Systems and methods for synthetic medical image generation
Systems and methods for synthetic medical image generation in accordance with embodiments of the invention are illustrated. One embodiment includes a synthetic medical image generation system, including a processor, and a memory containing an image synthesis application, where the image synthesis application directs the processor to obtain source image data generated by at least one medical imaging device, where the source image data describes a functional medical image taken of a patient administered with a first imaging agent, and synthesize a predicted medical image of the patient that depicts the patient as if they were administered with a second imaging agent, wherein the first imaging agent and the second imaging agent are different imaging agents.
1. A synthetic medical image generation system, comprising:
a processor; and
a memory containing an image synthesis application, where the image synthesis application directs the processor to:
obtain source image data generated by at least one medical imaging device, where the source image data describes a medical image taken of a patient administered with a first real chemical imaging agent; and
synthesize a predicted medical image of the patient based on the source image data, where the predicted medical image depicts the patient as if they were administered with a second real chemical imaging agent, wherein the first real chemical imaging agent and the second real chemical imaging agent are different real chemical imaging agents.
2. The system of claim 1 , wherein the at least one medical imaging device is a positron emission tomography (PET) scanner.
3. The system of claim 1 , wherein the first real chemical imaging agent is selected from the group consisting of 18 F-FDOPA, 18 F-FLT, 18 F-MPG, and 18 F-FDG.
4. The system of claim 1 , wherein the second real chemical imaging agent is selected from the group consisting of 18 F-FDOPA, 18 F-FLT, 18 F-MPG, and 18 F-FDG.
5. The system of claim 1 , wherein to synthesize a predicted medical image, the image synthesis application directs the processer to utilize a neural network.
6. The system of claim 5 , wherein the neural network is a generative adversarial network (GAN), comprising a generator and a discriminator.
7. The system of claim 6 , wherein the generator is implemented using a U-Net architecture.
8. The system of claim 6 , wherein the discriminator is implemented using a PatchGAN architecture capable of processing 3D image volumes.
9. The system of claim 1 , wherein the source image data further comprises an anatomical image.
10. The system of claim 1 , wherein the image synthesis application further directs the processor to generate at least one mask based on the source image data for use in synthesizing the predicted medical image.
11. A method for generating synthetic medical images, comprising:
obtaining source image data generated by at least one medical imaging device, where the source image data describes a medical image taken of a patient administered with a first real chemical imaging agent; and
synthesizing a predicted medical image of the patient based on the source image data, where the predicted medical image depicts the patient as if they were administered with a second real chemical imaging agent, wherein the first real chemical imaging agent and the second real chemical imaging agent are different real chemical imaging agents.
12. The method of claim 11 , wherein the at least one medical imaging device is a positron emission tomography (PET) scanner.
13. The method of claim 11 , wherein the first real chemical imaging agent is selected from the group consisting of 18 F-FDOPA, 18 F-FLT, 18 F-MPG, and 18 F-FDG.
14. The method of claim 11 , wherein the second real chemical imaging agent is selected from the group consisting of 18 F-FDOPA, 18 F-FLT, 18 F-MPG, and 18 F-FDG.
15. The method of claim 11 , wherein synthesizing a predicted medical image comprises utilizing a neural network.
16. The method of claim 15 , wherein the neural network is a generative adversarial network (GAN), comprising a generator and a discriminator.
17. The method of claim 16 , wherein the generator is implemented using a U-Net architecture.
18. The method of claim 16 , wherein the discriminator is implemented using a PatchGAN architecture capable of processing 3D image volumes.
19. The method of claim 11 , wherein the source image data further comprises an anatomical image; and the method further comprises generating at least one mask based on the source image data for use in synthesizing the predicted medical image.
20. A synthetic medical image generation system, comprising:
a processor; and
a memory containing an image synthesis application, where the image synthesis application directs the processor to:
obtain source image data comprising a functional medical image generated by a positron emission tomography (PET) scanner and an anatomical image, where the functional medical image describes a medical image taken of a patient administered with a first imaging agent, where the first imaging agent and the second imaging agent are different imaging agents;
co-register the functional medical image with the anatomical image;
generate a brain mask based on the anatomical image;
generate a tumor mask by extracting features from the anatomical image and the functional medical image scan using a feature extractor neural network;
synthesize a predicted medical image of the patient that depicts the patient as if they were administered with a second imaging agent by providing a generative adversarial network (GAN) with the source image data and the tumor mask, where the GAN comprises:
a generator conforming to a U-Net architecture; and
a discriminator conforming to a PatchGAN architecture capable of processing 3D image volumes; and
provide the predicted medical image via a display.