A Deep Siamese Convolution Neural Network for Multi-Class Classification of Alzheimer Disease

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) causes irreversible memory cell damage, leading to dementia, making early diagnosis challenging.
• Machine learning and deep convolutional neural networks (CNNs) show promise for analyzing brain MRI data in AD diagnosis.
• Overfitting due to limited data often hinders deep learning model performance in medical image analysis.

Purpose of the Study:

• To develop an accurate and efficient deep learning model for classifying dementia stages using MRI images.
• To address data scarcity and imbalance issues in deep learning models for AD detection.
• To improve the performance and accuracy of dementia stage classification compared to existing state-of-the-art methods.

Main Methods:

• A Siamese convolutional neural network (SCNN) model, inspired by VGG-16, was developed for dementia classification.
• Data augmentation techniques were employed to extend insufficient and imbalanced datasets.
• Experiments were conducted on the Open Access Series of Imaging Studies (OASIS) dataset.

Main Results:

• The proposed SCNN model achieved a high test accuracy of 99.05% for classifying dementia stages.
• The model demonstrated superior performance, efficiency, and accuracy compared to state-of-the-art models.
• Data augmentation effectively mitigated overfitting issues associated with limited medical imaging data.

Conclusions:

• The developed SCNN model offers a highly accurate and efficient solution for early dementia stage classification.
• The approach effectively handles data limitations in medical imaging datasets through augmentation.
• This research advances the application of deep learning in neuroimaging for Alzheimer's disease diagnosis.