Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Can a 15-word picture description checklist differentiate various subtypes of Primary Progressive Aphasia and Parkinson-plus disorders?
There is a Poster PDF for this presentation, but you must be a current member or registered to attend SNL 2023 to view it. Please go to your Account Home page to register.
Poster C4 in Poster Session C, Wednesday, October 25, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Shalom Henderson1,2, Siddharth Ramanan1, Karalyn Patterson1,2,3, James Rowe1,2,3, Matthew Lambon-Ralph1; 1MRC Cognition and Brain Sciences Unit, University of Cambridge, Cambridge, UK, 2Department of Clinical Neurosciences, University of Cambridge, Cambridge, UK, 3Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK
Picture description tasks are widely used in the assessment of aphasias. Connected speech samples are quick and easy to elicit, but analyses including full transcription are time-consuming, inconsistently conducted, and impractical for non-specialist settings. There is a pressing need for a simple, evidence-based analysis of connected speech that is clinically useful and easy to apply. Using connected speech samples of the Boston Diagnostic Aphasia Examination ‘cookie theft’ and Mini Linguistic Aphasia Examination (MLSE) ‘beach’ pictures from people with primary progressive aphasia (n=27 PPA), progressive supranuclear palsy (n=10 PSP), corticobasal syndrome (n=13 CBS), and healthy controls, we assessed the differences in quantitative language output and the psycholinguistic properties of words produced by patients and controls using a principal component analysis (PCA) to understand and simplify the covariance in the quantifiable (e.g., word count, timing) and lexico-semantic (e.g., frequency, semantic diversity) word properties of connected speech. We then optimised a 15-word checklist for the two pictures and tested the outputs of the least absolute shrinkage and selection operator (LASSO) models with out-of-sample predictive validity testing (26 PPA, 2 PSP, and 6 CBS). As an exploratory analysis, we examined the associations between whole-brain grey matter intensity and PCA-generated principal component scores using t-contrasts with clusters extracted using a threshold of p < 0.001 uncorrected for multiple comparisons. We found that the total language output was significantly reduced in patients with the non-fluent variant of PPA (nfvPPA), PSP, and CBS relative to those with the semantic variant of PPA (svPPA) and controls. Qualitative differences in word properties were found for patients with the logopenic variant of PPA (lvPPA) and svPPA with a disproportionately greater use of words that were more frequent and semantically diverse. Using our 15-word checklists, the LASSO models showed excellent accuracy for within-sample k-fold (over 95%) and out-of-sample validation between patients and controls (over 90%), and moderately good (59% - 70%) differentiation between primary motor (nfvPPA, PSP, CBS) and temporal/lexico-semantic groups (svPPA, lvPPA). When supplementing the LASSO models with the Addenbrooke’s Cognitive Examination – Revised and MLSE sub-scores along with the target words, diagnostic accuracy improved for svPPA versus lvPPA, which highlights the need for further language testing to capture these groups’ primary areas of impairment (e.g., semantic versus phonology/working memory) despite similar overall word usage. Results from our voxel-based morphometry (VBM) analyses revealed significant correlations between grey matter intensities in (i) the bilateral frontal lobe and language output, (ii) left frontal and superior temporal regions and articulatory variety, (iii) bilateral temporal, insula, and the right limbic lobe and phonology, and (iv) bilateral cingulate gyri, right caudate and putamen and age of acquisition. To conclude, we propose that (1) our word checklist is a suitable screening test to identify people with progressive aphasia, and (2) further specialist assessment is likely to be needed for some groups (e.g., svPPA and lvPPA) for accurate diagnoses.
Topic Areas: Language Production, Meaning: Lexical Semantics