Brain connectivity of the central control of breathing in humans using EEG and fMRI: integration of data and hypothesis driven approaches

This PhD thesis investigates brain networks involved in the physiological and pathological central control of breathing, exploiting respiratory signals, electroencephalography (EEG) and functional Magnetic Resonance Imaging (fMRI). Hypothesis driven and data driven approaches were integrated in different methodological frameworks to perform exploratory analyses of brain connectivity from EEG or fMRI. In particular, each framework addresses specific challenges arising from clinical and biomedical research. A first framework integrating i) Independent Component Analysis, ii) unsupervised clustering, iii) Multivariate Autoregressive modelling and iv) permutation-bootstrap statistics was developed for the group-level analysis of asymmetrical causal interactions from EEG recordings when the clinical setup involves low-density EEG caps, and when experimental protocols deal with resting-state/block design acquisitions. This framework was applied to two case studies. The first study involves patients with impaired breathing and affected by Cheyne-Stokes Respiration (CSR), and is focused on the assessment of the effects of typical CSR breathing pattern on brain connectivity. The second involves healthy subjects performing voluntary breath hold, and aims to investigate the effects at the brain level of the increase of carbon dioxide, as well as to establish potential differences in brain response to hypercapnia between the physiological and pathological case. The developed framework allowed to identify statistically significant differences in connectivity based on ventilatory conditions in both pathological and physiological case. Moreover, such differences occurred mostly in a frequency band associated with hypercapnic stimulation, holding for physiological plausibility of observed differences. A second framework integrating i) Independent Component Analysis, ii) component classification, iii) temporal and spatial correlation analysis and iv) regression analysis was developed for the study of functional brain connectivity on fMRI data from healthy subjects under CO2 stimulation through gas administration. The framework addresses several issues related to the study of brain function related to breathing. On one hand, it provides a pipeline of analysis of fMRI data when the expected time course of brain activation is difficult to be specified a priori due to poor physiological characterisation of delay between gas administration and brain response, on the other it defines an approach for disentangling non-specific effects related to gas stimulation from brain response to CO2. This framework was applied to a population of healthy subjects under hypercapnic-normoxyc stimulation induced by two different tasks: a voluntary breath hold and a carbon dioxide gas administration task. The study is focused on the evaluation of the sensitivity of different groups of chemoreceptors to different values of carbon dioxide. The framework allowed to identify brain regions more sensitive to differences in CO2. The methodological frameworks developed in this thesis raise from clinical and/or research challenges in brain imaging related to breathing. Nonetheless, the purpose of such frameworks is that of giving a methodology for the study of brain interactions in the presence of specific challenges that may arise also in other kind of studies. In this view, this thesis gives a pipeline of analysis of EEG data acquired with low-density caps and during resting-state/block-design that can be applied to other studies not strictly related to breathing. On the other hand, beyond the specific purpose of identifying a dose-dependent relationship between brain function and different levels of CO2, the fMRI pipeline presented in this thesis can be applied for the analysis of other breathing tasks inducing hypercapnia.