Noninvasive MRA-derived fractional flow for intracranial stenosis: methodological evaluation and hemodynamic insights

Objective Current fractional flow (FF) approaches in quantifying the hemodynamic severity of intracranial atherosclerotic stenosis (ICAS) are invasive or require angiography. This study primarily evaluated a noninvasive FF method derived from magnetic resonance angiography (MRA-FF) against established reference standards, with an exploratory assessment of its ability to reflect hemodynamically relevant impairment. Methods In this retrospective study of 50 patients with ICAS, MRA-FF was computed from 3D vascular models using patient-specific flow boundary conditions. Agreement with invasive pressure-wire-derived FF (PW-FF, n = 18) and angiography-based FF (DSA-FF, n = 50) was assessed using intraclass correlation coefficients (ICC). Equivalence to DSA-FF was evaluated using the two one-sided tests procedure (TOST; δ = ±0.05). An exploratory analysis examined the ability of MRA-FF to discriminate hypoperfusion using receiver operating characteristic (ROC) analysis in the subset of patients with available perfusion imaging ( n = 38). Results MRA-FF showed moderate-to-good agreement with PW-FF (ICC = 0.743; 95% CI, 0.437–0.895) and good agreement with DSA-FF (ICC = 0.784; 95% CI, 0.649–0.871). Equivalence to DSA-FF was confirmed by TOST (90% CI of difference, −0.016 to 0.035; p < 0.01). In the perfusion subset, MRA-FF was lower in patients with hypoperfusion than in those without ( p = 0.022). The area under the ROC curve for discriminating hypoperfusion was 0.723 (95% CI, 0.54–0.90). Within this dataset, the exploratory optimal cutoff value was 0.855. Conclusion MRA-FF demonstrates good agreement with invasive and angiographic reference methods and shows the ability to discriminate hypoperfusion in an exploratory analysis. These findings support the feasibility and potential clinical utility of a fully noninvasive FF approach for functional assessment and hemodynamic characterization, although further external validation is required.