80 as the reference, OCT-FFR showed an accuracy of 98.0%, sensitivity of 93.6%, specificity of 99.3%, and positive predictive value of 97.8%. It showed strong correlation (r = 0.95, p < 0.001) and agreement (mean difference: -0.001 ± 0.026) with FFR. Furthermore, it demonstrated superior diagnostic accuracy compared with previous methods without patient-specific blood pressure (98.0% vs. 90.0%) or minimal lumen area optimization (98.0% vs. 93.0%) (both p < 0.01). The new algorithm exhibits excellent diagnostic performance and strong concordance with invasive FFR, potentially optimizing revascularization strategies. However, larger prospective cohorts are needed to validate these findings and confirm generalizability. TRIAL REGISTRATION: ChiCTR2500098052.