Using Predictive Algorithms to Improve Recruitment in AD Clinical Trials

The goal of this project is to develop tools that would allow recruitment of more appropriately targeted subjects for clinical trials. The project aims to develop predictive algorithms that can be used to improve recruitment in Alzheimer’s clinical trials. The predictive algorithms are based on state-of-the-art Deep Learning techniques and use routinely collected data, such as brain structural Magnetic Resonance Imaging.

Aim 1 will develop algorithms to identify individuals at the earliest disease stage, while Aim 2 will develop algorithms to predict future cognitive performance in individuals with Alzheimer pathology. The developed algorithms will be trained and evaluated using multiple datasets.

Our proposal is innovative at many levels. First, it focuses on evaluating whether state-of-the-art Deep Learning methods are capable of identifying individuals at the preclinical AD stage and predicting future cognitive performance based on routinely collected imaging data. Second, it uses two large, existing, longitudinal multimodal datasets, which allows for independent training and testing of the developed algorithms.

First, the developed predictive tools may improve clinical trial recruitment by selecting homogeneous groups at the earliest stage of AD pathology. Second, the developed tools may provide an early and accurate diagnosis, and thus enable intervention at a stage that it might be more impactful.