Structural- and functional connectivity analyses for prediction of neurological outcome and cognitive functioning after cardiac arrest

Introduction: Neurological prognostocation in comatose patients after cardiac arrest remains uncertain for many patients, even with current multimodal approaches. Furthermore, there is a lack of predictors for long-term cognitive functioning in cardiac arrest survivors. Here, we evaluate the prognostic value of tractography-based structural connectivity and EEG-based functional connectivity in the prediction of neurological outcome and cognitive functioning after cardiac arrest. Methods: We used data from two ongoing prospective multicenter cohort studies on comatose patients and survivors after cardiac arrest. MRI data was acquired within 9 days and 1 month for the neurological outcome- and cognitive functioning analysis, respectively. Continuous EEG of comatose patients was recorded during the first 3 days. A 20 minute EEG was made of survivors included in the study on cognitive functioning. Neurological outcome was defined as good (cerebral performance category 1-2) and poor (cerebral performance category 3-5). cognitive functioning was defined using the Montreal cognitive Assessment and a neuropsychological examination. Diffusion MRI data was used to construct a tractography-based structural connectivity matrix. EEG data was source reconstructed to atlas-defined regions and the pairwise temporal irreversibility of the resulting timeseries was estimated to construct a functional connectivity matrix. Graph theory measures were computed of both the structural- and functional matrices. The connectivity measures were compared between patients with good- and poor neurological outcome and logistic regression models were made to assess predictive value. Mixed effect models were used to assess the relationship between connectivity measures and cognitive functioning. Results: We found significant differences in all investigated structural- and functional connectivity measures between patients with good- and poor neurological outcome. Combining the functional clustering coefficient and -modularity to a mean diffusivity measure and clinical features resulted in a significant increase in model performance (p=0.02) compared to a clinical model. Here, the sensitivity of poor outcome prediction at 100% specificity increased from 31% to 73%. There was no significant change in prediction when structural connectivity measures were added to a model based on mean diffusivity. We found no significant relationship between structural- or functional connectivity measures and cognitive functioning. Conclusion: EEG-based functional connectivity measures improve neurological outcome prediction in comatose patients after cardiac arrest. Diffusion MRI-based structural connectivity holds no predictive value in addition to mean diffusivity. No relationship was found between structural- or functional measures and cognitive functioning in survivors after cardiac arrest.