Digital Twins for Alzheimer’s Disease Patients

Teammates: Yiqiao Huang, Hongyi Lu, Yilin Ni

Advisor: Professor Thomas Bengtsson, UC Berkeley

We developed digital twins for Alzheimer’s Disease patients, i.e., predictions of outcomes for individual patients describing their likely progression in the future under standard-of-care given their baseline characteristics. We exploited the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database and extracted features including psychological assessments, biospecimen, genetics, and PET imaging. We trained a Conditional Restricted Boltzmann Machine (CRBM) to predict the distribution of a patient’s features at future time points given his/her features at baseline. We showed that our model effectively captures the distribution of variables and performs comparably to linear regression and random forest for time points near the baseline.