Image reconstruction method
A computer-implemented method of reconstructing a motion-compensated magnetic resonance image uses raw k-space data acquired at a first resolution over successive respiratory and/or cardiac cycles of a patient. After binning data based on corresponding motion states derived from these cycles, the resolution of the binned K-space data in each bin is reduced. This is done by selecting a sub-group of binned k-space data. Bin images are reconstructed from the reduced-resolution data, and histogram-equalised versions of the reconstructed reduced-resolution bin image generated for each bin. Motion fields are estimated and interpolated to the first resolution such that motion data can be incorporated into a final reconstruction of a motion compensated image.
1. A method of reconstructing a motion-compensated magnetic resonance image of a patient, the method comprising:
acquiring, at a first resolution, initial raw k-space data of a magnetic resonance image of a patient over at least one of successive respiratory or successive cardiac cycles of the patient;
grouping the initial raw k-space data, or a subset of the initial raw k-space data, into one or more bins, wherein each of the one or more bins represents a motion state at a specific point in at least one of a respiratory or cardiac cycle of the patient;
selecting a sub-group of binned k-space data from each of the one or more bins and reconstructing a reduced-resolution bin image from the sub-group of binned k-space data for each of the one or more bins, the one or more reduced-resolution bin images having a reduced resolution compared to the first resolution;
generating histogram-equalized versions of the one or more reduced-resolution bin images, reconstructed for the one or more bins from the sub-group of binned k-space data, and estimating motion between the histogram-equalized versions in a form of motion fields; and
incorporating the motion fields into a final reconstruction of the motion-compensated magnetic resonance image of the patient, taking into account at least a subset of the initial raw k-space data.
2. The method as claimed in claim 1 , wherein the acquiring initial raw k-space data of the magnetic resonance image comprises:
acquiring images during an acquisition window timed using an electrocardiogram of the patient.
3. The method as claimed in claim 1 , wherein the initial raw k-space data of the magnetic resonance image of the patient is one of a series of images that require registration with each other to create a final magnetic resonance image of a patient.
4. The method as claimed in claim 1 , wherein the selecting a sub-group of binned k-space data comprises:
resampling the binned k-space data to reduce a resolution of the binned k-space data by a percentage based on the first resolution of the initial raw k-space data.
5. The method as claimed in claim 1 , wherein the selecting a sub-group of binned k-space data comprises:
resampling the binned k-space data to reduce a resolution of the binned k-space data to a minimum resolution at which physiological features of interest are resolvable.
6. The method as claimed in claim 1 , wherein the selecting a sub-group of binned k-space data comprises:
resampling the binned k-space data to reduce a resolution by rejecting or zero-padding a variable number high-frequency samples in the binned k-space data.
7. The method as claimed in claim 1 , wherein the generating histogram-equalized versions of the one or more reduced-resolution bin images comprises:
increasing a global contrast of the one or more reduced-resolution bin images.
8. The method as claimed in claim 1 , further comprising:
forming a motion field for each of the one or more bins by distorting each of the one or more reduced-resolution bin images onto a reduced-resolution reconstructed bin image generated from binned k-space data in an adjacent bin to create a vector field representing the distortion.
9. The method as claimed in claim 1 , wherein reconstructing the one or more reduced-resolution bin images further comprises:
taking into account estimates of translation motion between the one or more bins.
10. The method as claimed in claim 1 , wherein the incorporating the motion fields comprises:
interpolating the motion fields to the first resolution.
11. The method as claimed in claim 1 , wherein the initial raw k-space data is undersampled.
12. The method as claimed in claim 1 , wherein the initial raw k-space data is acquired using a variable density trajectory imaging method, such that the sub-groups of binned k-space data, from which the one or more reduced-resolution bin images for the one or more bins are reconstructed, exhibit a lower relative undersampling than the first resolution of the initial raw k-space data.
13. The method as claimed in claim 1 , wherein the initial raw k-space data comprises a residual unsuppressed fat signal.
14. The method as claimed in claim 1 , wherein the initial raw k-space data is acquired using Dixon imaging.
15. The method as claimed in claim 1 , wherein the motion-compensated magnetic resonance image of the patient contains at least one of cardiac or vascular information.
16. A non-transitory computer-readable storage medium storing computer-executable instructions that, when executed by a computer, cause the computer to carry out the method of claim 1 .
17. A data processing apparatus adapted to reconstruct a motion-compensated magnetic resonance image of a patient, the data processing apparatus comprising:
an imaging device configured to acquire, at a first resolution, initial raw k-space data of a magnetic resonance image of a patient over at least one of successive respiratory or successive cardiac cycles of the patient; and
a processor configured to
group the at least a subset of the initial raw k-space data into one or more bins, wherein each of the one or more bins represents a motion state at a specific point in at least one of a respiratory or cardiac cycle of the patient,
select a sub-group of binned k-space data from each of the one or more bins and reconstruct a reduced-resolution bin image from the selected sub-group of binned k-space data for each of the one or more bins, the one or more reduced-resolution bin images having a reduced resolution compared to the first resolution,
generate histogram-equalized versions of the one or more reduced-resolution bin images, reconstructed for the one or more bins from the sub-group of binned k-space data, and estimate motion between the histogram-equalized versions in a form of motion fields, and
incorporate the motion fields into a final reconstruction of the motion-compensated magnetic resonance image of the patient, taking into account all, or a subset, of the initial raw k-space data.
18. The data processing apparatus as claimed in claim 17 , wherein the initial raw k-space data of the magnetic resonance image of the patient is one of a series of images that require registration with each other to create a final magnetic resonance image of a patient.
19. The data processing apparatus as claimed in claim 17 , wherein the processor is configured to select the sub-group of binned k-space data by resampling the binned k-space data to reduce a resolution of the binned k-space data by a percentage based on the first resolution of the initial raw k-space data.
20. The data processing apparatus as claimed in claim 17 , wherein the processor is configured to select the sub-group of binned k-space data by (i) resampling the binned k-space data to reduce a resolution of the binned k-space data to a minimum resolution at which physiological features of interest are resolvable, or (ii) resampling the binned k-space data to reduce the resolution by rejecting or zero-padding a variable number high-frequency samples in the binned k-space data.