Early Detection of Alzheimers Disease Using Deep Learning

Abstract

Alzheimer's Disease acts as a degenerative brain trauma that results in poor memory, alongside impairing mental functioning. Early detection of Alzheimer's Disease is aimed at controlling symptoms in addition to improving medical care for patients. Traditional diagnosis based on imaging procedures in conjunction with cognitive examinations takes a long time for experienced experts to analyze. Deep learning technology triggered interest in automated MRI data analysis techniques because they execute quickly and attain high accuracy levels. The application explores the development of an Alzheimer's Disease Diagnosis System as a web-based light application by using a custom Convolutional Neural Network (CNN) in its implementation. The system analyzes brain MRI scans to perform patient group assignment into Mild Demented, Moderate Demented, Very Mild Demented, and Non-Demented categories. The system shows a valid accuracy of 88.5%, prediction probabilities, and stage-dependent medical advice that generate downloadable diagnosis reports on the Streamlit platform. The system maintains user privacy through non-storage of their information and includes readable medical warning notices during use. The system is a working bridge between experimental technologies for Alzheimer's identification and available medical practices in clinical environments.