Digital speech markers for early detection of Alzheimer’s disease in a longitudinal racially and ethnically diverse cohort

PROJECT SUMMARY/ABSTRACT - modifying treatments become available. Traditional diagnostic methods, primarily episodic memory assessments, are often inadequate in the early stages, missing subtle cognitive shifts that occur years before clinical symptoms are apparent. Language impairments, such as difficulties with word retrieval, emerge as early markers of AD, yet remain underutilized in clinical settings. Automated speech analysis offers a non-invasive, scalable, and cost-effective approach to detecting these early signs by analyzing spontaneous speech patterns. The objective of this project is to develop and validate an AD-specific speech profile using automated speech analysis to improve early detection and monitoring of AD. Leveraging existing and newly collected longitudinal speech samples from culturally and educationally diverse populations, we will use cutting-edge machine learning algorithms to identify key linguistic and acoustic features indicative of cognitive decline. Our large and diagnostically diverse cohorts will enable us to distinguish between different stages of AD and differentiate AD from other neurodegenerative diseases such as primary progressive aphasia and frontotemporal dementia. Aim 1 focuses on developing a sensitive and specific AD-speech profile to classify clinical diagnoses and predict the transition from preclinical to clinical AD. Aim 2 investigates the clinical validity of the AD-speech profile by correlating it with established neuroimaging biomarkers (e.g., amyloid-PET, MRI) and cognitive measures while accounting for sociodemographic influences. Aim 3 aims to establish the reproducibility and generalizability of the speech profile by externally validating it with an independent cohort. This multidisciplinary project brings together expertise in neurology, linguistics, neuropsychology, biostatistics, and digital health to create an accessible, objective tool that enhances early detection and monitoring across racially and ethnically diverse populations. By focusing on subtle changes in language patterns, the project lays the foundation for automated speech analysis as a clinically viable and equitable method for early AD diagnosis and disease monitoring.