Listening to MS: AI-assisted speech analysis for diagnosis and fatigue prediction (COMMITMENT)

By
Helly Hammer
Monica Gonzalez-Machorro
Pascal Hecker
Uwe Reichel
Alisha Zmutt
Lisa Pedrotti
Andrew Chan
Florian Eyben
Hesam Sagha
Matthias Kahlau
Bert Arnrich
Björn W. Schuller
Robert Hoepner
May 29, 2026
0 min

Frontiers In Digital Health

Objective:

To identify vocal biomarkers to differentiate between people with Multiple Sclerosis (pwMS) with and without fatigue and to predict fatigue levels using AI-based speech analysis.

Key Findings:

Motor fatigue affected 50% and cognitive fatigue 40% of pwMS.
Five acoustic features were associated with general fatigue, five with motor fatigue, and twelve with cognitive fatigue.
The best classification models achieved specificities of 0.68–0.94 and sensitivities of 0.38–0.90.
Speech biomarkers distinguished pwMS from HCs with a specificity of 0.90 but a sensitivity of only 0.3.

Interpretation:

Speech in pwMS may serve as a potential biomarker for MS-associated fatigue and could help differentiate pwMS from healthy controls.

Limitations:

Sensitivity of the models was lower compared to specificity.
The study focused on a specific cohort with low disability levels (EDSS <4.0).

Conclusion:

AI-assisted speech analysis shows potential in identifying fatigue in pwMS and differentiating them from healthy individuals.

Listening to MS: AI-assisted speech analysis for diagnosis and fatigue prediction (COMMITMENT)

Objective:

Key Findings:

Interpretation:

Limitations:

Conclusion:

Original Source(s)

Listening to MS: AI-assisted speech analysis for diagnosis and fatigue prediction (COMMITMENT)

Related Content

Jump assessment on a force plate—an approach to quantify subtle lower limb neuromuscular deficits in people with multiple sclerosis

CSF Proteomics Identifies New Biomarkers for Multiple Sclerosis

The Relationship Between Employment and Self-Reported Health Among Individuals with Multiple Sclerosis in Sweden