Quantifying Resting-State Functional Connectivity in Critically Brain Injured Patients: A Graph-Theoretical Approach with fNIRS

Poster Session C - Sunday, April 14, 2024, 5:00 – 7:00 pm EDT, Sheraton Hall ABC

Ira Gupta¹ (igupta6@uwo.ca), Matthew Kolisnyk¹, Sergio L. Novi¹, Androu Abdalmalak¹, Loretta Norton², Derek B. Debicki¹, Adrian M. Owen¹; ¹Western University, ²King's College at Western University

Up to 20% of unresponsive patients with critical brain injuries in the ICU, believed to be unconscious, exhibit signs of awareness. However, there remains a need for objective and robust tools that are easily accessible to accurately characterize preserved cortical function and aid in predicting patient outcome. Advanced functional neuroimaging techniques are emerging to improve diagnostic and prognostic precision post-injury and assist in clinical decision-making. This study employs graph-theoretical analyses to quantify resting-state functional connectivity in brain injured patients using functional near-infrared spectroscopy (fNIRS). Mathematically calculated network parameters were used to quantify brain connectivity in patients and healthy controls. Resting-state data was collected for 6 minutes using a full head coverage, 129-channel fNIRS system from a group of unresponsive acutely brain injured ICU patients (N=16) and healthy controls (N=23). A localized comparison revealed significant group differences in clustering coefficient, local efficiency, and degree in several channels including those located in the left mid frontal, left postcentral parietal, and left mid temporal regions. These three network parameters were then integrated into machine learning algorithms to classify patients and healthy controls. Accuracy rates for a linear support vector machine classifier reached up to 75% using clustering coefficient. This study will enhance the understanding of brain connectivity in acute brain injury and ultimately hopes to offer a scientific foundation for clinicians and family members to make well-informed decisions regarding patient care.

Topic Area: METHODS: Neuroimaging