Method and system for accelerated acquisition and artifact reduction of undersampled MRI using a deep learning based 3D generative adversarial network
Systems and methods for generative adversarial networks (GANs) to remove artifacts from undersampled magnetic resonance (MR) images are described. The process of training the GAN can include providing undersampled 3D MR images to the generator model, providing the generated example and a real example to the discriminator model, applying adversarial loss, L2 loss, and structural similarity index measure loss to the generator model based on a classification output by the discriminator model, and repeating until the generator model has been trained to remove the artifacts from the undersampled 3D MR images. At runtime, the trained generator model of the GAN can be generate artifact-free images or parameter maps from undersampled MRI data of a patient.
1. A computer-implemented method for training a generator model of a generative adversarial network (GAN) to remove artifacts from undersampled magnetic resonance imaging (MRI) data, the GAN comprising the generator model and a discriminator model, the method comprising:
(a) obtaining, by a computer system, the undersampled 3D MRI data from fully-sampled 3D MRI data;
(b) providing, by the computer system, the undersampled 3D MRI data to the generator model, wherein the generator model is configured to output a generated example in response thereto;
(c) providing, by the computer system, the generated example and a real example based on the fully-sampled 3D MRI data to the discriminator model;
(d) applying, by the computer system, adversarial loss, L2 loss, and structural similarity index measure loss to the generator model based on a classification output by the discriminator model;
(e) repeating, by the computer system, (a)-(d) until the generator model has been trained to remove the artifacts from the undersampled 3D MRI data;
(f) generating, by the computer system, using the trained generator model, a parameter map for a biomarker from an undersampled MRI data set associated with a patient.
2. The method of claim 1 , wherein the undersampled 3D MRI data were created via kspace data undersampling.
3. The method of claim 1 , wherein the artifacts comprise at least one of streaking, aliasing, or ghosting.
4. The method of claim 1 , wherein the undersampled 3D MRI data were obtained via a Cartesian trajectory.
5. The method of claim 1 , wherein the undersampled 3D MRI data were obtained via a non-Cartesian trajectory.
6. The method of claim 1 , further comprising:
generating, by the computer system, using the trained generator model, an image of a patient from an undersampled MR image of the patient.
7. The method of claim 1 , wherein the trained generator model is configured to reduce a mean value and a standard deviation of measured values of the biomarker to suppress artificial values for the biomarker.
8. A computer system for training a generator model of a generative adversarial network (GAN) to remove artifacts from undersampled magnetic resonance imaging (MRI) data, the GAN comprising the generator model and a discriminator model, the computer system comprising:
a processor; and
a memory coupled to the processor, the memory storing instructions that, when executed by the processor, cause the computer system to:
(a) obtain the undersampled 3D MRI data from fully-sampled 3D MRI data;
(b) provide the undersampled 3D MRI data to the generator model, wherein the generator model is configured to output a generated example in response thereto;
(c) provide the generated example and a real example based on the fully-sampled 3D MRI data to the discriminator model,
(d) apply adversarial loss, L2 loss, and structural similarity index measure loss to the generator model based on a classification output by the discriminator model;
(e) repeat (a)-(d) until the generator model has been trained to remove the artifacts from the undersampled 3D MRI data; and
(f) generate, using the trained generator model, a parameter map for a biomarker from an undersampled MRI data set associated with a patient.
9. The computer system of claim 8 , wherein the undersampled 3D MRI data were created via k-space data undersampling.
10. The computer system of claim 8 , wherein the artifacts comprise at least one of streaking, aliasing, or ghosting.
11. The computer system of claim 8 , wherein the undersampled 3D MRI data were obtained via a Cartesian trajectory.
12. The computer system of claim 8 , wherein the undersampled 3D MRI data were obtained via a non-Cartesian trajectory.
13. The computer system of claim 8 , wherein the memory stores further instructions that, when executed by the processor, cause the computer system to:
generate, using the trained generator model, an image of a patient from an undersampled MR image of the patient.
14. A non-transitory medium storing program instructions for training a generator model of a generative adversarial network (GAN) to remove artifacts from undersampled magnetic resonance imaging (MRI) data, the GAN comprising the generator model and a discriminator model, the program instructions executable by a processing apparatus to:
a) obtain the undersampled 3D MRI data from fully-sampled 3D MRI data;
b) provide the undersampled 3D MRI data to the generator model, wherein the generator model is configured to output a generated example in response thereto;
c) provide the generated example and a real example based on the fully-sampled 3D MRI data to the discriminator model,
d) apply adversarial loss, L2 loss, and structural similarity index measure loss to the generator model based on a classification output by the discriminator model;
(e) repeat (a) (d) until the generator model has been trained to remove the artifacts from the undersampled 3D MRI data; and
(f) generate, using the trained generator model, a parameter map for a biomarker from an undersampled MRI data set associated with a patient.
15. The medium of claim 14 , wherein the undersampled 3D MRI data were created via k-space data undersampling.
16. The medium of claim 14 , wherein the undersampled 3D MRI data were obtained via a non-Cartesian trajectory.
17. The medium of claim 14 , the program instructions executable by a processing apparatus to:
generate, using the trained generator model, an image of a patient from an undersampled MR image of the patient.