CORE-PI is a general reconstrution method, suitable for image reconstruction from multi-coil (parallel imaging) acquisition of 2D Cartesian k-space data. This method was published in: Shimron, Webb, Azhari, "CORE-PI: Non-iterative convolution-based reconstruction for parallel MRI in the wavelet domain." Medical Physics 46.1 (2019):199-214
CORE-PI is a parameter-free method, so users do not need to calibrate any params!
CORE-PI is enables flexible 1D undersampling of a 2D Cartesian k-space. The toolbox includes demos with various undersampling schemes - periodic, varying-period, variable-density and random schemes.
A liscence for Matlab is required. The code was tested with Matlab2017R.
Clone or download the CORE-PI code.
Open the "main.m" function in Matlab, choose one example from the list, and run the code.
There are 9 reconstruction examples, divided to 3 groups:
-
Analyltical brain phantom demos - reconstructions were performed for different subsampling schemes, all with a reduction factor (sub-sampling rate) of R=6:
- Periodic
- Varying-period
- Variable-density
- Random subsampling In all these demos CORE-PI was impelmented with wavelet 'db2'.
-
Analyltical brain phantom demos in which CORE-PI was implemented using different wavelet types:
- haar
- coif1
- sym4.
-
In-vivo 7t brain scans demos - data was retrospectivly subsampled with R=4 (using periodic subsampling), and CORE-PI was impelmented with wavelet 'db2'.
Results for examples in group 1:
Example from group 3:
The in-vivo data is courtesy of Prof. Andrew G. Webb from Leiden University Medical Center (LUMC).
The Realistic Analytical Brain Phantom data was utilized here with permission from the authors of: Guerquin-Kern, Matthieu, et al. "Realistic analytical phantoms for parallel magnetic resonance imaging." IEEE Transactions on Medical Imaging 31.3 (2011): 626-636. If you use that data in your publications, please cite this paper.