Outcome prediction with resting-state functional connectivity after cardiac arrest.
Journal:
Scientific reports
PMID:
32678212
Abstract
Predicting outcome in comatose patients after successful cardiopulmonary resuscitation is challenging. Our primary aim was to assess the potential contribution of resting-state-functional magnetic resonance imaging (RS-fMRI) in predicting neurological outcome. RS-fMRI was used to evaluate functional and effective connectivity within the default mode network in a cohort of 90 comatose patients and their impact on functional neurological outcome after 3 months. The RS-fMRI processing protocol comprises the evaluation of functional and effective connectivity within the default mode network. Seed-to-voxel and ROI-to-ROI feature analysis was performed as starting point for a supervised machine-learning approach. Classification of the Cerebral Performance Category (CPC) 1-3 (good to acceptable outcome) versus CPC 4-5 (adverse outcome) achieved a positive predictive value of 91.7%, sensitivity of 90.2%, and accuracy of 87.8%. A direct link to the level of consciousness and outcome after 3 months was identified for measures of segregation in the precuneus, in medial and right frontal regions. Thalamic connectivity appeared significantly reduced in patients without conscious response. Decreased within-network connectivity in the default mode network and within cortico-thalamic circuits correlated with clinical outcome after 3 months. Our results indicate a potential role of these markers for decision-making in comatose patients early after cardiac arrest.