Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams

Automated markers of bulbar motor disease in ALS: Vowel articulation of natural speech

Poster C57 in Poster Session C, Friday, October 7, 10:15 am - 12:00 pm EDT, Millennium Hall

Sanjana Shellikeri¹, Sharon Ash¹, Mark Liberman², Sunghye Cho², Corey McMillan¹, Lauren Elman³, David Irwin¹, Murray Grossman¹, Naomi Nevler¹; ¹Penn Frontotemporal Degeneration Center, University of Pennsylvania, ²Linguistic Data Consortium, University of Pennsylvania, ³Penn Comprehensive ALS Center, University of Pennsylvania

Introduction: Bulbar motor impairment affects over 92% of people with amyotrophic lateral sclerosis (ALS), leading to a loss of speech and swallowing, with detrimental effects on survival and quality-of-life. Digital speech analytics show promise as automated, quantitative markers of bulbar disease. However, existing platforms analyze structured “pseudo-speech” tasks which lack face validity and therefore give limited insight into clinical outcomes such as intelligibility and swallowing difficulties, and may be confounded by cognitive impairment. This study tests a novel, automated analysis tool of natural speech that focuses on vowel acoustics which have direct lingual (tongue) articulatory underpinnings. We hypothesized that measures related to tongue articulation will be sensitive and specific to bulbar motor impairment in ALS, compared with neurodegenerative controls with behavioral variant frontotemporal dementia (bvFTD) and healthy controls (HC). Methods: We analyzed picture descriptions provided by 83 ALS (10 with ALS-FTD), 34 bvFTD, and 80 healthy controls (HC). Fifty-three ALS speakers had bulbar disease (ALS+bulbar, vs. n=28 ALS-bulbar), based on bulbar scales of ALS Functional Rating Scale-revised and Penn UMN score. Seventeen ALS had longitudinal data (n=9 ALS+bulbar at baseline). The automatic vowel analysis involved tagging vowels using a pronunciation dictionary, extracting their temporal (durational) and spectral acoustic properties (i.e., first and second formants, or F1 and F2, reflective of tongue position in high-low and front-back directions in the mouth), and calculating articulatory-acoustic measures: (1) vowel space area using the F1/F2 coordinates of corner vowels, (2) high-low vowel-pair distance (proxy of tongue elevation magnitude), (3) ΔF2/time during consonant-to-vowel transitions (i.e., F2 slope, measure of articulatory speed), and (4) average vowel duration. We compared ALS, bvFTD, vs. HC; and ALS+bulbar, ALS-bulbar vs. HC, covarying for demographic differences using ANCOVAs, and assessed discriminatory ability using logistic regression and receiver operating characteristic curve analysis. Linear regressions tested the associations of vowel features with bulbar motor and cognitive scores (Edinburgh Cognitive Assessment Scale). Linear mixed modelling assessed within-individual change over time and interactions with bulbar motor group at baseline, covarying for cognitive scores. We explored relations to MRI grey matter volume (GMv) and white matter diffusivity of speech network regions. Results: F2 slope was reduced, and vowel duration was longer in ALS vs. bvFTD (p = .003, p = .046) and HC (p = .013, p < .001), and correlated with bulbar motor scores (F2 slope: R = .46, p =.002; duration: R = -.37, p =.008). ALS+bulbar were impaired across all 4 measures compared to ALS-bulbar and HC (each p < .032). Vowel measures together classified ALS+bulbar vs. ALS-bulbar with area under the curve=.86. Longitudinally, ALS+bulbar significantly decreased F2 slope and increased vowel durations (both p<.001), but not ALS-bulbar. None of the measures differed between bvFTD and HC, nor were associated with cognitive impairment or demographics. Shallower F2 slope was related to reduced ventral precentral gyrus GMv (R=.36, p= .05) and corticospinal tract anisotropy (R=.30, p=.04). Summary: Automated vowel acoustic features of natural speech are sensitive, independent markers of bulbar motor disease in ALS spectrum disorders.

Topic Areas: Speech Motor Control, Disorders: Acquired