Research Spotlight: Digital Speech Markers Can Be Used to Classify Patients with Functional Speech Disorders

Jennifer Freeburn, MS, of the Departments of Speech, Language and Swallowing Disorders and Neurology at Massachusetts General Hospital (MGH), is the lead author of a paper published in Movement Disorders, “Using Digital Speech Markers to Classify Functional Speech Disorder: A Proof-of-Concept Pilot Study”. David Perez, MD, MMSc, chief of the Division of Behavioral Neurology at Mass General Brigham and director of the Functional Neurological Disorder Unit at MGH, and Neguine Rezaii, MD, of the MGH Department of Neurology, are co-senior authors of the paper.

Q: How would you summarize your study for a lay audience?

Functional neurological disorder (FND) is a commonly encountered, potentially disabling, and costly condition at the intersection of neurology and psychiatry, affecting about 80 per 100,000 people. Functional speech disorder (FND-speech) is a variant of FND, whereby adults experience new onset of stuttering, accented speech, or articulation difficulties, among other presentations. Patients diagnosed with functional speech difficulties often experience long diagnostic delays and limited access to evidence-based treatments.

Our paper extracts digital speech features from both patients with FND-speech and healthy controls who were recorded while describing a standard image. Using machine learning, we identified certain acoustic, rate-based, and linguistic features that could be digitally extracted from the recordings to robustly classify the group with FND-speech versus the healthy controls.

Our study demonstrates the proof-of-concept utility of digital speech markers to potentially serve as adjunctive markers, leading patients to be diagnosed earlier.

Q: What question were you investigating?

We investigated whether digital speech markers from audio recordings could distinguish patients with FND-speech compared to healthy controls using machine learning. 

Q: What methods or approach did you use?

Participants were asked to describe a standard image, and we recorded their speech. Speech features, such as an index of irregularities in voice intensity, were extracted from the audio recording based on previously established approaches refined by Dr. Rezaii. Using machine learning, we investigated the robust features that could be used to predictively classify individuals with FND-speech. 

Q: What did you find?

We found that acoustic features, such as local shimmer (changes in voice amplitude over a period of time), and formants (frequencies critical for distinguishing vowel sounds) were digital speech features that showed good ability to classify patients with FND-speech. These features underscore prominent motor disruption in FND-speech, resembling those in other FND, such as functional tremor.

Rate-based and language-based features also distinguished patients with FND-speech. The observation that language-based features could be used to classify FND-speech underscores that this condition involves functional disruptions in both the motoric and cognitive aspects of speech production.

Q: What are the next steps?

Out-of-sample prospective replication and larger-scale studies are needed as next steps. Our multidisciplinary team will apply for external funding to expand upon this fruitful research direction. Future studies could include looking at subjects with structural brain etiologies of dysarthria (a motor speech disorder) or aphasia (a communication disorder that impairs expressive/receptive language) in comparison to those with FND-Speech.

Authorship: In addition to Freeburn, Perez, and Rezaii, Mass General Brigham authors include Sara A. Finkelstein [co-first author] and Christiana Westlin.

Paper cited: Freeburn, J., et al., “Using Digital Speech Markers to Classify Functional Speech Disorder: A Proof-of-Concept Pilot Study.” Movement Disorders. DOI: 10.1002/mds.30256

Funding: This work was supported by the Mass General Neuroscience Transformative Scholar Award (GR0246510).

Disclosures: Perez has received honoraria for continuing medical education lectures and a textbook on functional neurological disorder; has received royalties from Springer Nature for a textbook on functional movement disorder; is on the editorial boards of Journal of Neuropsychiatry & Clinical Neurosciences (paid), Brain and Behavior (paid), Epilepsy & Behavior, and Cognitive and Behavioral Neurology; is a board member of the Functional Neurological Disorder Society and council member of the American Neuropsychiatric Association; and has received funding from the National Institutes of Health and Sidney R. Baer Jr. Foundation unrelated to this work.