Deep learning method for nonstationary image artifact correction

1. A method for magnetic resonance imaging that corrects non-stationary off-resonance image artifacts, the method comprising:

(a) performing by a magnetic resonance imaging (MRI) apparatus an MRI imaging acquisition using non-Cartesian trajectories within a field of view of the MRI apparatus to produce imaging data acquired using the non-Cartesian trajectories; and

(b) processing by the MRI apparatus the imaging data acquired using the non-Cartesian trajectories to produce a final MRI image; wherein the processing comprises:

i. reconstructing from the imaging data acquired using the non-Cartesian trajectories a complex-valued MRI image and

ii. using a convolutional neural network (CNN) to correct for non-stationary off-resonance artifacts in the complex-valued MRI image to produce the corrected complex-valued MRI image,

wherein an input to the CNN is the complex-valued MRI image and an output of the CNN is the corrected complex-valued MRI image;

iii. wherein the CNN is trained using a set of training MRI images comprising reference MRI images with off-resonance artifacts corrected using multifrequency autofocusing and input data generated by augmenting the reference MRI images with simulated zero-order off-resonance artifacts.

2. A method for magnetic resonance imaging (MRI) that corrects non-stationary off-resonance image artifacts, the method comprising:

(a) training a convolutional neural network (CNN) using a set of training MRI images, wherein the set of training MRI images comprises reference MRI images with off-resonance artifacts corrected using multifrequency autofocusing, and input data generated by augmenting the reference MRI images with simulated zero-order off-resonance artifacts;

(b) performing by a magnetic resonance imaging (MRI) apparatus an MRI imaging acquisition using non-Cartesian trajectories within a field of view of the MRI apparatus to produce imaging data acquired using the non-Cartesian trajectories;

(c) reconstructing by the MRI apparatus a complex-valued MRI image from the imaging data acquired using the non-Cartesian trajectories, and

(d) correcting non-stationary off-resonance artifacts in the complex-valued MRI image to produce a corrected complex-valued MRI image by inputting the complex-valued MRI image into the convolutional neural network (CNN) and obtaining the corrected complex-valued MRI image as an output of the CNN.

3. The method of claim 1 wherein the CNN is a residual network with multiple residual layers.

4. The method of claim 3 wherein the CNN comprises an input layer, followed by a 5×5×5 convolutional layer, followed by three consecutive residual layers, followed by an output layer, where each of the three consecutive residual layers comprises two 5×5×5 convolutional layers.

5. The method of claim 3 wherein an input layer of the residual network and an output layer of the residual network are complex-valued with the complex real and imaginary components split into two respective channels.

6. The method of claim 1 wherein the complex-valued MRI image input to the CNN has a non-zero real component and a zero imaginary component.

7. The method of claim 1 wherein the corrected complex-valued MRI image output of the CNN has a non-zero real component and a zero imaginary component.

8. The method of claim 1 wherein the processing comprises subtracting a complex-valued global mean from the complex-valued MRI image, and dividing the complex-valued MRI image by a global standard deviation.

9. The method of claim 1 wherein the complex-valued MRI image is 2D.

10. The method of claim 1 wherein the complex-valued MRI image is 3D.

11. The method of claim 1 wherein the non-Cartesian trajectory is a 2D spiral trajectory, a 2D radial trajectory, a 3D cones trajectory, or a 3D radial trajectory.

12. The method of claim 1 wherein performing the MRI imaging acquisition comprises using a gradient-echo sequence, a spoiled gradient-echo sequence, or a steady-state free precession sequence.