Accelerated magnetic resonance thermometry in the presence of uncertainties

Abstract

A model-based information theoretic approach is presented to perform the task of magnetic resonance (MR) thermal image reconstruction from a limited number of observed samples on k-space. The key idea of the proposed approach is to optimally detect samples of k-space that are information-rich with respect to a model of the thermal data acquisition. These highly informative k-space samples can then be used to refine the mathematical model and efficiently reconstruct the image. The information theoretic reconstruction was demonstrated retrospectively in data acquired during MR-guided laser induced thermal therapy (MRgLITT) procedures. The approach demonstrates that locations with high-information content with respect to a model-based reconstruction of MR thermometry may be quantitatively identified. These information-rich k-space locations are demonstrated to be useful as a guide for k-space undersampling techniques. The effect of interactively increasing the predicted number of data points used in the subsampled model-based reconstruction was quantified using the L2-norm of the distance between the subsampled and fully sampled reconstruction. Performance of the proposed approach was also compared with uniform rectilinear subsampling and variable-density Poisson disk subsampling techniques. The proposed subsampling scheme resulted in accurate reconstructions using a small fraction of k-space points, suggesting that the reconstruction technique may be useful in improving the efficiency of thermometry data temporal resolution.

Systematic review of reconstruction techniques for accelerated quantitative MRI

“…320 However, given the many possible combinations of acquisition settings and undersampling patterns, empirical optimization of in vivo precision is impractical. Hence, efforts for in-silico evaluation, such as predicting time efficiency, 12,266,[320][321][322][323][324][325] or accuracy, 118 are relevant. Recently, automated learning-based methodologies have been proposed to select an optimal sampling strategy independent of the model.…”

An information theory model for optimizing quantitative magnetic resonance imaging acquisitions

“…Compressed sensing allows measurement of a small number of signal samples without previous knowledge of the signal or image, whereas the technique described in this work provides a method to quantitatively optimize acquisition parameters given knowledge of the signal or imaging location. [6,[199][200][201] Furthermore, compressed sensing can be included in the signal model, so optimal acquisition parameters can be obtained for such acquisitions, and compressed sensing parameters can even be included in the optimization space. Appendix A…”

“…Acquisition parameters-the quantities that determine the acquisition sequence, such as excitation angle, TR, TE, and delay times-are currently selected on the basis of a combination of simple models and experience. As an alternative, this work develops a quantitative framework using mutual information to evaluate the information content [6] of quantitative MRI measurements and to guide acquisition parameter selection for the optimization of multi-parameter mapping.…”

“…Calculation of posterior variance can be accomplished through filters for parameter and state estimation. This has already been accomplished using the Kalman filter [6]. Several additional filters could be applied to this problem [161], including linear Gaussian filters (modified Kalman filter [162]), nonlinear Gaussian filters (extended Kalman filter [163], unscented Kalman filter [164], Gauss-Hermite filter [165,166], quadrature Kalman filter [165], cubature Kalman filter [167]), adaptive filtering (adaptive filter with gain adaptation [168,169], multiple models adaptive filtering [170][171][172]), and robust filtering (robust Kalman filter [173], variable structure filtering, H ∞ filtering [174]).…”

“…6 shows the change in M0 standard deviation in NIST phantom elements between a low-resolution pre-scan at one set of acquisition parameters and a high-resolution scan at a second set of acquisition parameters as a function of conditional mutual information. It can be seen that the change in M0 standard deviation is negatively correlated with conditional mutual information.…”