Assessing functional connectivity in the newborn brain using fNIRS

Functional connectivity represents a powerful approach to describe the intrinsic activity of the brain. It reveals the organization and correlations among anatomically separated regions supporting similar cognitive and sensory processes. Using functional Magnetic Resonance Imaging (fMRI), the recurrent spatial characteristics of these patterns have been extensively explored in the adult brain and their disruption has been found to be associated with psychiatric and developmental disorders. Unveiling the processes of emergence of resting state networks at a very early stage of life could shed light on the neuronal origins of these diseases. However, the study of the inception and development of functional connectivity in the newborn brain poses exceptional challenges, due to the complexity of dealing with non-compliant subjects. To this end, cortical activity at birth can be investigated using functional Near Infrared Spectroscopy (fNIRS) that represents a promising non-invasive neuroimaging method for developmental studies. In the present thesis, I applied fNIRS to assess functional connectivity in term neonates. The first part of the dissertation is dedicated to investigating the maturation of a specific resting state network, the Default Mode Network, within the first 48 hours of life. The study aimed to examine its emergence, for the first time, using optical imaging on newborns immediately after birth. While the majority of fMRI literature focused on large-scale spatial patterns, I took a different approach measuring an intrinsic and localized fingerprint feature of the network, consistently detected in adult subjects. In the second part of the dissertation, I aimed at improving the anatomical representation of brain connectivity, inferred only from signals collected at the scalp. Thus, I developed and validated a method for the reconstruction of spatially distributed functional signals on a dedicated template for term newborn subjects. The intent is to promote the shift from a sensor space description (one signal for each channel) to a source space representation in which the origin of the signal is reconstructed with better anatomical fidelity. The reliability of the reconstruction method was tested on synthetic and real data. In the former case, I simulated spatially correlated neural activity in the cortex, thus enabling assessment of the reconstructed images against a ground-truth map. Analyses of functional connectivity in both sensor and source space showed that the Default Mode Network is still immature at birth, with a lack of homotopic correlation in the lateral parietal cortices, and no evidence of anticorrelation with the Dorsal Attention Network, a well established feature in the adult brain. Overall the work presented in the thesis contributes to the understanding of functional connectivity in the infant’s brain and provides useful tools for source-based connectivity analysis and for probe design and optimization.