Free-Breathing and Ungated Dynamic MRI Using Navigator-Less Spiral SToRM

Area of Science:

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

Background:

• Dynamic MRI reconstruction from undersampled spiral acquisitions is challenging due to extensive motion.
• Traditional low-rank methods struggle when motion causes data matrices to exceed low-rank assumptions.
• Existing techniques often rely on explicit k-space navigators to guide reconstruction.

Purpose of the Study:

• To introduce a novel kernel low-rank algorithm for reconstructing free-breathing, ungated dynamic MRI.
• To overcome limitations of existing low-rank methods in handling complex motion artifacts.
• To enable accurate dynamic MRI reconstruction without the need for k-space navigators.

Main Methods:

• Developed a kernel low-rank matrix completion algorithm tailored for dynamic MRI.
• Exploited the non-linear structure of dynamic MRI data to form a low-rank kernel matrix.
• Applied the method to variable density spiral acquisitions, directly filling missing k-space data.

Main Results:

• The proposed kernel low-rank algorithm demonstrated improved reconstruction quality compared to classical low-rank and XD-GRASP methods.
• Validated performance using both simulated and in-vivo dynamic MRI data.
• Quantitative metrics from reconstructions showed agreement with breath-held cine data.

Conclusions:

• The kernel low-rank approach effectively reconstructs free-breathing, ungated dynamic MRI from spiral data.
• This method offers a navigator-free alternative, improving upon existing low-rank techniques.
• While image quality is comparable, minor degradations exist relative to breath-held scans.