Automated speech analysis enables MCI diagnosis

Abstract

Mild Cognitive Impairment (MCI) is a condition characterized by cognitive decline greater than expected for an individual's age and education level. In this study, we are investigating whether acoustic properties of speech production can improve the classification of individuals with MCI from healthy controls augmenting the Mini Mental State Examination, a traditional screening tool, with automatically extracted acoustic information. We found that just one acoustic feature, can improve the AUC score (measuring a trade-off between sensitivity and specificity) from 0.77 to 0.89 in a boosting classification task. These preliminary results suggest that computerized language analysis can improve the accuracy of traditional screening tools.

Open Brain AI: An AI Research Platform

“…Huminfra Conference 2024, Gothenburg, 10-11 January 2024. charalampos.themistokleous@isp.uio.no (C. Themistocleous) In our previous research, we demonstrated that a computational system with four computational pipelines for performing automated acoustic analysis, speech-to-text transcription, automatic morphosyntactic and linguistic analysis of transcripts, and machine learning could enable the identification of Swedish patients with Mild Cognitive Impairment and Alzheimer's Disease from healthy controls (12)(13)(14)(15)(16)(17) and the subtyping of patients with Primary Progressive Aphasia into variants (nonfluent PPA, semantic PPA, and logopenic PPA) (18). The machine learning model of the classification of patients with PPA was based on deep neural networks (DNN), and its performance was better than that of Random Forests, Support Vector Machines, Decision Trees, and expert clinicians' classifications (18).…”