MRI-Derived Patterns of Post-Cardiac Arrest Brain Injury: Linking Anatomical Injury and Cerebrovascular Physiology

Introduction Post-cardiac arrest brain injury (PCABI) patterns across patients are heterogeneous, and conventional clinical variables incompletely capture in vivo injury severity. Here, we evaluated whether anatomically resolved diffusion-weighted MRI (DWI) lesion patterns identify biologically distinct PCABI phenotypes. Methods We conducted a retrospective study of patients with PCABI who underwent brain MRI and intraparenchymal PbtO2; monitoring (n = 24). Ischemic lesion burden was quantified using an MNI atlas-based apparent diffusion coefficient pipeline with a voxel intensity threshold of <650 x 10-6 mm2/s and was summarized at whole-brain, hemispheric, tissue-class, and segment levels. Principal component analysis (PCA) and k-means clustering were used to identify MRI-derived injury patterns for downstream physiological analysis. Results Greater whole-brain lesion burden was associated with lower mean PbtO2 (r = -0.49, p = 0.014), with concurrent associations in the left hemisphere (r = -0.50, p = 0.012), right hemisphere (r = -0.48, p = 0.019), grey matter (r = -0.48, p = 0.018), and white matter (r = -0.49, p = 0.016). PCA identified a dominant lesion-pattern axis associated with PbtO2; (r = 0.48, p = 0.0169). K-means clustering in PC1-PC2 space identified three MRI-derived injury patterns with distinct regional signatures and progressively greater whole-brain lesion burden (p < 0.001). Discussion Greater MRI-defined lesion burden was associated with lower PbtO2 post-cardiac arrest. Anatomically resolved DWI/ADC lesion mapping identified three MRI-derived injury patterns that differed in cerebrovascular physiology, lesion patterning and whole-brain injury burden. Integrating quantitative MRI, intraparenchymal neuromonitoring, and blood biomarkers improved the characterization of the heterogeneity observed in PCABI patients.