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

Area of Science:

• Neurology
• Speech Science
• Artificial Intelligence

Background:

• Multiple Sclerosis (MS) fatigue is primarily assessed subjectively due to a lack of objective diagnostic tools.
• Developing objective biomarkers for MS-associated fatigue is crucial for accurate diagnosis and management.

Purpose of the Study:

• To identify vocal biomarkers for differentiating people with MS (pwMS) who experience fatigue from those who do not.
• To explore the potential of AI-based speech analysis in predicting MS fatigue and distinguishing pwMS from healthy controls (HCs).

Main Methods:

• A prospective, observational study (COMMITMENT trial) involving 50 relapsing MS patients and 20 healthy controls.
• Utilized automated artificial intelligence (AI)-based speech analysis to assess acoustic features.
• Employed a Leave-One-Speaker-Out (LOSO) cross-validation strategy for model evaluation.

Main Results:

• Five acoustic features correlated with general fatigue, independent of depression and sleepiness.
• Specific features were associated with motor (five features) and cognitive (12 features) fatigue.
• AI models achieved high specificity (0.68-0.94) in predicting fatigue, though sensitivity varied (0.38-0.90).
• Speech biomarkers differentiated pwMS from HCs with 0.90 specificity and 0.3 sensitivity.

Conclusions:

• Speech analysis, particularly AI-assisted, shows promise as an objective biomarker for MS-associated fatigue.
• Vocal biomarkers may aid in distinguishing pwMS with fatigue from those without, and from healthy individuals.
• AI-driven speech analysis could serve as a valuable complementary tool for existing fatigue assessments in MS.