Systems and methods for magnetic resonance imaging standardization using deep learning
A computer-implemented method for transforming magnetic resonance (MR) imaging across multiple vendors is provided. The method comprises: obtaining a training dataset, wherein the training dataset comprises a paired dataset and an un-paired dataset, and wherein the training dataset comprises image data acquired using two or more MR imaging devices; training a deep network model using the training dataset; obtaining an input MR image; and transforming the input MR image to a target image style using the deep network model.
1. A computer-implemented method for standardizing medical imaging, the method comprising:
acquiring one or more input images using one or more imaging devices, wherein the one or more imaging devices are of a same imaging modality and the one or more imaging devices correspond to one or more different image styles;
determining a standardized image style for processing the one or more input images;
predicting a synthesized output image by processing the one or more input images using a deep network model, wherein the synthesized output image has the standardized image style; and
displaying the synthesized output image with the standardized image style on a display device.
2. The computer-implemented method of claim 1 , wherein the standardized image style is determined by a user via a graphical user interface (GUI) rendered on the display device.
3. The computer-implemented method of claim 2 , wherein the standardized image style is determined by the user selecting an image style from a list of image styles.
4. The computer-implemented method of claim 1 , wherein the standardized image style is determined automatically based at least in part on the one or more input images.
5. The computer-implemented method of claim 1 , wherein the standardized image style comprises one or more characteristics including contrast, resolution, size, color, or image distortion.
6. The computer-implemented method of claim 1 , wherein the one or more imaging devices are magnetic resonance (MR) devices and are provided by one or more different vendors.
7. The computer-implemented method of claim 1 , wherein the one or more different image styles are different in at least one of contrast, resolution and image distortion.
8. The computer-implemented method of claim 1 , wherein the deep network model is trained using a combination of unsupervised learning and supervised learning.
9. The computer-implemented method of claim 8 , wherein the combination of unsupervised learning and supervised learning comprises training the deep network model using supervised training approach and further enhancing the deep network model using an unsupervised training approach.
10. The computer-implemented method of claim 8 , wherein the combination of unsupervised learning and supervised learning comprises using a combination of supervised loss and unsupervised loss.
11. A non-transitory computer-readable storage medium including instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
acquiring one or more input images using one or more imaging devices, wherein the one or more imaging devices are of a same imaging modality and the one or more imaging devices correspond to one or more different image styles;
determining a standardized image style for processing the one or more input images;
predicting a synthesized output image by processing the one or more input images using a deep network model, wherein the synthesized output image has the standardized image style; and
displaying the synthesized output image with the standardized image style on a display device.
12. The non-transitory computer-readable storage medium of claim 11 , wherein the standardized image style is determined by a user via a graphical user interface (GUI) rendered on the display device.
13. The non-transitory computer-readable storage medium of claim 12 , wherein the standardized image style is determined by the user selecting an image style from a list of image styles.
14. The non-transitory computer-readable storage medium of claim 11 , wherein the standardized image style is determined automatically based at least in part on the one or more input images.
15. The non-transitory computer-readable storage medium of claim 11 , wherein the standardized image style comprises one or more characteristics including contrast, resolution, size, color, or image distortion.
16. The non-transitory computer-readable storage medium of claim 11 , wherein the one or more imaging devices are magnetic resonance (MR) devices and are provided by one or more different vendors.
17. The non-transitory computer-readable storage medium of claim 11 , wherein the one or more different image styles are different in at least one of contrast, resolution and image distortion.
18. The non-transitory computer-readable storage medium of claim 11 , wherein the deep network model is trained using a combination of unsupervised learning and supervised learning.
19. The non-transitory computer-readable storage medium of claim 18 , wherein the combination of unsupervised learning and supervised learning comprises training the deep network model using supervised training approach and further enhancing the deep network model using an unsupervised training approach.
20. The non-transitory computer-readable storage medium of claim 18 , wherein the combination of unsupervised learning and supervised learning comprises using a combination of supervised loss and unsupervised loss.