Diffusion Imaging Protocol Heterogeneity Biases Ischemic Core Volume, Location, and Clinical Associations in Acute Stroke.

Abstract

Background: Diffusion-weighted magnetic resonance imaging is essential for diagnosing ischemic stroke and identifying targets for emergency revascularization. Apparent diffusion coefficient (ADC) maps derived from diffusion-weighted magnetic resonance imaging are commonly used to locate the infarct core, but they are not strictly quantitative and can vary across platforms and sites due to technical factors. This retrospective study was conducted to examine how differences in ADC map generation, resulting from varied protocols across platforms and sites, affect the determination of infarct core size, location, and related clinical outcomes in acute stroke.

Methods: In this retrospective study, 726 patients with acute anterior circulation stroke from a cohort of 1210 unique visits to the Lausanne University Hospital between May 2018 and January 2021 were selected, excluding patients with poor quality imaging or no magnetic resonance imaging or clinical information available. Diffusion-weighted magnetic resonance imaging data were used to generate ADC maps as they would appear from different protocols: 2 simulated with low- and medium-angular resolution (4 and 12 diffusion gradient directions) and 1 with high-angular resolution (20 directions). Using DEFUSE criteria and image postprocessing, ischemic cores were localized; core volume, location, and associations to the National Institutes of Health Stroke Scale and modified Rankin Scale scores were compared between the 2 imaging sequences.

Results: Significant differences were observed in the ADC distribution within white matter, particularly in the kurtosis and skewness, with the segmented infarct core volume being higher in protocols with reduced angular resolution compared with the 20-directions data (7.63 mL versus 3.78 mL). The volumetric differences persisted after correcting for age, sex, and type of intervention. Infarcted voxel's locations varied significantly between the 2 protocols. This variability affected associations between infarct core volume and clinical scores, with lower associations observed for 4-direction data compared with 20-direction data for the National Institutes of Health Stroke Scale at admission and after 24 hours, and modified Rankin Scale after 3 months, further confirmed by multivariate regression.

Conclusions: Imaging protocol heterogeneity leads to significant changes in the ADC distribution, ischemic core location, size, and association with clinical scores. Work is needed in standardizing imaging protocols to improve the reliability of ADC as an imaging biomarker in stroke management protocols to improve the reliability of ADC as an imaging biomarker in stroke management.