Sparsity adaptive reconstruction for highly accelerated cardiac MRI

Area of Science:

• Medical Imaging
• Cardiovascular Magnetic Resonance (CMR)
• Image Reconstruction

Background:

• Accelerated MRI acquisition is crucial for reducing scan times and improving patient comfort.
• Compressed Sensing (CS) techniques accelerate MRI but typically require manual tuning of regularization parameters.
• Parameter tuning in CS can be complex and time-consuming, limiting its widespread clinical adoption.

Purpose of the Study:

• To develop a parameter-free method for accelerated cardiovascular magnetic resonance (CMR) imaging.
• To enable faster and more efficient CMR data acquisition without compromising image quality.
• To present Sparsity adaptive Composite Recovery (SCoRe) as a novel, automated CS recovery technique.

Main Methods:

• Introduced Sparsity adaptive Composite Recovery (SCoRe), a technique leveraging sparsity in multiple transforms.
• Implemented a data-driven approach to automatically adjust the contributions of different sparsifying transforms.
• Validated SCoRe using digital phantoms and retrospectively/prospectively undersampled cine CMR datasets.

Main Results:

• SCoRe demonstrated lower root-mean-square error (RMSE) and higher structural similarity index (SSIM) compared to existing CS methods in simulations and retrospective undersampling.
• Expert reviewers rated SCoRe-reconstructed images significantly higher (p < 0.01) than those from other CS methods in prospective undersampling studies.
• The method achieved superior performance by adapting regularization weights based on noise and sparsity levels.

Conclusions:

• SCoRe enables accelerated cine CMR acquisition from highly undersampled data.
• The technique provides superior performance by adaptively adjusting regularization weights, eliminating the need for manual parameter tuning.
• SCoRe offers a robust, parameter-free solution for accelerating CMR imaging, enhancing clinical utility.