Automated Analysis of Written Language in the Three Variants of Primary Progressive Aphasia. (20th December 2022)

• Record Type: Journal Article
• Authors: Josephy‐Hernandez, Sylvia E.; Rezaii, Neguine; Jones, Amelia; Hochberg, Daisy; Quimby, Megan; Wong, Bonnie; Dickerson, Bradford C.
• Is Part Of: Alzheimer's & dementia. Volume 18(2022)Supplement 2
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-12-20
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.067739
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Despite the important role of written language in everyday life, abnormalities in functional written communication have been sparsely investigated in PPA. Method: We developed a Python program which uses a language parser and quantifies content units (CU) and total utterances in language samples. The program was used to analyze written descriptions of the WAB Picnic scene, based on a pre‐defined CU corpus, from a cohort with 95 participants (20 controls, 28 nonfluent variant Primary Progressive Aphasia (nfvPPA), 30 logopenic variant (lvPPA), 17 semantic variant (svPPA)). We then compared written functional communication between the three PPA variants and controls. We analyzed the measurements in relation to aphasia severity (Progressive Aphasia Severity Scale Sum of Boxes (PASS SoB) and Clinical Dementia Rating Supplemental Language (CDR‐L) scores), and dementia severity (Clinical Dementia Rating Sum of Boxes (CDR SoB)). Result: The reliability of the automated analysis was high as compared to manual scoring of 239 samples (99.6%). One‐way ANOVAs revealed patients with lvPPA and svPPA wrote fewer CU than controls (mean+/‐SD 11.97+/‐7.87 and 10+/‐4.5, vs 18.5+/‐9.32) (F(3, 91)=[4.897], p=0.003). There was a trend towards a decrease in CU in nfvPPA (13.89+/‐6.33) (p=0.144). Patients with all PPA subtypes (nfvPPA mean+/‐SD 34.64+/‐18.89; lvPPA 38+/‐24.54; svPPA 28.88+/‐20.36) wrote fewer total utterances than controls (61.7+/‐27.86) (F(3, 91)=[7.778], p<0.001). The CU/utterance ratio was higher in nfvPPA and svPPA than in controls and lvPPA (0.42+/‐0.14 and 0.43+/0.22 vs 0.3+/‐0.05 and 0.32+/‐0.12 respectively) (F(3, 91)=[5.592], p=0.001). Considering all PPA samples, regression analysis showed a decrease in CU relative to an increase in PASS SoB (R=0.24, p=0.04), CDR‐L (R=0.22, p=0.06), and CDR SoB (R=0.34, p=0.004). Conclusion: As expected, we found impairments in informative written language in patients with PPA. Automated CU analysis is a tool that can improve assessment of functional written communication, and may be an objective tool to track aphasia progression. More studies are required to assess individual changes over time and whether these findings could be extrapolated to other forms of cognitive impairment.