An integrated predictive model for Alzheimer's disease progression from cognitively normal subjects using generated...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder with early cognitive changes.
• Early diagnosis of AD is critical for timely intervention and management.
• Predicting AD progression from cognitively normal (CN) stages remains a challenge due to limited longitudinal data.

Purpose of the Study:

• To develop an integrated AI framework for predicting Alzheimer's disease progression from cognitively normal stages.
• To leverage ensemble transfer learning, generative modeling, and ROI extraction for AD prediction.
• To enhance model transparency by identifying key brain regions involved in disease progression.

Main Methods:

• Utilized the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.
• Employed a three-stage process: ensemble transfer learning for CN to MCI transition probability estimation, ViT-GANs for simulating future MRI images, and 3D CNN with isotonic regression for AD prediction.
• Applied Grad-CAM for interpreting critical regions of interest (ROIs).

Main Results:

• Achieved high accuracy (0.85) and F1-score (0.86) in predicting CN to AD conversion up to 10 years.
• Successfully generated synthetic MRI images simulating disease progression.
• Identified key brain regions associated with AD progression through ROI interpretation.

Conclusions:

• The proposed integrated framework shows significant potential for early Alzheimer's disease diagnosis.
• The approach addresses data limitations by generating synthetic images and improves interpretability.
• Offers a promising tool for personalized intervention strategies in Alzheimer's disease management.