Generalized approximate message passing algorithms for sparse magnetic resonance imaging reconstruction

Patent metadata

Field	Value
Publication number	US-9542761-B2
Application number	US-201514630712-A
Country	US
Kind code	B2
Filing date	Feb 25, 2015
Priority date	Feb 25, 2015
Publication date	Jan 10, 2017
Grant date	Jan 10, 2017

Abstract

A method for reconstructing magnetic resonance imaging data includes acquiring a measurement dataset using a magnetic resonance imaging device and determining an estimated image dataset based on the measurement dataset. An iterative reconstruction process is performed to refine the estimated image dataset. Each iteration of the iterative reconstruction process comprises: updating the measurement dataset and a sparse coefficient dataset based on the estimated image dataset and a plurality of belief propagation terms, incorporating a noise prior dataset into the measurement dataset, incorporating a sparsity prior dataset into the sparse coefficient dataset, updating the plurality of belief propagation terms based on the measurement dataset and the sparsity prior dataset, and updating the estimated image dataset based on the plurality of belief propagation terms. A reconstructed image and confidence map are generated using the estimated image dataset.

First claim

We claim:

1. A method for reconstructing magnetic resonance imaging data, the method comprising:
acquiring a measurement dataset using a magnetic resonance imaging device;
determining an estimated image dataset based on the measurement dataset;
performing an iterative reconstruction process to refine the estimated image dataset, wherein each iteration of the iterative reconstruction process comprises:
updating the measurement dataset and a sparse coefficient dataset based on the estimated image dataset and a plurality of belief propagation terms, wherein the sparse coefficient dataset comprises mean sparse coefficient data and variance sparse coefficient data,
incorporating a noise prior dataset into the measurement dataset,
incorporating a sparsity prior dataset into the sparse coefficient dataset by (a) updating the mean sparse coefficient data by applying a soft thresholding operator to the mean sparse coefficient data and (b) updating the variance sparse coefficient data by applying a derivative soft thresholding operator to the mean sparse coefficient data,
updating the plurality of belief propagation terms based on the measurement dataset and the sparsity prior dataset, and
updating the estimated image dataset based on the plurality of belief propagation terms; and

generating a reconstructed image and confidence map using the estimated image dataset.

2. The method of claim 1 , wherein the measurement dataset comprises mean measurement data and variance measurement data.

3. The method of claim 2 , wherein updating measurement dataset based on the estimated image dataset and the plurality of belief propagation terms by a process comprises:
determining the mean measurement data based on a mean of the estimated image dataset; and
determining the variance measurement data based on a variance of the estimated image dataset and one or more of the plurality of belief propagation terms.

4. The method of claim 2 , wherein incorporating the noise prior dataset into the measurement dataset is performed using a maximum a posteriori probability (MAP) estimation process.

5. The method of claim 1 , wherein updating sparse coefficient dataset based on the estimated image dataset and the plurality of belief propagation terms by a process comprises:
determining the mean sparse coefficient data based on a mean of the estimated image dataset; and
determining the variance sparse coefficient data based on a variance of the estimated image dataset and one or more of the plurality of belief propagation team.

6. The method of claim 1 , wherein the plurality of belief propagation terms comprise:
a first belief propagation term corresponding to mean measurement data;
a second belief propagation term corresponding to variance measurement data;
a third belief propagation term corresponding to mean sparse coefficient data; and
a fourth belief propagation term corresponding to variance sparse coefficient data.

7. The method of claim 1 , wherein the estimated image dataset comprises mean estimated image data and variance estimated image data.

8. The method of claim 7 , wherein the updating of the estimated image dataset based on the plurality of belief propagation terms comprises:
updating the mean estimated image data by applying a first transform operator to a first belief propagation term and a second transform operator to a second belief propagation teem; and
updating the variance estimated image data by applying the first transform operator to a third belief propagation term and the second transform operator to a fourth belief propagation term and multiplying by the mean estimated image data.

9. The method of claim 8 , wherein the first transform operator is an adjoint of a Fourier transform operator and the second transform operator is an adjoint of a wavelet transform operator.

10. The method of claim 7 , wherein the reconstructed image is generated based on the mean estimated image data and the confidence map is generated based on the variance estimated image data.

11. A method for reconstructing magnetic resonance imaging data, the method comprising:
obtaining k-space scan data captured by a MRI system, the k-space scan data being representative of an undersampled region over time; and
reconstructing an image dataset from the k-space scan data by applying generalized approximate message passing (GAMP) to solve an optimization problem which applies a Fourier transform and a wavelet transform to the k-space scan data, wherein GAMP is applied to solve the optimization problem by a process comprising:
determining a plurality of image distribution values based on the k-space scan data;
determining a plurality of measurement coefficient values using the plurality of image distribution values and a first belief propagation correction term;
incorporating a first prior data value into the plurality of measurement coefficient values;
determining a plurality of sparse coefficient values using the plurality of image distribution values and a second belief propagation correction term;
incorporating a second prior data value into the plurality of sparse coefficient values; and
updating the plurality of image distribution values based on the plurality of measurement coefficient values and the plurality of sparse coefficient values.

12. The method of claim 11 , wherein the image dataset comprises a reconstructed image and a mean squared error (MSE) map.

13. The method of claim 11 , wherein the first prior data value is incorporated into the plurality of measurement coefficient values using a maximum a posteriori (MAP) probability estimate.

14. The method of claim 13 , wherein the first prior data value comprises a Gaussian distribution of a noise and the second prior data value comprises a Laplacian distribution of a regularization term.

15. A system for reconstructing magnetic resonance imaging data, the system comprising:
an imaging device comprising a plurality of coils configured to acquire k-space scan data representative of an undersampled region over time; and

a central control computer unit configured to reconstruct an image dataset from the k-space scan data by applying generalized approximate message passing (GAMP) to solve an optimization problem which applies a Fourier transform and a wavelet transform to the k-space scan data, wherein the central control computer unit applies GAMP is applied using a process comprising:
determining a plurality of image distribution values based on the k-space scan data,
determining a plurality of measurement coefficient values using the plurality of image distribution values and a first belief propagation correction term,
incorporating a first prior data value into the plurality of measurement coefficient values;
determining a plurality of sparse coefficient values using the plurality of image distribution values and a second belief propagation correction term,
incorporating a second prior data value into the plurality of sparse coefficient values, and
updating the plurality of image distribution values based on the plurality of measurement coefficient values and the plurality of sparse coefficient values.

16. The system of claim 15 , wherein the central control computer unit is further configured to:
generate a reconstructed image based on the image dataset; and
generate a mean squared error (MSE) map based on the image dataset.

17. The system of claim 16 , further comprising:
a display configured to simultaneously present the reconstructed image and the MSE map.

Assignees

• Tan Jin
• Mailhe Boris
• Wang Qiu
• Nadar Mariappan S
• Siemens Healthcare Gmbh

Inventors

• Tan Jin
• Mailhe Boris
• Wang Qiu
• Nadar Mariappan S

Classifications

• G06T12/00 Primary

  Tomographic reconstruction from projections · CPC title

• G06T7/0012

  Biomedical image inspection · CPC title

• G06T11/003 Primary

  Physics · mapped topic

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title

• G01R33/5611 Primary

  Parallel magnetic resonance imaging, e.g. sensitivity encoding [SENSE], simultaneous acquisition of spatial harmonics [SMASH], unaliasing by Fourier encoding of the overlaps using the temporal dimension [UNFOLD], k-t-broad-use linear acquisition speed-up technique [k-t-BLAST], k-t-SENSE (structural details of arrays of sub-coils G01R33/3415) · CPC title