A Brief Review of Image Classification Techniques for Alzheimer’s Disease Detection

Abstract

Alzheimer’s disease (AD) is a neurodegenerative disease that causes memory loss, and commonly occurs among the older population. This type of dementia is preferably detected by non-invasive imaging techniques available, such as magnetic resonance imaging (MRI), positron emission tomography (PET), and computed tomography (CT) scan. MRI is the most common modality used for categorization and training purposes. Several studies have focused on detecting and visualization of the imaging content for early detection of the disease using the head prototype, pixel counting-based method, and efficient fuzzy C means adaptive thresholding and deep-learning based reconstruction approach. In recent advancements, convolutional neural network (CNN) is being utilized to detect and train AD classification. CNN provides a more extensive classification range as compared to the state-of-the-art predictors. The prominent techniques proposed using CNN include 3D CNN, random forest, divNet, linear discrimination analysis, support vector machine, decision tree, and K-nearest neighbor. Significant comparisons were made between AD, cognitively normal (CN), mild cognitive impairment (MCI), normal control (NC), and healthy control (HC). In this chapter, we summarize the earlier studies based on the dataset and modality used, techniques applied to various classifications, and observed results. For comparing the state-of-the-art techniques, we primarily classify the reported techniques as non-CNN-based techniques and CNN-based techniques. The results of the non-CNN-based techniques are reported based on peak signal-to-noise ratio and structural similarity index measure index values, and for CNN-based techniques, respective accuracies have been mentioned.