Real time dynamic MRI by exploiting spatial and temporal sparsity.

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging
• Image Reconstruction

Background:

• Online imaging reconstruction typically relies on previous frames, limiting real-time applications.
• Dynamic magnetic resonance imaging (dMRI) requires fast and accurate frame reconstruction for clinical utility.
• Existing methods often struggle with the computational demands of real-time dMRI.

Purpose of the Study:

• To develop a novel, parallelizable scheme for real-time dynamic MRI reconstruction.
• To introduce an online dynamic total variation model for exploiting spatio-temporal sparsity.
• To design an efficient algorithm for rapid reconstruction of individual frames.

Main Methods:

• A parallelizable reconstruction scheme where frames (except the first) are independent.
• Implementation of dynamic total variation for joint spatial and temporal sparsity.
• A reweighted least squares algorithm accelerated by preconditioned conjugate gradient descent.

Main Results:

• The proposed method enables parallel processing, allowing immediate reconstruction post-data acquisition.
• Experimental results on cardiac dMRI datasets demonstrate significant improvement over existing online methods.
• Reconstruction accuracy is comparable to state-of-the-art offline methods.

Conclusions:

• The novel scheme offers a viable solution for real-time dMRI reconstruction.
• The dynamic total variation model effectively captures spatio-temporal information.
• The accelerated algorithm ensures fast and accurate image reconstruction, suitable for clinical applications.