Extreme MRI: Large-scale volumetric dynamic imaging from continuous non-gated acquisitions

Area of Science:

• Medical Imaging
• Computational Imaging
• Magnetic Resonance Imaging (MRI)

Background:

• Dynamic MRI requires reconstructing large volumetric datasets from limited k-space data.
• Existing methods struggle with massive data sizes and computational demands of continuous acquisitions.

Purpose of the Study:

• To develop a framework for reconstructing large-scale volumetric dynamic MRI from rapid, continuous, non-gated acquisitions.
• To enable applications in pulmonary imaging and dynamic contrast-enhanced (DCE) imaging.

Main Methods:

• Leveraging 3D radial and 3D cones trajectories for incoherent k-space coverage.
• Employing compressed representation via multiscale low-rank matrix factorization for problem constraint and memory reduction.
• Utilizing stochastic optimization to enhance computation, memory locality, and inter-processor communication.

Main Results:

• Demonstrated feasibility on DCE and pulmonary imaging datasets.
• The proposed technique reveals transient dynamics missed by gating-based methods.
• Improved image feature sharpness, especially with irregular or non-repetitive motions.

Conclusions:

• Successfully demonstrated a method for reconstructing massive 3D dynamic image series.
• Addressed extreme undersampling and computational challenges in dynamic MRI reconstruction.