Predicting Conversion to Dementia Using Machine Learning and Neuropsychological Test Data

Abstract

Future disease-modifying treatments will target early stages of Alzheimer’s disease and related conditions to mitigate functional loss, which will require identification of pre-clinical individuals at risk for developing dementia. Machine learning algorithms using various parameters such as neuroimaging, genetics, and neuropsychological test data have shown promise in recent years. Here, we examine the performance of several such algorithms in a large, well-characterized national sample. To evaluate the potential for future clinical implementation, we used machine learning techniques to predict conversion to dementia using readily available and cost-effective neuropsychological test scores, basic health history, and demographic information.