Phenotyping Behavioral Disorders in REM Sleep: Study of Digital Biomarkers of Neurodegeneration

Neurodegenerative disorders remain a significant challenge in modern medicine, with early diagnosis and intervention being critical for improving patient outcomes. This thesis focuses on advancing the understanding and detection of REM Sleep Behavior Disorder (RBD), a recognized prodromal symptom of alpha-synucleinopathies, among which Parkinson's Disease (PD). By integrating computational approaches with physiological insights, this research presents novel tools and methods for identifying early-stage biomarkers and understanding autonomic dysfunction in RBD. Heart Rate Variability (HRV) and QT Variability (QTV), studied across different sleep stages, were investigated as potential digital biomarkers of neurodegeneration. Apart from traditional HRV analysis, this thesis also considered the influence of respiration on spectral HRV analysis. A major outcome of HRV studies carried out in this thesis is that patients with idiopathic RBD (iRBD) exhibit distinct HRV profiles during sleep, specifically during N2 and REM phase, characterized by reduced autonomic modulation and decreased complexity compared to healthy controls and PD patients, with or without RBD. iRBD patients showed early sympathetic dominance, likely due to initial neurodegenerative changes in parasympathetic control centers such as the dorsal motor nucleus of the vagus nerve (DMNV), while PD patients exhibited a more balanced autonomic activity due to broader neurodegeneration. These findings highlight the potentialities of HRV analysis during sleep as an early biomarker for neurodegenerative progression in alpha-synucleinopathies. Moreover, respiration guided HRV analysis also revealed the disruption in automatic regulation also in N3 phase which is typically not observed in conventional spectral analysis. The QTV studies also showed that both iRBD and PD patients with RBD exhibited adaptive physiological changes after sleep, potentially reflecting compensatory mechanisms for poor sleep quality. PD-RBD patients demonstrated more pronounced alterations in autonomic regulation, likely due to broader neurodegenerative changes, which were distinct from those seen in iRBD patients. Furthermore, physiological analysis during the REM sleep phase provided insights into autonomic dysfunction in PD and its subtypes, suggesting progressive disruption in regulatory processes. In order to ease the HRV analysis in sleep studies, during my Ph.D. I also developed and released (open-source) the UNICA-HRV tool, a customizable, user-friendly MATLAB user interface designed to analyze HRV metrics across different sleep phases, by exploiting the hypnogram information. Moreover, during my research period in the industry (Micromed), I contributed to the development of signal processing methods for real-time denoising of EEG signals from polysomnographic data. A novel algorithm, the Adaptive Sigmoid Normalized Least Mean Square (AS-NLMS), was developed to address the challenges of artifact removal in dynamic and non-linear environments. The findings presented in this thesis underscore the transformative potential of interdisciplinary research, where engineering and medicine converge to solve complex problems. By introducing new tools, refining analytical methods, and proposing novel approaches to digital phenotyping, this work contributes to the broader effort of early detection and personalized treatment of neurodegenerative diseases.