Machine Learning algorithms for assessing REM Sleep Without Atonia in patients with Parkinson’s Disease and Narcolepsy

REM Sleep Behavior Disorder (RBD) is a sleep disorder characterized by harmful motor behaviors occurring in conjunction with vivid dream experiences. Recent research has increasingly recognized RBD as an early manifestation of neurodegenerative conditions such as Parkinson’s Disease, Multiple System Atrophy, and Dementia with Lewy bodies. Diagnosis traditionally relies on clinical history, polysomnography, and manual-visual scoring methods, which are time-consuming and require specialized expertise. This thesis explores the application of automated algorithms, particularly the REM Sleep Atonia Index (RAI), and Dissociation Index (DI) to identify and evaluate RBD. RBD is characterized by a loss of muscle atonia during REM sleep, known as REM Sleep Without Atonia (RSWA). The development of a continuous Dissociation Index (DI) has introduced a quantitative measure of RBD severity, offering potential benefits in monitoring and treatment. This study aims to assess the effectiveness of the DI, particularly in the context of RSWA observed in Parkinson’s Disease and narcoleptic patients. The link between RBD and Parkinson’s Disease is strong, with RBD often preceding the clinical onset of neurodegeneration. Early identification of RSWA could serve as an invaluable marker for early intervention and treatment. Lifestyle modifications and emerging neuroprotective drugs have demonstrated potential in slowing disease progression, highlighting the need for early markers of neurodegeneration. Furthermore, this research addresses the relationship between RBD and Narcolepsy Type 1 (NT1), specifically examining the phenomenon of Narcoleptic REM Sleep Behavior Disorder (N-RBD) in individuals with narcolepsy. Recent studies indicate that RBD may affect a significant proportion of NT1 patients, emphasizing the importance of efficient and practical diagnostic methods. While automated RSWA assessment methods show promise, they require high-quality polysomnography recordings and further validation, especially in larger patient groups. Machine learning techniques have been instrumental in classifying RBD patients from healthy controls, and this study assesses a method for the blind identification of RSWA by analyzing spectral patterns of electromyogram (EMG) data during REM sleep. The Dissociation Index, measuring the degree of dissociation between mind and body during RSWA episodes, holds potential as a tool for enhancing the evaluation and long-term monitoring of RBD. However, its robustness and reliability need comprehensive validation. This research seeks to advance our understanding of RBD and its association with neurodegenerative diseases and narcolepsy, ultimately contributing to the development of more effective diagnostic and monitoring tools for these conditions.