Slide Slam

< Slide Slam Sessions

Slide Slam A19

Longitudinal speech markers of motor and cognitive disease in ALS-FTD spectrum

Slide Slam Session A, Tuesday, October 5, 2021, 12:30 - 3:00 pm PDT Log In to set Timezone

Sanjana Shellikeri1, Sunghye Cho2, Mark Liberman2, Corey McMillan1, Lauren Elman1, Sharon Ash1, Murray Grossman1, Naomi Nevler1; 1Penn Frontotemporal Degeneration Center, University of Pennsylvania, 2Linguistic Data Consortium, University of Pennsylvania

Introduction: Speech is controlled by motor processes associated with the planning and execution of articulators in producing utterances and cognitive processes of selecting and arranging appropriate vocabularies to convey a message. Clinical markers that are quantitative and domain-specific to individual-level changes are crucial for delivering targeting clinical care and to track therapeutic trial outcomes in conditions with mixed motor and cognitive disorders such as amyotrophic lateral sclerosis–frontotemporal dementia (ALS-FTD). Automated digitized speech measures can serve as inexpensive, non-invasive, and specific markers of disease. In this study, we analyzed longitudinal changes in speech acoustic and lexical-semantic measures in ALS-FTD using an objective and automated method. We hypothesized that cognitive and motor factors would show partially differential longitudinal changes in patients with ALS-FTD disorders. Methods: We analyzed digitized speech samples of Cookie Theft picture descriptions longitudinally from n=23 ALS-FTD patients over a course of 2-5 years (77% ALS, 23% ALS-FTD). The automatic speech analysis involved: (1) segmenting the acoustic signal into speech and silent pauses and pitch-tracking to extract duration measures and f0 range as properties of speech timing and prosody; (2) forced-aligning the transcript to the acoustic signal, tagging vowels based on established word pronunciation, and extracting vowel formants to derive measures of articulatory-acoustic working space and vowel-consonant transition speeds; and (3) tagging part-of-speech (POS) categories of tokenized words, and rating lexical-semantic characteristics to establish POS usage and lexical diversity. We examined within-individual changes and relations with clinical scales of cognitive (ECAS ALS Specific), bulbar motor (Penn UMN bulbar scores), and respiratory impairments (%FVC), and explored relations to MRI cortical thinning. Results: Articulatory rate (syllables/sec) (p = 0.037) and second formant transition slopes (p = 0.041) significantly declined over time in all patients, indicating neuromuscular slowing of tongue articulatory movements. f0 range interacted with bulbar scores covarying for cognitive scores (p = 0.021) showing that f0 range declined over time only in patients with bulbar disease after controlling for cognitive severity. Lexical-semantic measures interacted with cognitive scores after controlling for bulbar motor severity: reductions in average age of acquisition (p = 0.004), word ambiguity (p = 0.050), and word length (in number of phonemes) (p = 0.020), and an increase in word concreteness (p = 0.012) was observed only for patients with ALS-FTD. Summary: With the implementation of automated speech analysis methods, natural speech can provide independent markers of motor and cognitive disease in ALS-FTD. Our current report demonstrates the value of these digital speech markers in longitudinal patient follow-up.

< Slide Slam Sessions

SNL Account Login

Forgot Password?
Create an Account

News