|
Abstract:
|
Existing methods in univariate longitudinal data or multivariate cross-sectional data are limited in the power of analyzing multi-domain outcome association at a single time point and modeling a disease progression at the individual level. A generalized joint mixed-effect model is developed in this paper to exploit multidomain (i.e. Neuropsychological assessments, Functional and behavioral assessments, Neuroimaging measures, and biomarkers concentration) longitudinal data and develop disease progression estimates for each individual. In addition, a novel horizontal modeling approach is applied to make inferences on the disease latent timing for individuals’ specific trajectories besides the vertical modeling of time and age. Individual-level predictions of longitudinal trajectories for the testing data were performed, with results of the estimated subject-specific latent time-shifts showing that progression latent times are in concordance with the sub-groups diagnosis labels. Different levels of correlation (0.7< r< 1;0.3< r< 0.7;0< r< 0.3) are uncovered in multi-domain measurements in AD progression in subgroups, revealing possible pathological patterns in AD patients.
|
