Amyloid Prediction in Early Stage Alzheimer's Disease Through Speech Phenotyping

The primary objective of the study is to evaluate whether a set of algorithms analysing acoustic and linguistic patterns of speech can detect amyloid-specific cognitive impairment in early stage Alzheimer's disease, as measured by the AUC of the receiver operating characteristic (ROC) curve of the binary classifier distinguishing between amyloid positive (Arms 1 and 3) and amyloid negative (Arms 2 and 4) Arms. Secondary objectives include (1) evaluating whether similar algorithms can detect amyloid-specific cognitive impairment in the cognitively normal (CN) and MCI Arms respectively, as measured on binary classifier performance; (2) whether they can detect MCI, as measured on binary classifier performance (AUC, sensitivity, specificity, Cohen's kappa), and the agreement between the PACC5 composite and the corresponding regression model predicting it in all Arms pooled (Wilcoxon signed-rank test, CIA); (3) evaluating variables that can impact performance of such algorithms of covariates from the speaker (age, gender, education level) and environment (measures of acoustic quality).