Visually weighted reconstruction of compressive sensing MRI.

Area of Science:

• Medical Imaging
• Signal Processing
• Computer Vision

Background:

• Compressive sensing (CS) allows magnetic resonance (MR) imaging with reduced scan times by reconstructing images from undersampled k-space data.
• Conventional CS reconstruction methods can introduce quality distortions compared to fully sampled images.
• Improving the visual quality of CS-reconstructed MR images remains a key challenge.

Purpose of the Study:

• To develop and evaluate a weighted optimization algorithm for enhancing the visual quality of MR images reconstructed using compressive sensing.
• To investigate the impact of incorporating human visual system (HVS) perceptual characteristics into the reconstruction process.
• To leverage wavelet transforms for improved data sparsity utilization in both spatial and frequency domains.

Main Methods:

• Variable density random undersampling in the phase encoding direction of k-space.
• Application of a weighted optimization algorithm incorporating visual weights, particularly for regions of interest (ROIs).
• Utilizing wavelet transform for image analysis and data sparsity exploitation across spatial and frequency domains.
• Employing ℓ1 norm minimization with prioritized coefficients based on visual weights.

Main Results:

• The proposed weighted optimization algorithm significantly improves the visual quality of reconstructed MR images.
• Images reconstructed with visual weights demonstrate superior local and global quality compared to conventional methods.
• Objective quality assessment metrics confirm the enhanced performance of the weighted reconstruction approach.

Conclusions:

• The developed weighted optimization algorithm effectively enhances the visual quality of compressive sensing-based MR images.
• Incorporating human visual system characteristics into the reconstruction process is a promising strategy for improving MR image fidelity.
• This method offers a viable approach to achieve high-quality MR imaging with reduced scan times.