Deep Multi-Branch CNN Architecture for Early Alzheimer's Detection from Brain MRIs.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging Analysis

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder causing dementia and significant cognitive decline.
• Early diagnosis of AD is crucial for intervention and slowing disease progression.
• The economic and social burden of AD is substantial, highlighting the need for effective diagnostic tools.

Purpose of the Study:

• To review existing early detection methods for Alzheimer's disease.
• To propose and evaluate a novel deep convolutional neural network (CNN) for the early diagnosis of AD.
• To assess the accuracy of the proposed CNN model in classifying dementia severity.

Main Methods:

• A comprehensive review of current AD early detection strategies.
• Development of a deep CNN architecture with 7,866,819 parameters, featuring three distinct convolutional branches with varying kernel sizes.
• Training and validation of the CNN model on a relevant dataset for AD diagnosis.

Main Results:

• The proposed deep CNN model achieved a three-class accuracy of 99.05% in predicting dementia severity (non-demented, mild-demented, moderately demented).
• The model's architecture, with its multi-branch convolutional approach, proved highly effective for AD classification.
• Demonstrated exceptional performance in distinguishing between different stages of dementia.

Conclusions:

• The developed deep CNN model represents a significant advancement in the early and accurate diagnosis of Alzheimer's disease.
• This AI-driven approach holds considerable potential for improving patient care and management of AD.
• Highlights the efficacy of deep learning in addressing complex neurodegenerative disease diagnostic challenges.