Alzheimer's disease (AD) is marked by progressive neuropathological changes that begin decades before cognitive and functional symptoms, and thus efforts have been focused on developing innovative tools and biomarkers for early identification of pre-dementia stages. To date, clinical ability to identify those with pre-dementia stages of AD has been limited and requires expensive (Amyloid PET) or invasive (Lumbar Punctures, LP) testing. However, subtle changes in connected speech may be detectable years before overt disease symptoms present. Our team has developed an approach that uses machine learning and natural language processing combined with advanced acoustic phonetic and lexical-semantic analyses. Preliminary data show promise in identifying AD biomarker status and predicting 2-year cognitive progression. In the proposed study, we leverage our success in collecting CSF biomarkers, neuroimaging and detailed cognitive phenotyping combined with audio-recordings of participants in the Brain Stress, Hypertension and Aging Research Program cohort. This cohort, now in its second year of follow-up, consists of 400 individuals 50 years or older with normal cognition or MCI. We plan to extend this cohort of 400 participants for 3 more years to collect additional waves of voice recordings, cognitive assessments, and follow-up CSF biomarkers and neuroimaging. Our overarching hypothesis is that the derived novel features reflecting poor lexical-semantic connectedness or acoustic perturbations are significantly different between biomarker positive and negative participants, have better diagnostic performance than traditional cognitive tests (e.g. confrontation naming), and are associated with a longitudinal change in cognition and AD-related biomarkers. The Specific Aims are: 1) Determine the accuracy of the derived digital biomarkers in detection of in-vivo AD pathology in the B-SHARP cohort; 2) Investigate longitudinally the association of the derived features with cognitive decline and their ability to reflect changes in AD biomarkers; and 3) using resting state functional MRI, identify the networks in the brain that map to derived lexical semantic and acoustic features with brain connectivity at baseline and during follow-up. This project will provide needed insight into the use of non-invasive digital biomarkers to improve the ability to detect and track longitudinal changes in cognitive and functional status in AD.