Systems and methods of noise reduction in magnetic resonance images
A computer-implemented method of reducing noise in magnetic resonance (MR) images is provided. The method includes executing a neural network model of analyzing MR images, wherein the neural network model is trained with a pair of pristine images and corrupted images. The pristine images are the corrupted images with noise reduced, and target output images of the neural network model are the pristine images. The method also includes receiving first MR signals and second MR signals, reconstructing first and second MR images based on the first MR signals and the second MR signals, and analyzing the first MR image and the second MR image using the neural network model. The method further includes deriving a denoised MR image based on the analysis, wherein the denoised MR image is a combined image based on the first MR image and the second MR image and outputting the denoised MR image.
1. A computer-implemented method of reducing noise in magnetic resonance (MR) images, comprising:
executing a neural network model of analyzing MR images, wherein the neural network model is trained with a pair of pristine images and corrupted images, wherein the pristine images are the corrupted images with noise reduced, and target output images of the neural network model are the pristine images;
receiving first MR signals of a portion of a subject and second MR signals of the portion of the subject, the first MR signals and the second MR signals acquired using a pulse sequence of an MR system;
reconstructing a first MR image and a second MR image based on the first MR signals and the second MR signals, the first MR image and the second MR image being MR images of the portion of the subject;
analyzing the first MR image and the second MR image using the neural network model;
deriving a denoised MR image based on the analysis, wherein the denoised MR image is a combined image based on the first MR image and the second MR image; and
outputting the denoised MR image.
2. The method of claim 1 , wherein executing a neural network model further comprises executing the neural network model, wherein an input of the neural network model includes a complex image.
3. The method of claim 1 , wherein the first MR image has a different noise distribution from the second MR image.
4. The method of claim 1 , wherein receiving first MR signals and second MR signals further comprises:
receiving the first MR signals and the second MR signals, wherein the first MR signals are signals of a first echo of a first chemical species and a second chemical species, the second MR signals are signals of a second echo of the first chemical species and the second chemical species, wherein the first echo and the second echo have different echo times.
5. The method of claim 4 , wherein the first echo is an in-phase echo of the first chemical species and the second chemical species, and the second echo is an out-of-phase echo of the first chemical species and the second chemical species.
6. The method of claim 4 , wherein the first chemical species comprises water, and the second chemical species comprises fat.
7. The method of claim 1 , wherein:
reconstructing a first MR image and a second MR image further comprises:
reconstructing the first MR image based on the first MR signals; and
reconstructing the second MR image based on the second MR signals;
analyzing the first MR image and the second MR image further comprises:
combining the first MR image and the second MR image into a third image; and
analyzing the third image using the neural network model; and
deriving a denoised MR image further comprises:
deriving the denoised MR image based on the analysis of the third image.
8. The method of claim 1 , wherein:
reconstructing a first MR image and a second MR image further comprises:
reconstructing the first MR image based on the first MR signals; and
reconstructing the second MR image based on the second MR signals;
analyzing the first MR image and the second MR image further comprises:
analyzing the first MR image using the neural network model to derive a first denoised MR image; and
analyzing the second MR image using the neural network model to derive a second denoised MR image; and
deriving a denoised MR image further comprises:
deriving the denoised MR image by combining the first denoised MR image and the second denoised MR image.
9. The method of claim 1 , wherein:
reconstructing a first MR image and a second MR image further comprises:
reconstructing the first MR image based on the first MR signals; and
reconstructing the second MR image based on the second MR signals;
analyzing the first MR image and the second MR image further comprises:
jointly analyzing the first MR image and the second MR image using the neural network model; and
deriving a denoised MR image further comprises:
deriving the denoised MR image based on the joint analysis of the first MR image and the second MR image.
10. The method of claim 1 , wherein receiving first MR signals and second MR signals further comprises receiving first MR signals and second MR signals acquired with a multi-channel MR system.
11. The method of claim 1 , wherein the neural network model is configured to analyze three-dimensional (3D) MR datasets.
12. A magnetic resonance (MR) noise reduction system, comprising a noise reduction computing device, the noise reduction computing device comprising at least one processor in communication with at least one memory device, and the at least one processor programmed to:
execute a neural network model of analyzing MR images, wherein the neural network model is trained with a pair of pristine images and corrupted images, wherein the pristine images are the corrupted images with noise reduced, and target output images of the neural network model are the pristine images;
receive first MR signals of a portion of a subject and second MR signals of the portion of the subject, the first MR signals and the second MR signals acquired using a pulse sequence of an MR system;
reconstruct a first MR image and a second MR image based on the first MR signals and the second MR signals, the first MR image and the second MR image being MR images of the portion of the subject;
analyze the first MR image and the second MR image using the neural network model;
derive a denoised MR image based on the analysis, wherein the denoised MR image is a combined image based on the first MR image and the second MR image; and
output the denoised MR image.
13. The system of claim 12 , wherein the at least one processor is further programmed to:
execute the neural network model, wherein an input of the neural network model includes a complex image.
14. The system of claim 12 , wherein the first MR image has a different noise distribution from the second MR image.
15. The system of claim 12 , wherein the at least one processor is further programmed to:
receive the first MR signals and the second MR signals, wherein the first MR signals are signals of a first echo of a first chemical species and a second chemical species, the second MR signals are signals of a second echo of the first chemical species and the second chemical species, wherein the first echo and the second echo have different echo times.
16. The system of claim 15 , wherein the first echo is an in-phase echo of the first chemical species and the second chemical species, and the second echo is an out-of-phase echo of the first chemical species and the second chemical species.
17. The system of claim 15 , wherein the first chemical species comprises water, and the second chemical species comprises fat.
18. The system of claim 12 , wherein the at least one processor is further programmed to:
reconstruct the first MR image based on the first MR signals;
reconstruct the second MR image based on the second MR signals;
combine the first MR image and the second MR image into a third image;
analyze the third image using the neural network model; and
derive the denoised MR image based on the analysis of the third image.
19. The system of claim 12 , wherein the at least one processor is further programmed to:
reconstruct the first MR image based on the first MR signals;
reconstruct the second MR image based on the second MR signals;
analyze the first MR image using the neural network model to derive a first denoised MR image;
analyze the second MR image using the neural network model to derive a second denoised MR image; and
derive the denoised MR image by combining the first denoised MR image and the second denoised MR image.
20. The system of claim 12 , wherein the at least one processor is further programmed to:
jointly analyze the first MR image and the second MR image using the neural network model; and
derive the denoised MR image based on the joint analysis of the first MR image and the second MR image.