Benchmarking Gaussian processes for prediction and data assimilation of Alzheimer's disease progression

The ability to predict the trajectory of disease progression with high resolution for individual patients can enhance clinical trial design, enabling personalized, data-driven medical approaches. In this study, we deployed a kernel/Gaussian process-based dynamic model to predict Alzheimer's disease progression. Our numerical results demonstrate that the dynamic method outperforms static linear regression, improving the prediction of ADAS-Cog 11 subscores over extended periods by effectively incorporating intermediate data observations. This approach highlights the potential of computational models in enhancing clinical trial design and advancing personalized medicine for Alzheimer's disease.