Detecting bulbar amyotrophic lateral sclerosis (ALS) using automatic acoustic analysis.

Area of Science:

• Neurology
• Speech-Language Pathology
• Biomedical Engineering

Background:

• Amyotrophic Lateral Sclerosis (ALS) presents significant challenges in clinical practice and trials.
• Acoustic speech analysis shows promise for detecting speech motor impairments in ALS.
• Current methods face limitations due to lab-based assessments and analysis opacity, hindering disease staging.

Purpose of the Study:

• To evaluate an automated speech assessment app for detecting ALS.
• To determine if acoustic features can differentiate between ALS disease stages and severity.
• To explore the potential of remote screening and monitoring for ALS patients.

Main Methods:

• Collected speech samples from 119 ALS patients and 22 healthy controls.
• Analyzed 53 acoustic features from readings of a standardized passage.
• Stratified ALS patients into early (ALS-E) and late (ALS-L) stages using the ALS Functional Ratings Scale-Revised bulbar score.
• Employed sparse Bayesian logistic regression for data analysis.

Main Results:

• The automated acoustic features distinguished ALS patients from controls with an AUROC of 0.85.
• Early-stage ALS (ALS-E) patients were well separated from controls (AUROC = 0.78).
• Reasonable separation was achieved between early (ALS-E) and late (ALS-L) stage patients (AUROC = 0.70).

Conclusions:

• Automated acoustic analysis holds significant potential for early ALS detection.
• This technology can aid in stratifying ALS patients for clinical trials and practice.
• Remote, app-based speech assessments offer a feasible approach for ALS monitoring.