Bayesian model for highly accelerated phase-contrast MRI
Methods and systems for accelerated Phase-contrast magnetic resonance imaging (PC-MRI). The technique is based on Bayesian inference and provides for fast computation via an approximate message passing algorithm. The Bayesian formulation allows modeling and exploitation of the statistical relationships across space, time, and encodings in order to achieve reproducible estimation of flow from highly undersampled data.
1. A method of image acquisition and reconstruction in a magnetic resonance imaging (MRI) apparatus, comprising:
acquiring, during a single breath-hold, undersampled phase-contrast magnetic resonance imaging (PC-MRI) data associated with physiological activity in an area of interest of a subject using Variable density incoherent spatiotemporal acquisition (VISTA) sampling;
determining conditional distributions for velocity-encoded measurements in the PC-MRI data for a velocity-free region and a velocity-containing region;
performing a Bayesian modeling of the PC-MRI data and conditional distributions to determine a posterior distribution;
determining a mean squared error (MMSE) estimate of image magnitude and velocity-encoded phase from the posterior distribution;
performing parallel image reconstruction from the estimate of image magnitude and velocity-encoded phase; and
displaying the reconstructed images corresponding to the physiological activity in the area of interest.
2. The method of claim 1 , wherein the Bayesian modeling is visualized as a factor graph.
3. The method of claim 2 , further comprising applying a sum-product rule using approximate message passing on the factor graph.
4. The method of claim 3 , further comprising determining approximate marginal posterior distributions of x b , x v , θ v , and v,
wherein x b represents an image associated with a velocity-compensated measurement, wherein x v represents an image associated with a velocity-encoded measurement, where θ v represents a velocity-encoded phase, and wherein v represents a Bernoulli random variable.
5. The method of claim 1 , wherein the Bayesian modeling provides for automatic tuning of at least one parameter associated with the image reconstruction.
6. The method of claim 5 , wherein the at least one parameter includes a prior probability of flow at each pixel and frame, and a variability between encodings.
7. The method of claim 1 , wherein PC-MRI data is downsampled to obtain acceleration factors R=1, 2, 4, 6, 8, 10 12, 14 and 16.
8. A magnetic resonance imaging (MRI) system, comprising:
an MRI machine that acquires, during a single breath-hold, undersampled phase-contrast magnetic resonance imaging (PC-MRI) data associated with physiological activity in an area of interest of a subject using Variable density incoherent spatiotemporal acquisition (VISTA) sampling and generates MRI image data;
a computing device that receives the PC-MRI data, determines conditional distributions for velocity-encoded measurements in the PC-MRI data for a velocity-free region and a velocity-containing region, performs a Bayesian modeling of the PC-MRI data and conditional distributions to determine a posterior distribution, determines a mean squared error (MMSE) estimate of image magnitude and velocity-encoded phase from the posterior distribution, and performs parallel image reconstruction from the estimate of image magnitude and velocity-encoded phase; and
a display that receives and displays the reconstructed images corresponding to the physiological activity in the area of interest.
9. The system of claim 8 , wherein the Bayesian modeling is visualized as a factor graph.
10. The system of claim 9 , wherein the computing device further applies a sum-product rule using approximate message passing on the factor graph.
11. The system of claim 10 , wherein the computing device further determines approximate marginal posterior distributions of x b , x v , θ v , and v,
wherein x b represents an image associated with a velocity-compensated measurement, wherein x v represents an image associated with a velocity-encoded measurement, where θ v represents a velocity-encoded phase, and wherein v represents a Bernoulli random variable.
12. The system of claim 8 , wherein the Bayesian modeling provides for automatic tuning of at least one parameter associated with the image reconstruction.
13. The system of claim 12 , wherein the at least one parameter includes a prior probability of flow at each pixel and frame, and a variability between encodings.
14. The system of claim 8 , wherein PC-MRI data is downsampled to obtain acceleration factors R=1, 2, 4, 6, 8, 10 12, 14 and 16.
15. A non-transitory computer readable medium containing computer executable instructions that when executed by a processor of a computing device causes the computing device to perform a method of image acquisition and reconstruction in a magnetic resonance imaging (MRI) apparatus, comprising:
acquiring, during a single breath-hold, undersampled phase-contrast magnetic resonance imaging (PC-MRI) data associated with physiological activity in an area of interest of a subject using Variable density incoherent spatiotemporal acquisition (VISTA) sampling;
determining conditional distributions for velocity-encoded measurements in the PC-MRI data for a velocity-free region and a velocity-containing region;
performing a Bayesian modeling of the PC-MRI data and conditional distributions to determine a posterior distribution;
determining a mean squared error (MMSE) estimate of image magnitude and velocity-encoded phase from the posterior distribution;
performing parallel image reconstruction from the estimate of image magnitude and velocity-encoded phase; and
displaying the reconstructed images corresponding to the physiological activity in the area of interest.
16. The non-transitory computer readable medium of claim 15 , wherein the Bayesian modeling is visualized as a factor graph.
17. The non-transitory computer readable medium of claim 16 , further comprising instructions for applying a sum-product rule using approximate message passing on the factor graph.
18. The non-transitory computer readable medium of claim 17 , further comprising instructions for determining approximate marginal posterior distributions of X b , X v , θ v , and v,
wherein x b represents an image associated with a velocity-compensated measurement, wherein x v represents an image associated with a velocity-encoded measurement, where θ v represents a velocity-encoded phase, and wherein v represents a Bernoulli random variable.
19. The non-transitory computer readable medium of claim 15 , wherein the Bayesian modeling provides for automatic tuning of at least one parameter associated with the image reconstruction.
20. The non-transitory computer readable medium of claim 15 , wherein PC-MRI data is downsampled to obtain acceleration factors R=1, 2, 4, 6, 8, 10 12, 14 and 16.