Enhanced Detection of Acute Ischemic Stroke With Low-Field MRI

BACKGROUND: Portable, low-field magnetic resonance imaging (MRI) has the potential to expand access to neuroimaging in environments where conventional MRI is limited. However, diffusion-weighted imaging at low magnetic field is challenged by a low signal-to-noise ratio and gradient strength, which may limit diagnostic confidence in acute ischemic stroke evaluation, particularly for very small strokes. In this study, we evaluated a combination of novel pulse sequences and low-field MRI hardware to enhance lesion detection. METHODS: Patients with a suspected diagnosis of acute ischemic stroke were prospectively enrolled at 3 centers. Diffusion-weighted imaging was performed using a single-direction (SD) or a custom multi-direction (MD) sequence comprising 3 orthogonal directions. Imaging was acquired on 2 0.064 T hardware versions: a first-generation C-arm system (Swoop v1) and a next-generation H-arm system featuring optimized gradient amplifiers, form factor, and cooling system (Swoop v2). Diagnostic accuracy and the lower limit of lesion detection were calculated for both SD and MD images on each system compared with ground-truth MRI (1.5–3 T). RESULTS: A total of 95 patients (n=62 confirmed acute ischemic stroke; n=33 stroke mimics) were included. On SD images, agreement between assessors regarding lesion detection was κ =0.72, and κ =0.84 on MD images. The positive predictive value for differentiating acute ischemic stroke from stroke mimics was 78.2% on SD and 95% on MD images. For SD images, a lesion volume cut point of 0.6 mL yielded a sensitivity of 89% and specificity of 88%. For MD images, the lesion volume cut point was 0.4 mL, with a corresponding sensitivity of 86% and specificity of 83%. MD imaging on the next-generation v2 system improved image uniformity ( P