A multimodal fusion method for Alzheimer's disease based on DCT convolutional sparse representation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Multimodal medical imaging, combining MRI and PET, is crucial for Alzheimer's disease (AD) diagnosis.
• Traditional sparse representation-based fusion methods suffer from energy loss, low contrast, and spatial inconsistency.

Purpose of the Study:

• To propose a novel multimodal fusion algorithm for AD diagnosis.
• To address limitations of existing fusion techniques, enhancing diagnostic accuracy.

Main Methods:

• The proposed algorithm utilizes multi-scale Discrete Cosine Transform (DCT) decomposition of source images.
• Sparse coefficients are optimized using the Alternating Directional Method of Multipliers (ADMM).
• Fused images are reconstructed using an improved L1 parametric rule and Novel Sum-Modified Spatial Frequency (NMSF).

Main Results:

• The method demonstrates significant improvements in contrast enhancement.
• It effectively retains crucial texture and contour information in fused images.
• Experimental results validate the algorithm's superior performance.

Conclusions:

• The DCT convolutional sparse representation algorithm offers a robust solution for multimodal medical image fusion in AD diagnosis.
• This approach enhances image quality, supporting more accurate and reliable diagnostic outcomes.