Disease Delineation for Multiple Sclerosis, Friedreich Ataxia, and Healthy Controls Using Supervised Machine Learning...

Area of Science:

• Neurology
• Speech Science
• Computational Linguistics

Background:

• Neurodegenerative diseases frequently impact speech production.
• Acoustic speech analysis offers objective markers for pathological conditions.
• Prior studies often excluded comorbidities, limiting the scope of speech markers.

Purpose of the Study:

• To investigate the utility of multiple acoustic speech features in delineating neurodegenerative diseases.
• To differentiate between healthy individuals, those with multiple sclerosis, and those with Friedreich ataxia using speech patterns.
• To identify specific acoustic features indicative of neurodegenerative pathology.

Main Methods:

• Utilized supervised machine learning (CatBoost) for classification.
• Analyzed 74 spectral and temporal prosodic features from diadochokinetic tasks.
• Compared speech recordings from healthy controls, multiple sclerosis patients, and Friedreich ataxia patients.

Main Results:

• Achieved high accuracy (over 82%) in identifying Friedreich ataxia, multiple sclerosis, and healthy controls.
• Identified 21 acoustic features (spectral qualia, spectral power, speech rate) as strong disease markers.
• Demonstrated the capability of speech markers to accurately distinguish between healthy and pathological speech.

Conclusions:

• Speech acoustic markers can effectively delineate neurodegenerative diseases and distinguish them from healthy speech.
• Speech analysis is a valuable tool for assessing indicators of neurodegenerative disease.
• Further research with large-scale initiatives can expand speech marker applications to other neurological disorders and affective conditions.