System and method for iteratively calibrated reconstruction kernel for accelerated magnetic resonance imaging
A method for iteratively calibrating a reconstruction kernel for use in accelerated magnetic resonance imaging (MRI) is provided. An MRI system is used to acquire k-space data from multiple slice locations following the application of a multiband radio frequency (RF) excitation pulse. An initial reconstruction kernel is generated from the acquired k-space data, and this initial reconstruction kernel is used to produce an initial image for each of the multiple slice locations by applying the initial reconstruction kernel to the acquired k-space data. The average phase of each slice location is then calculated from these images, and used to shift the phase values of the subsequently acquired k-space data. From the phase-shifted k-space data, an updated reconstruction kernel is then generated. This process is repeated iteratively until a stopping criterion is satisfied.
1. A method for reconstructing an image using a magnetic resonance imaging (MRI) system, the steps of the method comprising:
a) acquiring, using the MRI system, first k-space data from a plurality of different slice locations, wherein spins in each slice location are excited using a single band radio frequency (RF) excitation pulse;
b) acquiring, using the MRI system, second k-space data from the plurality of different slice locations, wherein spins in the plurality of different slice locations are excited using a multiband RF excitation pulse;
c) generating an initial reconstruction kernel from the first k-space data acquired in step a);
d) producing an initial image for each of the multiple slice locations by applying the initial reconstruction kernel generated in step c) to the second k-space data acquired in step b);
e) producing an updated reconstruction kernel by deriving one of phase information and magnitude information from the initial images produced in step d), generating updated k-space data by applying the derived one of phase information and magnitude information to the first k-space data, and generating the updated reconstruction kernel from the updated k-space data; and
f) producing an image for each slice location by applying the updated reconstruction kernel to the second k-space data acquired in step b).
2. The method as recited in claim 1 , wherein step e) includes deriving phase information from the initial images and applying the phase information to the first k-space data to produce phase-shifted k-space data.
3. The method as recited in claim 2 , wherein the phase information includes an average phase value calculated for each slice location from the respective initial image for that slice location.
4. The method as recited in claim 3 , wherein calculating the average phase value includes averaging phase values over each channel in a multichannel receiver.
5. The method as recited in claim 4 , wherein calculating the average phase value includes averaging over each spatial frequency in addition to over each channel in the multichannel receiver.
6. The method as recited in claim 2 , wherein the phase information includes a spatially varying phase value calculated from the initial images.
7. The method as recited in claim 2 , wherein the phase-shifted k-space data is produced by producing phase-shifted images by multiplying the initial images produced in step d) by the phase information, and Fourier transforming the phase-shifted images into k-space.
8. The method as recited in claim 1 , wherein the second k-space data acquired in step b) is undersampled along a phase-encoding direction.
9. The method as recited in claim 1 , wherein steps d) and e) are repeated to iteratively update the reconstruction kernel until a stopping criterion is satisfied.
10. The method as recited in claim 1 , wherein the initial reconstruction kernel generated in step c) includes a first initial reconstruction kernel and a second initial reconstruction kernel, the first initial reconstruction kernel being associated with a first portion of k-space and the second initial reconstruction kernel being associated with a second portion of k-space that is different that the first portion of k-space.
11. The method as recited in claim 10 , wherein the first portion of k-space is associated points in k-space where a source point is matched and spatially aligned with a target point, and the second portion of k-space is associated with points in k-space where a source point is matched but not spatially aligned with a target point.