GRASP-Pro: imProving GRASP DCE-MRI through self-calibrating subspace-modeling and contrast phase automation

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Medical Imaging Technology
• Radiology

Background:

• Dynamic Contrast-Enhanced MRI (DCE-MRI) is crucial for assessing tissue perfusion and characterizing lesions.
• Accelerated imaging techniques are needed to improve patient comfort and reduce motion artifacts.
• Existing methods like GRASP (golden-angle radial sparse parallel imaging) offer improvements but can be further optimized.

Purpose of the Study:

• To introduce GRASP-Pro, a novel, accelerated, high-resolution DCE-MRI technique.
• To enhance image quality and reconstruction speed through joint sparsity and self-calibrating subspace constraints.
• To automate the selection of relevant contrast phases using a neural network.

Main Methods:

• GRASP-Pro utilizes a combination of low-rank subspace and temporal sparsity constraints for reconstruction.
• The temporal basis for the subspace is learned directly from the low-resolution k-space data.
• A convolutional neural network analyzes arterial enhancement curves to automatically select optimal contrast phases.

Main Results:

• GRASP-Pro demonstrated significantly improved image sharpness and reduced noise compared to standard GRASP.
• Finer delineation of prostate structures was achieved without increasing reconstruction time.
• The automated contrast phase selection using the neural network showed good correlation with manual selection.

Conclusions:

• GRASP-Pro offers superior reconstruction performance for rapid and continuous DCE-MRI.
• The technique enables high spatiotemporal resolution imaging with enhanced image quality.
• Automated contrast phase selection streamlines the GRASP MRI workflow.