Automated and objective analysis of speech in premanifest and early-stage Huntington’s disease

Abstract

AbstractBackgroundClinical markers that show change in performance in people with Huntington’s disease (HD) during the presymptomatic and prodromal stages remain a target of investigation in clinical medicine. Alongside genetic and neuroimaging initiatives, digital speech analytics has shown promise as a sensitive clinical marker of premanifest HD.ObjectiveTo investigate the sensitivity of digital speech measures for detecting subtle cognitive-linguistic and fine motor features in people carrying the expanded HD gene, with and without symptoms.MethodsSpeech data were acquired from 110 participants (55 people with the expanded HD gene including 16 presymptomatic HD; 16 prodromal HD; 14 early-stage HD; 9 mid-stage HD; and 55 matched healthy controls). Objective digital speech measures were derived from speech tasks that fit along a continuum of motor and cognitive complexity. Acoustic features quantified speakers’ articulatory agility, voice quality and speech-timing. Subjects also completed the tests of cognition and upper limb motor function.ResultsSome presymptomatic HD (furthest from disease onset) differed to healthy controls on timing measures derived from the syllable repetition and monologue. Prodromal HD presented with reduced articulatory agility, reduced speech rate and longer and variable pauses. Speech agility correlated with poorer performance on the upper limb motor test.ConclusionTasks with a mix of cognitive and motor demands differentiated prodromal HD from their matched control groups. Motor speech tasks alone did not differentiate groups until participants were relatively closer to disease onset or symptomatic. Data demonstrated how ubiquitous behaviors like speech, when analyzed objectively, provide insight into disease related change.