Exclusion of Arteriovenous Shunts in Intracerebral Hemorrhage Using MRI Arterial Spin Labeling: A Prospective Multicenter Study

Identifying intracranial arteriovenous shunts (AVSs), chiefly arteriovenous malformations (AVMs) and dural arteriovenous fistulas (dAVFs), after nontraumatic intracerebral hemorrhage (ICH) is critical to guide management and prevent recurrence. AVSs are the leading secondary causes of ICH but typically require digital subtraction angiography (DSA) for diagnosis, an invasive, ionizing radiation-based procedure that carries a non-negligible risk of complications and is not readily available in many centers. We assessed the diagnostic accuracy of arterial spin labeling (ASL), a noninvasive MRI technique for detecting AVSs during the etiologic workup of acute ICH.

We conducted a prospective multicenter diagnostic accuracy study, across 3 French tertiary hospitals within the ongoing OPTICH cohort (October 2022-October 2025). Consecutive adults with nontraumatic ICH who underwent standardized MRI including ASL within the first week were included. ASL perfusion maps were independently evaluated by 2 blinded neuroradiologists and rated positive when focal hyperperfusion was observed adjacent to the hematoma. Final ICH etiology was adjudicated using all available imaging and follow-up data, with DSA confirmation for all AVSs. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy were calculated, using the final etiologic diagnosis as the reference standard.

Among 537 enrolled patients, 277 met inclusion criteria (median age 71 years [interquartile range 60-80]; 163 (58.8%) men). An AVS was confirmed in 24 patients (8.7%; 23 AVMs and 1 dAVFs). Interobserver agreement for ASL rating 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), PPV of 62.2% (46.1-75.9), and NPV of 99.6% (98.1-99.9), with an overall accuracy of 94.6% (91.3-96.7). All AVSs identified on initial DSA were detected by ASL; 3 ASL-positive AVSs were occult on initial DSA.

In an unselected, low-prevalence adult ICH population, ASL demonstrated excellent diagnostic accuracy and a near-perfect NPV, indicating that a negative ASL reliably excludes AVS. Main limitations include selective DSA use, few non-AVM shunts, and modest scanner heterogeneity. Pending external validation, these findings support systematic incorporation of ASL as a gatekeeper to DSA into standard MRI protocols during ICH etiologic workup.