Machine learning: assessing neurovascular signals in the prefrontal cortex with non-invasive bimodal electro-optical neuroimaging in opiate addiction

Abstract

AbstractChronic and recurrent opiate use injuries brain tissue and cause serious pathophysiological changes in hemodynamic and subsequent inflammatory responses. Prefrontal cortex (PFC) has been implicated in drug addiction. However, the mechanism underlying systems-level neuroadaptations in PFC during abstinence has not been fully characterized. The objective of our study was to determine what neural oscillatory activity contributes to the chronic effect of opiate exposure and whether the activity could be coupled to neurovascular information in the PFC. We employed resting-state functional connectivity to explore alterations in 8 patients with heroin dependency who stayed abstinent (>3 months; HD) compared with 11 control subjects. A non-invasive neuroimaging strategy was applied to combine electrophysiological signals through electroencephalography (EEG) with hemodynamic signals through functional near-infrared spectroscopy (fNIRS). The electrophysiological signals indicate neural synchrony and the oscillatory activity, and the hemodynamic signals indicate blood oxygenation in small vessels in the PFC. A supervised machine learning method was used to obtain associations between EEG and fNIRS modalities to improve precision and localization. HD patients demonstrated desynchronized lower alpha rhythms and decreased connectivity in PFC networks. Asymmetric excitability and cerebrovascular injury were also observed. This pilot study suggests that cerebrovascular injury in PFC may result from chronic opiate intake.