A Promising Subject-Level Classification Model for Acute Concussion Based on Cerebrovascular Reactivity Metrics

Concussion imaging research has primarily focused on neuronal disruption with lesser emphasis directed toward vascular dysfunction. However, blood flow metrics may be more sensitive than measures of neuronal integrity. Vascular dysfunction can be assessed by measuring cerebrovascular reactivity (CVR)-the change in cerebral blood flow per unit change in vasodilatory stimulus. CVR metrics, including speed and magnitude of flow responses to a standardized well-controlled vasoactive stimulus, are potentially useful for assessing individual subjects following concussion given that blood flow dysregulation is known to occur with traumatic brain injury. We assessed changes in CVR metrics to a standardized vasodilatory stimulus during the acute phase of concussion. Using a case control design, 20 concussed participants and 20 healthy controls (HCs) underwent CVR assessment measuring blood oxygen-level dependent (BOLD) magnetic resonance imaging using precise changes in end-tidal partial pressure of CO₂ (PETCO₂). Metrics were calculated for the whole brain, gray matter (GM), and white matter (WM) using sex-stratification. A leave-one-out receiver operating characteristic (ROC) analysis classified concussed from HCs based on CVR metrics. CVR magnitude was greater and speed of response faster in concussed participants relative to HCs, with WM showing higher classification accuracy compared with GM. ROC analysis for WM-CVR metrics revealed an area under the curve of 0.94 in males and 0.90 in females for speed and magnitude of response respectively. These greater than normal responses to a vasodilatory stimulus warrant further investigation to compare the predictive ability of CVR metrics against structural injury metrics for diagnosis and prognosis in acute concussion.