Deep convolutional generative adversarial network for Alzheimer's disease classification using positron emission...

Area of Science:

• Biomedical imaging
• Artificial intelligence
• Neurology

Background:

• Deep learning and computer vision have advanced biomedical tasks.
• Supervised deep learning requires large labeled datasets, which are challenging to obtain for diseases like Alzheimer's.
• Limited data hinders the performance of automated disease diagnosis models.

Purpose of the Study:

• To develop a novel approach for generating synthetic brain positron emission tomography (PET) images for three stages of Alzheimer's disease (AD): normal control (CN), mild cognitive impairment (MCI), and AD.
• To improve the accuracy of automated disease diagnosis models by synthesizing essential training data.

Main Methods:

• Utilized deep convolutional generative adversarial networks (DCGANs) to synthesize brain PET images.
• Generated images representing three distinct stages of Alzheimer's disease: normal control (CN), mild cognitive impairment (MCI), and Alzheimer's disease (AD).

Main Results:

• The proposed DCGAN model demonstrated superior performance in synthesizing brain PET images across all three disease stages.
• A classification model trained on synthetic images achieved 72% accuracy in disease diagnosis.
• Quantitative evaluation using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) yielded high scores, indicating good image synthesis quality.

Conclusions:

• The novel DCGAN-based approach effectively synthesizes brain PET images for Alzheimer's disease stages.
• The generated synthetic data can significantly improve the performance and accuracy of automated disease diagnosis models.
• This method addresses the challenge of limited labeled data in medical imaging AI.