Incorporating Regional Brain Connectivity Profiles into the Inference of Exposure-Related Neurobehavioral Burden in Explosive Ordnance Disposal Veterans

Conventional data-driven investigation into novel and complex health outcomes is often impeded by sparse and incomplete data fragmented across platforms and studies. To address this, we propose a knowledge-driven framework for assembling mechanistically informed regulatory networks and simulating their dynamic behavior. We apply the Natural Language Processing (NLP) engine MedScan to the Elsevier ontology and full text corpus (>7.2 million journal articles) to extract documented relationships linking 40 bilateral Brodmann Area (BA) regions, to 9 self-reported neurobehavioral measures of mood, quality of life, resilience and symptom burden. A Constraint Programming problem was defined to determine the direction and mode of action for each network interaction as well as logic parameters describing signal transmission thresholds and decisional weights dictating each node’s state transition. Parameter values were identified such that the predicted behavior of this integrated neurobehavioral regulatory network would jointly explain 1) two distinct neurobehavioral profiles associated with subjective military exposure histories in a pilot cohort (n = 13) of deployed Explosive Ordnance Disposal (EOD) veterans, and 2) EEG regional source activation patterns (theta/alpha spectral power ratio) previously reported during acute mild Traumatic Brain Injury (acute mTBI), chronic mild-moderate Traumatic Brain Injury (chronic mmTBI), as well as control subjects from a publicly available database. Outcomes from the resulting family of competing models unanimously predicted distinct shifts in neural activity in the dorsolateral prefrontal cortex (Brodmann Areas 9 and 46) presenting primarily in the right hemisphere for self-reported exposure to ionizing radiation sources and in the left hemisphere in the case of non-ionizing radiation.