Title: Improved performance for Alzheimer's disease earlier detection and diagnosis using deep learning algorithms

Authors: T. Deenadayalan; S.P. Shantharajah

Addresses: School of Information Technology and Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, 632014, India ' School of Information Technology and Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, 632014, India

Abstract: As our society ages, cognitive impairment and dementia in the elderly, particularly Alzheimer's disease (AD), have become more prevalent. After clinical symptoms arise, senile cognitive impairment will advance to irreversible dementia, finally leading to death as a multifactor, multistage, and clinical syndrome with concomitant disorders. Alzheimer's disease is currently irreversible, and scientific trials for effective treatments are lacking. The progression of a patient's condition will go through numerous stages, so early detection is critical. Early Alzheimer's disease treatments can effectively decrease disease development while also reducing the burden on patients' families and society. The research presents a deep learning-based strategy for early detection and screening of Alzheimer's disease. The method involves slicing a three-dimensional magnetic resonance picture of the human brain into a two-dimensional image, then using an object recognition network called faster region with convolutional neural networks (R-CNN) to detect shrinkage in the hippocampal region of the brain to make an AD diagnosis. To get feature maps and to get 100% high-precision detection of AD samples, a new network is constructed and optimised based on VGG16, which is the basic network of faster R-CNN. At the same time, the validation set achieves 97.67% of the detected picture correctness.

Keywords: disease detection; feature extraction; CNN; deep learning algorithm; high-precision.

DOI: 10.1504/IJESMS.2024.139541

International Journal of Engineering Systems Modelling and Simulation, 2024 Vol.15 No.4, pp.174 - 180

Received: 29 Oct 2021
Accepted: 21 Feb 2022
Published online: 04 Jul 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article