Alzheimer’s Dementia: Diagnosis and Prognosis using Neuro-Imaging Analysis

Abstract

Alzheimer’s disease (AD) is the most common type of progressive neurological disorder that leads to the death of brain cells over the time. It causes memory loss and decline in the cognitive skills among the elderly subjects. Early diagnosis of the progressive diseases plays a vital role in the healthcare community. Machine learning (ML) algorithms and various multivariate data exploratory tools are employed in the field of AD research. The main purpose of this work is to analyse the importance of features selection which in turn enhances the classification accuracy of the models. The hyper parameter tuning for Support Vector Machine (SVM) classification and Boruta algorithm for Random Forest (RF) classification are applied for the selection of optimal set of features. In this work, a five-stage ML pipeline with each stage further categorized into different sub-levels is proposed. Initially, the data collected from the Open Access Series of Imaging Studies (OASIS-2) dataset of Magnetic Resonance Imaging (MRI) brain images is explored and pre-processed using the imputation technique. Feature scaling of the pre-processed data is done using the Min-max scaling technique. Then, the classification techniques such as logistic regression, Decision Tree (DT) classification, Random Forest (RF) classification, Support Vector Machine (SVM) classification and AdaBoost Classification are applied to classify the data and finally the performance of the classifiers are compared in terms of accuracy, Area under the curve of the Receiver Operating Characteristic (AUC) curve and recall measures. From the performance analysis, it is concluded that the Random Forest (RF) classifier yields maximum accuracy, recall and AUC values. The hyperparameter tuning and Boruta algorithm added significance to the SVM and RF classification, thereby resulting in a F-score of 91% and 92% respectively.