Biomarkers.

Area of Science:

• Neurology
• Computational Linguistics
• Gerontology

Background:

• Advances in language technology offer potential for early Alzheimer's disease (AD) detection and monitoring.
• Clinical interpretability of automatic speech and language features needs assessment against patient outcomes.
• Research must expand beyond English to ensure equitable access to language-based health technologies.

Purpose of the Study:

• To investigate the clinical interpretability of automatically extracted speech and language features.
• To relate these features to patient-centered outcomes like speech fluency and word-finding difficulties.
• To assess automatic speech and language analysis in Icelandic speakers for Alzheimer's disease research.

Main Methods:

• 48 native Icelandic speakers were grouped into four cohorts: cognitively healthy (amyloid negative/positive), mild cognitive impairment, and mild Alzheimer's disease.
• Participants completed oral picture description tasks recorded over three sessions.
• Acoustic and lexical features were automatically extracted, alongside expert ratings by two speech-language pathologists.

Main Results:

• Analysis is ongoing, focusing on inter-rater reliability between speech-language pathologists.
• Correlations between automated features and expert ratings will be presented.
• Cohort classification accuracy using both methods will be compared.

Conclusions:

• Automatic speech and language analytics offer a promising alternative to subjective and burdensome clinical trial endpoints for Alzheimer's disease.
• These objective, non-invasive assessments target core AD clinical features.
• This study advances the development of language-based health technology for global clinical trials.