Abstract Background and aims Current etiological work-up after non-traumatic intracerebral hemorrhage (ICH) relies on digital subtraction angiography (DSA) to detect arteriovenous shunts (AVS) that may require urgent treatment, despite DSA invasiveness and the low prevalence of AVS. A noninvasive technique able to safely exclude AVS could fundamentally change diagnostic pathways. We evaluated the diagnostic performance of arterial spin labeling (ASL), a noninvasive MRI technique, for AVS detection in the acute phase of ICH. Methods This prospective multicenter diagnostic accuracy study was conducted within the OPTICH cohort between October 2022 and October 2025 across three French tertiary centers. Consecutive adults with non-traumatic ICH underwent standardized MRI including ASL at day 3 post-ICH. ASL images were independently analyzed by two neuroradiologists. Diagnostic performance metrics were calculated using the final etiological diagnosis as reference. Results Among 277 patients analyzed (median age 71 years; 58.8% men), the reference diagnosis identified 24 AVS (8.7%). Interobserver agreement for ASL interpretation was excellent (κ = 0.88). ASL demonstrated a sensitivity of 95.8% (95% CI 79.8–99.3), specificity of 94.5% (90.9–96.7), negative predictive value of 99.6% (98.1–99.9), and overall accuracy of 94.6% (91.3–96.7). All AVS detected by DSA were identified by ASL; three ASL-positive AVS were not detected on initial DSA, and no ASL-negative patient had a positive initial DSA. Conclusions ASL MRI provides a reliable, noninvasive strategy to exclude AVS in acute ICH and supports a paradigm shift in etiological work-up, in which routine DSA may be avoided in ASL-negative patients without compromising diagnostic safety. Conflict of interest All authors declare that they have nothing to disclose.
Beaufort et al. (2026) studied this question.