Dissociation Index (DI) for the evaluation of REM Sleep Behaviour Disorder in ALS patients

The aim of this thesis work is to develop a machine-learning algorithm able to recognize and predict the progression of amyotrophic lateral sclerosis (ALS), which is commonly known as Lou Gehrig's disease. There is evidence that rapid eye movement (REM) sleep behaviour disorder (RBD) may serve as a precursor or early indicator of ALS, owing to the shared motor system dysfunction between the two conditions. Genetic or environmental factors might also contribute to the development of both diseases. Hence, this thesis examines several studies about RBD and REM sleep without atonia (RSWA), as they have been associated with various neurodegenerative pathologies, primarily those in the alpha-synucleinopathies group, such as Parkinson's disease and Dementia with Lewy bodies. For this purpose, comprehensive datasets on ALS and RBD patients' sleep patterns, motor abilities, and health records have been collected and analyzed. The study could help develop a precise prediction system for ALS using non-invasive measures such as polysomnography, which is a sleep study that records brain waves, eye movements, and other indicators that can reveal the state of a person's nervous system. Unfortunately, acquiring and then analysing polysomnographies has some drawbacks: patients need to spend at least one night in the hospital and this may affect the quality of their sleeping and neurophysiologists manually score eight hours (or even more) records of sleeping for each patient. This requires the physiologists to be extremely accurate and expert in the field of sleep scoring. Therefore, a machine learning approach could improve and facilitate the task, thus enhancing its speed and precision. As a preliminary step, manual evaluation of the RBD and its state of the art were analyzed, focusing on how the RAI, SINBAR and Montreal indices are usually computed and estimated. In a second step, those results were compared to those obtained by the Dissociation Index showing the goodness of such an algorithm. Data were acquired from a set of patients of the Regional Centre for Sleep Medicine at the Molinette Hospital in Turin in three different stages: at the first check-up and later on, at the 6- and 12-month follow-ups. Since ALS is a rapid-course disease with a very low life expectancy, some patients in the final stage were not able to attend follow-ups and unfortunately, others had already passed away. Research shows that RBD symptoms may appear several years in advance with respect to those affected by ALS; in this scenario, such a prediction model can significantly enhance the ALS diagnosis and treatment, potentially paving the way for novel therapies and improving patients' quality of life.