Predicting the risk of cognitive decline and speech impairment in the preclinical stage of Alzheimer's disease using structural MRI

Abstract Background Speech abnormalities are increasingly recognized as a manifestation of cognitive deficits in Alzheimer's disease (AD) and its preclinical and prodromal stages. Here, we investigated whether MRI measures of brain atrophy, specifically in the basal forebrain and cortical language areas, can predict cognitive decline and speech difficulties in older adults within the AD spectrum. Method The ongoing Prospect‐AD study aims to develop an algorithm to automatically identify speech biomarkers in individuals with early signs of AD. We recruited participants for the Prospect‐AD study from the German national cohort studies DELCODE and DESCRIBE. Within Prospect‐AD, we captured speech recordings via automated phone calls in healthy controls and people with preclinical and prodromal AD. Speech features were extracted from the semantic verbal fluency (SVF) task of automated telephone conversations, using artificial intelligence‐based analysis of the audio recordings. We combined the speech data from Prospect‐AD with the clinical and imaging data from DELCODE and DESCRIBE, including socio‐demographic information, baseline AD diagnosis, neuropsychology, CSF amyloid status, and MRI scans. We used mixed‐effects regression to estimate the association of rates of cognitive decline and traits of automatically obtained speech features with regional atrophy. Result To date, we enrolled 220 participants in the Prospect‐AD study, 103 of whom are included in this report. Our participants underwent baseline clinical examinations and MRI scans between 2014 and 2022 (mean age = 68.1, SD=5.7 years, 67% female, mean MMSE score = 29.2, SD=1.3, CDR global score ≤0.5) with annual follow‐up over 4.9 years (SD=2.5 years) and successfully completed the baseline phone assessment in 2022‐2023. The results will be presented at the conference. Conclusion Automated speech‐based cognitive testing may assist in low‐threshold identification of older adults at higher risk for clinical progression to AD which in turn may assist in allocation of clinical resources. In future analyses, we will further investigate associations of cognitive performance and speech features with structural or functional connectivity changes in the brain using DTI or functional MRI data.