Methodologies of speech analysis for neurodegenerative diseases evaluation

Area of Science:

• Neurology
• Speech Pathology
• Biomedical Engineering

Background:

• Neurodegenerative diseases progressively damage nerve cells, often causing speech impairment.
• Vocal signal analysis is a key method for assessing speech disorders like dysarthria.
• Distinguishing between healthy and pathological voices aids in diagnosing neurological conditions.

Purpose of the Study:

• To propose and evaluate two vocal signal analysis methods for dysarthria detection.
• To compare speech patterns in healthy subjects, Parkinson's Disease (PD) patients, and Multiple Sclerosis (MS) patients.
• To identify acoustic and vowel features indicative of neurological disease-related speech impairment.

Main Methods:

• Collected voice signals from 153 participants: 39 healthy (HS), 60 PD, and 54 MS.
• Extracted acoustic features (F0, jitter, shimmer, NHR) and vowel metrics (tVSA, qVSA, FCR).
• Applied statistical tests to analyze feature differences between groups.

Main Results:

• Significant differences were found in most extracted vocal features between pathological and healthy voices.
• Features like F0, jitter, shimmer, NHR, tVSA, and FCR proved statistically significant.
• These features can serve as diagnostic indicators for dysarthria in PD and MS.

Conclusions:

• The proposed vocal analysis methods are efficient for characterizing speech differences in neurological disorders.
• These methods can support physicians in evaluating disease progression and diagnosing comorbidities.
• The study aligns with precision medicine principles for personalized diagnosis and treatment.