Interpolated Compressed Sensing for Calibrationless Parallel MRI Reconstruction

Parallel magnetic resonance imaging (pMRI) in clinical study are commonly acquired in multiple slices; parallely along different channels. Since, MRI traditionally suffers from slow data acquisition, reconstruction of images in clinical pMRI would be further slower. Compressed sensing MRI (CS-MRI) has successfully demonstrated its potential in reducing the scan time of pMRI by manifolds. Due to high correlation of adjacent slices in multislice sequence, interpolation of multi-slice data may be carried out to support non-uniform undersampling based CS reconstruction of slices in k-space. Exploiting intra/inter slice as well as multichannel data redundancy of multi-slice pMRI, it is possible to accelerate the scan time further. These correlations can be well modeled by introducing multidimensional wavelet forest sparsity and joint total variation regularization during the CS reconstruction. To validate our claim, a number of experiments are carried out with real pMRI datasets and results are compared with the state-of-the-art.