AI Models for Modeling and Simulation of Clinical Studies for Alzheimer's and Parkinson's Disease

Abstract

Neurodegenerative diseases (NDDs) have a complex structure and most of them are untreatable that’s why more research studies undertake translational paths for getting better insights into prevention, early detection, and better treatment options. A longitudinal understanding of disease development and progression across all biological scales is required for translational research of these diseases. However, due to the complexity underlying these diseases and their heterogeneous nature, there is a need for a comprehensive picture of a specific disease. For this purpose, multiple studies need to be compared and analyzed and several observational cohort studies and clinical trials are available for this purpose. Many of these clinical studies aim at early prognosis, drug development, and treatment of the disease. However, legal and ethical constraints typically do not allow for sharing of sensitive patient data. In consequence, there exist data silos, which slow down the overall scientific progress in translational research. <br /> In our work, we suggest artificial intelligence (AI) based methods that are generative in nature and help to model and simulate the clinical studies for Alzheimer’s disease (AD) and Parkinson’s disease (PD). The key idea here is to describe a longitudinal patient cohort with the help of a bayesian network (BN), in conjunction with deep learning methods. Our approach allows for incorporating arbitrary multi-scale, multi-modal data. As our method is generative in nature, we try to solve the problem of data sharing and data silos by generating synthetic data. We show that with the help of such a model, we can simulate subjects that are largely indistinguishable from real ones. Moreover, we demonstrate the possibility to simulate counterfactual interventions in a synthetic cohort. We also unravel the complexities underlying NDDs by disentangling and quantifying the connections between different clinical parameters.