A Methodical and Performance-based Investigation of Alzheimer Disease Detection on Magnetic Resonance and Multimodal Images

Abstract

Background: In recent years, Alzheimer's Disease (AD) has received more attention in the field of medical imaging, which leads to cognitive disorders. Physicians mainly rely on MRI imaging to examine memory impairment, thinking skills, judge functional abilities, and detect behavioral abnormalities for diagnosing Alzheimer's disease. Objective: Early diagnosis of AD has become a challenging and strenuous task with conventional methods. The diagnostic procedure becomes complicated due to the structure and heterogeneous dimensions of the brain. This paper visualizes and analyzes the publications on AD and furnishes a detailed review based on the stages involved in the early detection of the disease. Methods: This paper also focuses on assorted stages of disease detection such as image preprocessing, segmentation, feature extraction, classification, and optimization techniques that have been used in the diagnosis of AD during the past five years. It also spotlights the deep learning models used in assorted stages of detection. This paper also highlights the benefits of each method for assorted modalities of images. Results: AD has been analyzed with various computational methods on a few datasets, which leads to high computation time and loss of important features. Hybrid methods can perform better in every diagnosis stage of AD than others. Finally, the assorted datasets used for the diagnosis and investigation of Alzheimer's disease were analyzed and explored using a computerized system for future scope. Conclusion: From the review papers, we can conclude that DNN has greater accuracy in MR images and CNN +AEC has the best accuracy in the multimodal images.