MRI image based Ensemble Voting Classifier for Alzheimer's Disease Classification with Explainable AI Technique

Abstract

Alzheimer's is one of the causes of dementia, which causes memory loss, problem-solving disability, speaking, and a lot more difficulties in day-to-day life. Generally, dementia is a loss of memory, problem-solving ability, language fluency, and other thinking abilities that severely affect day-to-day life. Alzheimer's creates a huge impact on family life, the economy, and finally, the country as a whole is affected. According to statistics every 3 seconds, one person develops dementia in the world and the estimates say that by 2030, 78 million people will be affected, and by 2050 139 million people will have dementia. Estimates say that the economic impact due to dementia by 2030 in the US will be $2.8 Trillion which causes a huge loss and needs to be avoided.Alzheimer's can be diagnosed at various stages, with different datasets like Magnetic Resonance Imaging (MRI) Test images, Speech Tests, Symptoms, genes, and other data. Several models are developed to diagnose, but doctors expect proper insights about results apart from diagnosis, so the paper explains the results using various explainable methods like SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME).Data Sets used are MRI Features data extracted with generic information, Cross-sectional MRI data, and Longitudinal MRI Data. The step-by-step data processing includes data balancing using SMOTEENN, and then data transferred, using Quantile Transformer and PCA dimension reduction technique for 6 features, and Meta machine learning model, first level six key machine learning methods and finally voting classifier with hyperparameter tuning to get performance, 97.6 %, Precision 95.8%, recall 97.9% and finally F1 Score 96.8%.