Residual-Based Multi-Stage Deep Learning Framework for Computer-Aided Alzheimer's Disease Detection

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's Disease (AD) is a leading cause of dementia globally, affecting over 50% of elderly Japanese.
• Current AD detection methods struggle with the complexity of deep learning models.
• There is a critical need for automated and accurate AD detection systems.

Purpose of the Study:

• To introduce a novel multi-stage deep neural network for enhanced Alzheimer's Disease detection.
• To address the limitations of existing hierarchical convolutional neural networks (CNNs) in AD analysis.
• To improve the accuracy and efficiency of automated AD detection systems.

Main Methods:

• A five-stage deep neural network architecture utilizing residual functions for feature enhancement.
• Integration of a deep learning-based feature selection module with batch normalization, dropout, and fully connected layers to prevent overfitting.
• Classification using machine learning algorithms: Support Vector Machines (SVM), Random Forest (RF), and SoftMax.

Main Results:

• The proposed model achieved high accuracy rates: 99.47% on ADNI1, 99.10% on MIRAID, and 99.70% on OASIS Kaggle datasets.
• The system demonstrated superior performance compared to existing methods in binary classification tasks.
• The multi-stage architecture effectively enhanced feature extraction and model depth.

Conclusions:

• The novel multi-stage deep neural network offers a significant advancement in Alzheimer's Disease analysis.
• The model's high accuracy suggests its potential for reliable automated AD detection.
• This approach paves the way for more effective early diagnosis and management of Alzheimer's Disease.