# Predictive value of resting-state fMRI graph measures in hypoxic encephalopathy after cardiac arrest

## TL;DR

This study examines how resting-state fMRI can predict brain function recovery after cardiac arrest, finding that it correlates with EEG patterns but offers limited additional predictive value.

## Contribution

The study evaluates the added value of graph-theoretical measures from fMRI in predicting outcomes after cardiac arrest, finding them largely redundant.

## Key findings

- Whole-brain functional connectivity and clustering coefficient were significantly lower in patients with poor outcomes.
- Functional connectivity in posterior brain areas most strongly correlated with neurological outcomes.
- Graph measures did not provide additional predictive value beyond clinical and EEG-based models.

## Abstract

•fMRI-based whole brain functional connectivity is related to EEG patterns in postanoxic coma.•Whole brain functional connectivity discriminates between good and poor outcome, but additional prognostic value is small.•Extraction of graph measures appears redundant.

fMRI-based whole brain functional connectivity is related to EEG patterns in postanoxic coma.

Whole brain functional connectivity discriminates between good and poor outcome, but additional prognostic value is small.

Extraction of graph measures appears redundant.

Current multimodal prediction models can determine the prognosis of about half of comatose cardiac arrest patients. We investigated whether whole-brain graph-theoretical measures from early resting-state functional magnetic resonance imaging (fMRI) three days after cardiac arrest discriminate between good and poor outcome and improve outcome prediction.

We conducted a prospective cohort study on comatose cardiac arrest patients on intensive care units. Resting-state fMRI three days after cardiac arrest was used to quantify whole-brain functional connectivity, global efficiency, clustering coefficient, and modularity. Neurological outcome at six months was classified as good or poor (Cerebral Performance Category 1–2 vs 3–5). Logistic regression models were used to examine between-group differences and study the additional value of graph-theoretical measures to clinical and EEG-based prediction.

In seventy included patients (good outcome n = 44, poor n = 26), whole-brain functional connectivity and clustering coefficient (but not global efficiency and modularity) were significantly lower in patients with poor outcome. Connectivity of nodes in posterior brain areas most prominently correlated with outcome. Clustering coefficient showed strong correlation with whole-brain functional connectivity. Patients with continuous EEG patterns differed in whole-brain functional connectivity levels from those with suppressed or epileptiform patterns. Combining functional connectivity or graph measures with clinical and EEG-based predictors slightly improved outcome prediction.

fMRI-based whole-brain functional connectivity is a sensitive measure for encephalopathy severity after cardiac arrest, according to relations with established EEG categories and discrimination between good and poor outcome. Additional predictive values for outcome seem small. Graph measures do not provide complementary information.