Addressing Cognitive and Behavioral Heterogeneity in Amyotrophic Lateral Sclerosis Through the Identification of Multidimensional Speech Biomarkers Using the VOC-ALS Smartphone Platform

Amyotrophic Lateral Sclerosis (ALS) is increasingly recognized as a multisystem neurodegenerative disorder characterized not only by progressive motor neuron degeneration but also by cognitive and behavioral impairment. Among the clinical manifestations of ALS, dysarthria represents one of the most disabling symptoms and has emerged as a promising source of digital biomarkers for disease monitoring and patient stratification. This thesis explored ALS through a multidimensional framework integrating neuropsychological assessment, structured speech data collection, acoustic analysis, and computational modeling. First, cognitive and behavioral alterations were investigated to improve the characterization of extra-motor involvement across the ALS clinical spectrum. Subsequently, a structured voice signal database was developed to support the systematic study of speech impairment through multiple speech tasks and acoustic dimensions. Building upon this dataset, machine learning and statistical approaches were applied to identify clinically meaningful acoustic markers associated with dysarthria severity. Finally, the relationship between speech production and cognition was explored using integrated acoustic–cognitive models. Acoustic and cognitive variables were organized into physiologically and clinically meaningful domains and summarized through Principal Component Analysis (PCA). The findings suggest that articulatory speech alterations are associated with global cognitive functioning independently of bulbar motor severity, supporting a multidimensional interpretation of speech impairment in ALS. Overall, this thesis highlights the potential of digital speech analysis as a non-invasive and scalable tool for improving the characterization of ALS. By combining cognitive, acoustic, and computational approaches, this work contributes to the development of more interpretable digital biomarkers and provides further insight into the interaction between motor and extra-motor mechanisms in neurodegenerative disease.