A genetic programming Rician noise reduction and explainable deep learning model for Alzheimer's diseases severity...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Magnetic Resonance Imaging (MRI) quality is often compromised by noise, particularly Rician noise.
• This degradation poses a significant challenge for accurate medical image analysis and disease diagnosis.

Purpose of the Study:

• To develop an effective method for Rician noise reduction in MRI images.
• To enhance the accuracy of Alzheimer's disease (AD) detection using a deep learning framework.

Main Methods:

• A genetic programming (GP) technique was employed for pre-processing MRI images to reduce Rician noise.
• An explainable deep learning framework utilizing SHAP values was developed for AD diagnosis, incorporating an oversampling strategy for class imbalance.

Main Results:

• The proposed model demonstrated superior performance compared to existing methods, including DenseNet169, VGGNet15, and Inceptionv3.
• The approach effectively handled images with limited spectral features, exhibited lower computational complexity, and reduced overfitting.

Conclusions:

• The research successfully addressed Rician noise in MRI and improved AD severity prediction.
• The combination of GP for noise reduction and explainable AI for diagnosis offers a promising approach for neurodegenerative disease research.