Dementia Prediction Using OASIS Data for Alzheimer’s Research

Abstract

The role of data science and machine learning in the medical field has increased manifold in recent years. However there lies a vast scope in conditions like dementia. Complex machine learning models are in place to analyze brain images but these fail to perform on numeric biological and social data of the patients. This work analyses the longitudinal brain data of patients collected by OASIS for Alzheimer research. A graphical analysis is performed on the data and several conclusions regarding dementia have been drawn. Multilayer Perceptron and Decision Tree both provided an accuracy of 0.839 and a recall of 0.836 and 0.800 respectively thereby providing the most efficient model for dementia prediction. Machine learning models can help predict dementia using social and biological data of patients to a fairly accurate degree without the requirement of brain MRI images.