Ultrafast brain MRI at 3 T for suspected acute ischemic stroke: interchangeability of the ultrafast and standard protocols

Abstract Purpose To evaluate the interchangeability of the super-resolution deep learning reconstruction (SR-DLR)-based ultrafast protocol with the standard protocol for acute ischemic stroke detection at 3 T. Materials and methods In this prospective study, 100 patients (mean age, 66.1 ± 16.2 years; 65 men) with suspected acute ischemic stroke underwent both the standard and ultrafast protocols (diffusion-weighted, fluid-attenuated inversion recovery, and T2*-weighted imaging). All images were reconstructed using SR-DLR; the ultrafast protocol employed accelerated acquisition with a reduced scan time. Two radiologists independently assessed both protocols for acute ischemic lesion detection, vascular territory involvement, and clinically relevant secondary findings. Interchangeability was evaluated using the individual equivalence index (IEI), with a prespecified ± 0.05 equivalence margin. Results Acute ischemic stroke was confirmed in 30 patients (30%). The ultrafast protocol reduced the mean total acquisition time by 61.7% (1 min 43 s vs. 4 min 29 s; P < 0.001). Interchangeability was demonstrated for patient-level acute ischemic lesion detection (IEI, 0.010 [90% confidence interval (CI): −0.006, 0.026]), vascular territory–level detection (IEIs, − 0.007 to 0.015), and clinically relevant secondary findings (IEI, 0.005 [90% CI: −0.025, 0.035]). Qualitative image quality scores were slightly but consistently higher for the ultrafast protocol across all four assessment items (all P ≤ 0.002). No significant differences were observed in ischemic lesion size or relative signal characteristics between protocols. Conclusion The ultrafast protocol was interchangeable with the standard protocol for detecting acute ischemic stroke, with reduced scan time and comparable or marginally improved image quality; however, this work-in-progress diffusion-weighted imaging algorithm warrants further validation.