ki:elements

Automated speech analysis for fatigue detection in multiple sclerosis

Marcelo Dias, Susett Garthof, Simona Schäfer, Julia Elmers, Nicklas Linz, Tina Boggiano, James Overell, Johannes Tröger, Anja Dillenseger, Christian Beste, Björn Tackenberg, Jan Wesiack & Tjalf Ziemssen

* Poster presented at the ECTRIMS 2024 Congress, Copenhagen (Denmark)

Abstract

INTRODUCTION: Multiple sclerosis (MS) is a chronic autoimmune disease that significantly impacts neurological function, leading to a variety of symptoms, including fatigue. Approximately 75% of MS patients have fatigue (Lerdal et al., 2007), which significantly impacts their daily life activities. Subjective reports of fatigue are the most prevalent among persons with multiple sclerosis (pwMS), and patients often report communicative impairment concurrent with the onset of fatigue (Blaney & Lowe-Strong, 2009), indicating a potential relationship between these two constructs. Despite its prevalence, fatigue is still under addressed due to the lack of objective assessment methods. Automated speech analysis has
emerged as a potential low burden and non-invasive method to detect fatigue in MS patients.

OBJECTIVE/AIMS: This study aims to evaluate the efficacy of automated speech analysis in identifying fatigue among pwMS, thereby contributing to improved disease monitoring and management strategies.

METHODS: We collected a broad range of speech samples and clinical questionnaires, notably the Fatigue Scale for Motor & Cognition (FSMC), from 145 healthy controls and 137 MS patients. Acoustic and linguistic features were extracted and correlated with the FSMC measures.

RESULTS: The analysis of narrative speech shows that speed and voice quality features are associated with overall fatigue (total FSMC) (p<0.05) and that this correlation is driven by cognitive fatigue (FSMC
Cognition) but not motor fatigue (FSMC motor). Additionally, speech features extracted from the Picture Description task were able to accurately distinguish between severely fatigued and non-fatigued participants (p<0.05). Severely fatigued participants show a degradation in voice quality and an overall slowing of speech.

CONCLUSION: Our findings suggest that automated speech analysis offers a promising tool for detecting fatigue in MS patients. This approach could enhance clinical assessment strategies by providing a low burden, non-invasive and objective measure of fatigue, potentially leading to more personalized and effective management of MS symptoms.

Share this article