Hybrid CS-DMRI: Periodic Time-Variant Subsampling and Omnidirectional Total Variation Based Reconstruction

Area of Science:

• Medical Imaging
• Image Reconstruction
• Signal Processing

Background:

• Compressive sensing (CS) accelerates dynamic MRI (DMRI).
• Online CS-DMRI offers speed, while offline CS-DMRI provides higher accuracy.
• A need exists for DMRI reconstruction balancing both speed and accuracy.

Purpose of the Study:

• To propose a hybrid CS-DMRI method for enhanced image reconstruction performance.
• To improve both the speed and accuracy of DMRI.

Main Methods:

• A hybrid CS-DMRI approach with periodic time-variant subsampling.
• Utilizing reference frames for frame prediction.
• Employing 2-D and 3-D omnidirectional total variation (OTV) regularization for exploiting data correlations.
• Solving the optimization model with iterative reweighted least squares and pre-conditioned conjugate gradient methods.

Main Results:

• The proposed hybrid method achieves superior reconstruction accuracy compared to existing techniques.
• The method demonstrates low computational complexity, comparable to online CS-DMRI.
• Omnidirectional total variation effectively leverages data correlations in multiple directions.

Conclusions:

• The hybrid CS-DMRI method successfully balances reconstruction speed and accuracy.
• The novel OTV regularization enhances image quality by exploiting spatio-temporal coherence.
• This approach offers a significant advancement for accelerated DMRI.