Calibrationless parallel imaging reconstruction for multislice MR data using low-rank tensor completion.

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Signal Processing

Background:

• Accelerated MRI acquisition is crucial for reducing scan times.
• Reconstructing highly undersampled multislice data presents significant challenges.
• Existing single-slice reconstruction methods struggle with artifacts in accelerated multislice imaging.

Purpose of the Study:

• To develop a joint, calibrationless parallel imaging reconstruction method for highly accelerated multislice 2D MR k-space data.
• To leverage similarities between adjacent slices for improved image reconstruction.
• To enhance image quality in accelerated multislice MRI.

Main Methods:

• Multichannel k-space data from multiple slices were formed into a block-wise Hankel tensor.
• Tensor low-rankness was promoted using higher-order Singular Value Decomposition (SVD).
• The method was applied to 2D spiral and Cartesian k-space undersampling and evaluated on human brain MR data.

Main Results:

• The multislice block-wise Hankel tensor completion robustly reconstructed highly undersampled data.
• The proposed method produced substantially lower artifact levels compared to single-slice reconstruction.
• Quantitative analysis showed significantly improved performance in reducing residual artifacts and root mean square error.

Conclusions:

• The developed method effectively exploits similarities in coil sensitivity and image content within multislice MR data.
• This tensor completion framework offers a novel and effective approach for calibrationless reconstruction of highly undersampled multislice MR data.
• The technique holds promise for faster and higher-quality MRI acquisition.