Joint calibrationless reconstruction of highly undersampled multicontrast MR datasets using a low-rank Hankel tensor...

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Image Reconstruction

Background:

• Undersampled MRI data requires advanced reconstruction techniques.
• Multicontrast MRI datasets contain shared information beneficial for joint reconstruction.
• Existing methods may not fully exploit correlations across different contrasts.

Purpose of the Study:

• To develop a novel framework for joint reconstruction of highly undersampled multicontrast 2D MRI datasets.
• To leverage shared information across multiple contrasts for improved reconstruction accuracy and efficiency.
• To enable reconstruction without the need for coil-sensitivity calibration.

Main Methods:

• Proposed a multicontrast Hankel tensor completion (MC-HTC) framework.
• Organized multicontrast k-space data into a block-wise Hankel tensor.
• Employed higher-order singular value decomposition (HOSVD) for low-rank approximation, exploiting structural correlations and shared coil sensitivities.
• Recovered missing k-space data through iterative low-rank approximation and data consistency enforcement.
• Evaluated on multicontrast multichannel 2D human brain datasets with various undersampling schemes.

Main Results:

• The MC-HTC framework achieved high acceleration in image reconstruction.
• Demonstrated significantly lower residual errors compared to single-contrast SAKE and multicontrast J-LORAKS.
• Successfully applied to Cartesian uniform undersampling with a novel complementary k-space sampling strategy.
• Orthogonally alternated phase-encoding directions among contrasts for enhanced reconstruction.

Conclusions:

• The MC-HTC approach is an effective tensor completion framework for joint reconstruction.
• Enables high-quality reconstruction of highly undersampled multicontrast 2D datasets.
• Eliminates the requirement for coil-sensitivity calibration in the reconstruction process.