Dynamic causal modelling for functional near-infrared spectroscopy using spatial priors derived from diffuse optical tomography

Functional near-infrared spectroscopy (fNIRS) is an optical neuroimaging technique that measures brain activity by detecting changes in oxygenated and deoxygenated hemoglobin concentrations. Although methods exist for estimating causal interactions among brain regions using fNIRS data, current approaches typically rely on node locations defined at the sensor measurement level near the cortical surface. To address this limitation, this study extends dynamic causal modeling (DCM) for fNIRS by incorporating source-level locations estimated from diffuse optical tomography (DOT). The proposed method was applied to an experimental dataset comprising 104 participants recorded during a Go/No-Go response inhibition task. Bayesian model selection confirmed that DCM models using DOT-informed neuronal source locations yielded superior model evidence compared to models using sensor-level locations. Furthermore, the posterior means of effective connectivity parameters validated the inhibitory influence of the right inferior frontal gyrus (rIFG) on regions within the motor network during response inhibition. Due to limited depth sensitivity, subcortical areas were not explicitly modeled; consequently, the reported connectivity represents a net inhibitory influence of the rIFG on the motor areas, which may encompass indirect influences mediated via unmeasured subcortical pathways. By providing depth-dependent localization of source activation informed by statistical parametric mapping (SPM) of DOT-reconstructed data, the proposed DCM approach enables more accurate inference of causal (directed) connectivity at the neuronal level from optical density measurements. Given the compact and portable nature of fNIRS systems, this method can readily be applied in realistic, naturalistic environments to investigate the network-level modulation of brain responses.