GAN-enhanced deep learning for improved Alzheimer's disease classification and longitudinal brain change analysis

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging Analysis

Background:

• Traditional Alzheimer's disease (AD) diagnosis is time-consuming and prone to errors.
• Existing methods struggle with large datasets, leading to slower and potentially inaccurate classifications.
• Advancements in AI, machine learning (ML), and deep learning (DL) offer potential solutions for improved AD detection.

Purpose of the Study:

• To enhance the accuracy and efficiency of Alzheimer's disease (AD) detection and classification.
• To develop an AI model capable of predicting AD progression.
• To integrate deep learning with brain simulation for deeper insights into AD mechanisms.

Main Methods:

• Utilized deep learning architectures, including ResNet101 with novel layers (PDPO, DCK) for feature extraction from ADNI and OASIS datasets.
• Employed Long Short-Term Memory (LSTM) networks for classifying individuals into cognitively normal (CN), mild cognitive impairment (MCI), and AD categories.
• Integrated Generative Adversarial Networks (GANs) to determine AD progression (progressive vs. non-progressive).
• Incorporated brain simulation models for visualizing and predicting AD evolution.

Main Results:

• Achieved high accuracy rates: 0.9931 on the ADNI dataset and 0.9985 on the OASIS dataset.
• The GAN model successfully identified the progressive nature of AD.
• The integrated approach offers a nuanced understanding of AD for personalized treatment.

Conclusions:

• The proposed AI framework significantly improves AD diagnosis accuracy and efficiency.
• This approach facilitates personalized treatment strategies by predicting disease progression.
• The integration of AI and brain simulation provides valuable insights into AD's neural mechanisms.