To study acoustic speech signals in ALS and identify specific acoustic-prosodic markers of motor and cognitive impairments.

In amyotrophic lateral sclerosis (ALS), motor and cognitive deficits may disrupt speech. We implement automated speech recognition methods to characterize acoustic-prosodic properties of speech in patients with ALS spectrum disorders.

We recruited 44 patients with motor ALS, 23 with ALS-frontotemporal degeneration (ALS-FTD) as well as normal and FTD (non-ALS) controls. Speech samples of picture descriptions were automatically segmented with a speech activity detector. Continuous speech segments were pitch-tracked and duration measures for speech and silent pause segments were extracted. We calculated acoustic measures, including fundamental frequency (f0) range, mean speech and total speech durations, and pause rate. We compared groups and related performance on acoustic measures to clinical tests of cognitive and motor functions and cortical atrophy in MRI-T1 scans of ALS.

f0 range was impaired and related to bulbar disease (beta=-0.59, p=0.012) in ALS. f0 range was associated with MRI atrophy in primary motor cortex and left peri-Sylvian regions. Shortened speech segments (beta=0.01, p=0.02) and total speech duration (beta=0.38, p=0.006) were related to cognitive impairment. Total speech time was associated with atrophy in the frontal opercula and IFG bilaterally, the left anterior insula and the superior temporal gyri (STG). Only timed measures of speech, including pause rate and shortened speech segments, were related to respiratory restriction and associated with atrophy in the right frontal operculum and left STG.

Speech samples in ALS spectrum disorders can provide highly objective and reproducible markers of disease derived purely from the acoustic signal. Such acoustic markers relate to prosodic elements of natural language such as fluency and intonation and reflect specific motor and cognitive impairments in ALS.