Non uniform sampling for sparse 2D correlated MRS: a quantitative point of view
Résumé
2D MR spectroscopy has the advantage to increase spectral resolution by unraveling the spectral information along a second dimension and resultsin a gain in quantification precision at the cost of increased acquisition time. To overcome this issue, non uniform sampling (NUS) allowing acquisition acceleration has recently been revisited as a solution in combination with the compressed sensing (CS) reconstruction or maximum entropy. NUS scheme is generally designed with randomly chosen indirect dimension step (t1). Besides, by taking advantage of the prior knowledge on the spectral support of sparse spectra, one can optimize the indirect dimension NUS regarding the reconstructed spectral information in COSY (Correlation spectroscopy) type spectra. The 2D time domain signal is a sum of separable terms resulting in separable spectrum with same spectral support in the two frequency domains. The sequential backward selection (SBS) algorithm can be used for choosing the indirect dimension step leading to decreasing noise and error in the reconstruction of the spectrum. The aim of this work is to demonstrate and study the feasibility of time domain quantification for NUS applied on Localized COSY (L-COSY) spectra by using this SBS approach.