Classification of Neurodegenerative Disease Stages using Ensemble Machine Learning Classifiers

Abstract

Existing research works for Alzheimer’s disease (AD) can predict the prevalence of disease only after the advancement of the disease. With these existing prediction models, it is possible only to reduce and delay the symptoms of the disease. The exact usefulness is when the presence of the disease is identified at an early stage and this early detection makes a great impact in subjects’ recovery. Thus, early detection of controls at high risk of development of Alzheimer’s disease is of a key objective of the proposed work. Existing machine learning and deep learning algorithms derive only limited predictive accuracy. Also, they derive results based on expensive machine learning algorithms that had hard-to-collect features and classifying becomes complex with numerous overfitting in choosing decision boundaries. The proposed study intends to develop a learning algorithm for the prediction of Alzheimer’s disease at an early stage. It also classifies the features if the subjects with Mild Cognitive impairment (MCI) and Pre-Mild Cognitive Impairment (Pre-MCI)has the likelihood to develop Alzheimer’s disease. A dataset of AD controls was used to train different machine learning algorithms. Onset information like social behavior, demographic characteristics, neurological test scores, clinical cardiovascular index, and brain atrophy ratio can also be used as the extract predictor. A validation procedure was applied to identify a relevant subset of predictors. The conversion to AD in MCI and Pre MCI subjects are based only on non-invasively and effectively collectible predictors.