Machine learning to forecast patient risk of developing Alzheimer's

 
   Lead Investigator:    Andy Lee
   Institution      :    Ikigai
   E-Mail           :    andylee024@gmail.com
   Proposal ID      :    1077



   Proposal Description:
   For many Alzheimer's drugs, the ideal patient is someone who exhibits early signs of dementia, but is not too deep in his disease progression. This is based on a common belief that no drug can slow down or stop disease pathology once a patient has lost too many brain cells.  This study aims to determine the feasibility of machine learning models for forecasting a patient's risk of developing dementia based on various biomarkers. In particular, we would like to evaluate the use of machine learning methods for predicting a patient's clinical dementia rating and timeline of disease progression using a combination of biomarker data (imaging, sequencing, protein measurements). State of the art machine learning models work well on multimodal data sources, so we expect that multiple biomarkers on a patient will be more predictive than single biomarkers.    Our specific objectives are given as follows.  1. Evaluate multiple machine-learning models that predict a patient's clinical dementia rating and future disease progression based on multiple patient biomarkers       2. Evaluate the accuracy and robustness of these machine learning models  3. Determine if the best machine learning model can help pharmaceutical efforts better stratify patient populations in order to improve the success rate of clinical trials.