Automated analysis of natural speech in amyotrophic lateral sclerosis spectrum disorders.

摘要: Objective We implemented automated methods to analyze speech and evaluate the hypothesis that cognitive motor factors impair prosody in partially distinct ways patients with amyotrophic lateral sclerosis (ALS). Methods recruited 213 participants, including 67 ALS (44 ALS, 23 frontotemporal degeneration [FTD]), 33 healthy controls, neurodegenerative reference groups behavioral variant FTD (n = 90) nonfluent/agrammatic primary progressive aphasia 23). Digitized, semistructured samples obtained from picture descriptions were automatically segmented a Speech Activity Detector; continuous segments pitch-tracked; duration measures for silent pause extracted. Acoustic calculated, fundamental frequency (f0) range, mean segment durations, total duration, rate (pause count per minute of speech). Group comparisons related performance on acoustic clinical scales impairments explored MRI cortical thinning ALS-FTD. Results The f0 range was significantly impaired spectrum disorders bulbar disease, regression analyses this thickness cortex perisylvian regions. Impaired degree impairment disorders, regressions bilateral frontal opercula left anterior insula. Conclusion Automated properties dissociate components deficits disorders.