Portable 0.23-T MRI for acute brain injury after extracorporeal cardiopulmonary resuscitation: a prospective cohort study

Acute brain injury (ABI) after extracorporeal cardiopulmonary resuscitation (ECPR) is difficult to characterise early because conventional MRI is rarely feasible during extracorporeal membrane oxygenation. We assessed the feasibility and safety of protocolised portable 0.23-T MRI within 72 h after return of spontaneous circulation (ROSC) and described ABI phenotypes and 90-day neurological outcomes. In this prospective cohort study, adults with out-of-hospital cardiac arrest who remained comatose underwent portable 0.23-T MRI within 72 h after ROSC. The primary outcome was MRI completion without serious adverse events. Secondary outcomes were image quality, MRI-defined ABI phenotypes adjudicated by blinded neuroradiologists, and 90-day Cerebral Performance Category (CPC). Of 58 screened patients, 36 were included. MRI was completed in all patients (36/36, 100%; exact 95% CI 90.3–100.0) at a median of 1.8 days after ROSC. One transient decrease in extracorporeal membrane oxygenation flow occurred because of cannula compression by the scanner cover (1/36, 2.8%; 95% CI 0.1–14.5); no serious adverse events occurred. MRI identified ABI in 28 patients (77.8%; 95% CI 60.8–89.9): hypoxic–ischaemic brain injury in 20 (55.6%), ischaemic stroke in 7 (19.4%), and intracranial haemorrhage in 1 (2.8%). Favourable neurological outcome (CPC 1–2) occurred in 7 patients. Overall, 25 of 36 patients (69.4%) died, with multiple organ failure being the leading cause in the non-ABI group (2/3, 66.7%) and brain death in the ABI group (11/22, 50.0%). Specific lesion imaging characteristics were associated with unfavorable neurological outcomes. Protocolised portable 0.23-T MRI was feasible in comatose ECPR patients. Early MRI revealed a high ABI burden, predominantly hypoxic–ischaemic injury. Identifying specific imaging features of these heterogeneous lesions may provide exploratory information for prognostic assessment.