Coil Sketching for Fast and Efficient 4D Lung MRI Reconstruction

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging
• Computational Imaging

Background:

• Respiratory-resolved 4D MRI offers vital insights into lung structure and function.
• High computational and memory requirements limit current 4D lung MRI acquisition and reconstruction.
• Accelerated reconstruction is crucial for widespread clinical adoption of advanced lung MRI techniques.

Purpose of the Study:

• To develop and evaluate a memory-efficient and accelerated reconstruction framework for 4D lung MRI.
• To enable high-quality motion-compensated low-rank (MoCo-LR) reconstructions using coil sketching and Toeplitz approximation.
• To achieve these reconstructions on clinically accessible GPU hardware.

Main Methods:

• Extended the 3D coil sketching framework to 4D (3D + time) for reduced GPU memory consumption and reconstruction time.
• Implemented Toeplitz approximation to accelerate the normal encoding operator, further decreasing computational load.
• Utilized a MoCo-LR regularization technique with forward deformations for enhanced reconstruction stability.

Main Results:

• The 4D coil sketching method achieved an approximate 3-fold reduction in memory usage compared to fully sampled reconstructions.
• High-resolution MoCo-LR reconstructions were completed in under 10 minutes on GPUs with less than 48 GB of memory.
• Coil sketching preserved lung parenchymal signal and structural fidelity, outperforming conventional coil compression, especially in low-SNR regions.

Conclusions:

• Coil sketching and Toeplitz approximation offer a generalizable and hardware-efficient solution for 4D lung MRI reconstruction.
• These techniques effectively reduce computational barriers and enhance reconstruction speed while maintaining image quality.
• The developed framework paves the way for broader clinical implementation of respiratory-resolved lung MRI.