Time-specific comparison of gray-white matter ratio models for predicting outcomes after cardiac arrest: basal ganglia as the leading model.

BACKGROUND To identify optimal time-specific gray-white matter ratio (GWR) models for predicting poor neurological outcomes after cardiac arrest (CA). METHODS We retrospectively analyzed a prospective single-center cohort of comatose adults who survived CA. Brain computed tomography (CT) scans were performed at three predefined phases following the return of spontaneous circulation (ROSC): the early phase (<12 h), intermediate phase (12-72 h), and recovery phase (>72 h to 1 week). Patients with ≥1 scan were included. Five GWR models (basal ganglia [BG], cortical [CO], average [AV], simplified [SI], and cerebellar [CE]) and a combined model (BG+CE) were evaluated for their ability to predict poor neurological outcomes (cerebral performance category of 3-5) at discharge. RESULTS In all, 173 patients were included, 40 (23.1%) of whom achieved favorable neurological outcomes at discharge. The five GWR values were consistently lower in the poor-outcome group across multiple phases. In the early phase, the area under the curves (AUCs) for GWR-BG, GWR-AV, GWR-CE, GWR-SI, and GWR-CO were 0.83 (95% CI 0.75-0.91), 0.78 (95% CI 0.68-0.87), 0.73 (95% CI 0.61-0.84), 0.72 (95% CI 0.62-0.82), and 0.64 (95% CI 0.51-0.74), respectively. In the intermediate phase, the AUCs increased to 0.87 (95% CI 0.77-0.95), 0.88 (95% CI 0.78-0.96), 0.84 (95% CI 0.73-0.93), 0.81 (95% CI 0.70-0.92), and 0.79 (95% CI 0.66-0.88), respectively, while in the recovery phase, they decreased to 0.76 (95% CI 0.63-0.88), 0.75 (95% CI 0.61-0.86), 0.74 (95% CI 0.60-0.87), 0.64 (95% CI 0.49-0.77), and 0.67 (95% CI 0.52-0.80), respectively. GWR-BG remained comparable to the other individual models. The combined model achieved AUCs of 0.87, 0.90, and 0.80 across the three phases. In the early phase, the combined model significantly outperformed the other four individual models and exhibited numerically higher performance than GWR-BG. CONCLUSION GWR-BG was superior or comparable to the other GWR models across all three post-ROSC phases; GWR-CE was validated as a predictor of poor neurological outcomes, and the combination of GWR-BG and GWR-CE further enhanced early-phase prediction.