Compressed sensing based dynamic MR image reconstruction by using 3D-total generalized variation and tensor...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Image Reconstruction

Background:

• Dynamic cardiac MR imaging (DCMRI) requires accelerated acquisition due to the fast motion of the heart.
• Compressed Sensing (CS) techniques leverage data sparsity and low-rank properties to reconstruct images from undersampled k-space data.
• Existing CS methods for DCMRI face challenges in balancing reconstruction accuracy and acceleration factors.

Purpose of the Study:

• To develop and evaluate a novel CS algorithm for improved DCMRI reconstruction quality.
• To minimize k-space data acquisition while enhancing image reconstruction accuracy.
• To investigate the synergistic integration of 3D total generalized variation (3D-TGV) and high order singular value decomposition (HOSVD) for sparse representation.

Main Methods:

• The proposed k-t TGV-TD method integrates 3D-TGV for localized sparsity and HOSVD for low-rank structure of 3D dynamic cardiac MR data.
• The Fast Composite Splitting Algorithm (FCSA) is employed to efficiently solve the resulting low-rank and sparse problem.
• Performance is evaluated on cardiac perfusion and cine MR datasets, comparing against state-of-the-art methods.

Main Results:

• The k-t TGV-TD method demonstrated superior reconstruction accuracy compared to existing methods, evidenced by higher Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM).
• Significantly better and more stable reconstruction results were achieved, particularly for cardiac perfusion datasets.
• The method effectively handles various acceleration factors, maintaining high reconstruction quality.

Conclusions:

• The k-t TGV-TD method is an effective sparse representation technique for DCMRI.
• It significantly improves reconstruction accuracy and image quality in accelerated DCMRI.
• The proposed method offers a promising solution for faster and more accurate cardiac MR imaging.