Adaptive sampling and non linear reconstruction for cardiac magnetic resonance imaging

We show how an adaptive acquisition sequence and a non linear reconstruction can be efficiently used to reconstruct undersampled cardiac MRI data. We demonstrate that, by using the adaptive method and L0-homotopic minimization, we can reconstruct an image with a number of samples which is very close to the sparsity coefficient of the image without knowing a-priori the sparsity of the image. We highlight two important aspects: 1) how the shape and the cardinality of the starting dataset influence the acquisition/reconstruction process; 2) how well the termination criteria allows to fit the optimal number of coefficients. The method is tested on MRI cardiac images and it is also compared to the weighted Compressed Sensing. All the experiments and results are reported and discussed. © 2014 Springer International Publishing.

We show how an adaptive acquisition sequence and a non linear reconstruction can be efficiently used to reconstruct undersampled cardiac MRI data. We demonstrate that, by using the adaptive method and L 0-homotopic minimization, we can reconstruct an image with a number of samples which is very close to the sparsity coefficient of the image without knowing a-priori the sparsity of the image. We highlight two important aspects: 1) how the shape and the cardinality of the starting dataset influence the acquisition/reconstruction process; 2) how well the termination criteria allows to fit the optimal number of coefficients. The method is tested on MRI cardiac images and it is also compared to the weighted Compressed Sensing. All the experiments and results are reported and discussed.