Real time dynamic MRI with dynamic total variation.

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging
• Image Reconstruction

Background:

• Dynamic magnetic resonance imaging (dMRI) enables real-time visualization of physiological processes.
• Traditional dMRI reconstruction methods can be computationally intensive and slow.
• Online reconstruction methods aim to improve speed but often sacrifice accuracy.

Purpose of the Study:

• To develop a novel, parallelizable real-time dMRI reconstruction scheme.
• To improve reconstruction accuracy for dynamic MRI while maintaining online processing capabilities.
• To introduce a new dynamic total variation model and an accelerated algorithm for sparse MRI.

Main Methods:

• A novel parallelizable dMRI reconstruction scheme where frames after the first are reconstructed independently.
• Utilizing a dynamic total variation model to exploit spatial and temporal sparsity.
• Employing an accelerated reweighted least squares algorithm tailored for sparse MRI reconstruction.

Main Results:

• The proposed method allows for parallel processing of dMRI frames after the first frame is reconstructed.
• Demonstrated significant outperformance compared to existing online dMRI reconstruction methods.
• Achieved reconstruction accuracy comparable to state-of-the-art offline methods on in-vivo cardiac dMRI datasets.

Conclusions:

• The novel scheme offers an efficient and accurate solution for real-time dMRI reconstruction.
• The dynamic total variation model and accelerated algorithm effectively address reconstruction challenges.
• This approach advances the capabilities of dynamic MRI for clinical and research applications.