Advances in Signal Processing Methods to Investigate Functional Brain-Heart Interplay

In this doctoral dissertation, advances in the study of functional Brain-Heart Interplay (BHI) are reported. While the crucial role of the interaction between Central and Autonomous Nervous Systems has been highlighted in several clinical and physiological studies by investigating their anatomical, biochemical, and functional links, a few methodological endeavors reported on a quantitative description of such a fundamental interplay. Here, I report on a novel, fully parametric methodological framework for the directional quantification of functional BHI using non-invasive brain and heartbeat monitoring, and experimental results are gathered in different physiological and pathological conditions. The dissertation is organized as follows. An overview of fundamental BHI physiology is in the first Chapter, with evidences and significance from clinical studies. The second Chapter reports on the state of the art signal processing techniques that have been employed in modelling brain and heartbeat dynamics, which are mainly recorded through the electroencephalogram, EEG, and heart rate variability, HRV, respectively. Chapter three reports on brain- and heartbeat-related features and coupling metrics that may be critical for the definition of comprehensive functional BHI estimation techniques. Following a thorough description of the current state of the art, the proposed computational framework based on generative synthetic data modelling of the joint brain and heart activity is detailed. From the model backward formulation, novel, directional, ad-hoc BHI biomarkers that are defined for EEG and heartbeat dynamics can be derived. Further directional, probabilistic BHI methods based on inhomogeneous point processes, as well as novel BHI quantifiers based on multifractal spectral analysis are described as well. A collection of experimental results using the aforementioned methodology is reported in the fourth Chapter, which is divided in results coming from the application of model-based approaches, and model-free ones. Experimental datasets used in this dissertation, which includes concurrent EEG and ECG series recordings, include: • sympathovagal changes through a Cold Pressor Test (CPT); • emotional elicitation through images and videos; • different type of movement performances; • dreaming experience during REM sleep; • resting state of subclinical depressive subjects. Regarding the model-based applications proposed in this thesis, the computational generative model of the joint brain and heart activity is first validated through synthetic data generation and the analysis of related time series. The family of new biomarkes derived from the designed model is then computed in several cases: a comparison between CPT and resting state; an analysis on valence and arousal differences induced through emotional video elicitations; a study on BHI correlates of a dream recall after REM sleep; a BHI investigation on dysphoric subjects; BHI metrics exploited to discriminate classes of upper limb movements involving different interaction with objects. Furthermore, the exploitation of model-free techniques based on multifractal (MF) and nongaussian (NG) features is proposed to quantify the functional BHI in a nonlinear fashion. The last Chapter summarises and discuss overall the experimental results to highlight common findings and milestone neuroscientific point of interests, with a critical analysis of the advantages and limitations of the proposed framework that lead to future developments of this research.