System and method for synthesizing magnetic resonance images
Methods and systems for synthesizing contrast images from a quantitative acquisition are disclosed. An exemplary method includes performing a quantification scan, using a trained deep neural network to synthesize a contrast image from the quantification scan, and outputting the contrast image synthesized by the trained deep neural network. In another exemplary method, an operator can identify a target contrast type for the synthesized contrast image. A trained discriminator and classifier module determines whether the synthesized contrast image is of realistic image quality and whether the synthesized contrast image matches the target contrast type.
1. A method for synthesizing a magnetic resonance (MR) contrast image, the method comprising:
performing a quantification scan, wherein the quantification scan is the measurement of MR signals reflecting absolute values of physical parameters of tissues being scanned;
using a trained deep neural network to synthesize the MR contrast image from a quantitative acquisition obtained by the quantification scan;
outputting the MR contrast image synthesized by the trained deep neural network; and
wherein the quantification scan is the input to the trained deep neural network.
2. The method of claim 1 , wherein performing the quantification scan comprises acquiring MR signals with a multi-delay multi-echo (MDME) sequence.
3. The method of claim 1 , wherein the quantitative acquisition is multiple complex images reconstructed from MR signals acquired by the quantification scan.
4. The method of claim 1 , wherein the MR contrast image synthesized by the trained deep neural network is one of T1-weighted image, T-2 weighted image, T1-FLAIR image, T2-FLAIR image, proton density weighted image, and STIR image.
5. The method of claim 1 , wherein the MR contrast image synthesized by the trained deep neural network is a T2*-weighted image.
6. The method of claim 1 , further comprising:
receiving, at the trained deep neural network, a target contrast type for the synthesized MR contrast image.
7. The method of claim 6 , further comprising:
using a trained discriminator to determine whether the synthesized MR contrast image is of realistic image quality; and
in response to determining that the synthesized MR contrast image is not of realistic image quality, outputting an indication to indicate that.
8. The method of claim 6 , further comprising:
using a trained classifier to determine the contrast type of the synthesized MR contrast image;
determining whether the determined contrast type matches the target contrast type; and
in response to determining that the determined contrast type does not match the target contrast type, outputting an indication to indicate that.
9. A magnetic resonance imaging (MRI) system comprising:
a gradient coil assembly configured to generate gradient magnetic fields;
a radio frequency (RF) coil assembly configured to generate RF pulses;
a display;
a controller in communication with the gradient coil assembly, the RF coil assembly, and the display and configured to:
instruct the gradient coil assembly and RF assembly to generate a sequence to perform a quantification scan, wherein the quantification scan is the measurement of MR signals reflecting absolute values of physical parameters of tissues being scanned;
instruct a trained deep neural network to synthesize a MR contrast image from a quantitative acquisition obtained by the quantification scan;
output the MR contrast image synthesized by the trained deep neural network at the display; and
wherein the quantification scan is the input to the trained deep neural network.
10. The MRI system of claim 9 , wherein the sequence for the quantification scan is an MDME sequence.
11. The MRI system of claim 9 , wherein the target contrast type is any of T1-weighted, T2-weighted, T2*-weighted image, T1-FLAIR, T2-FLAIR, proton density weighted, and STIR.
12. The MRI system of claim 9 , further comprising a memory storing the trained deep neural network.
13. The MRI system of claim 12 , wherein the trained deep neural network is a multi-scale U-Net convolutional neural network (CNN) with an encoder-decoder structure.
14. A method for synthesizing a magnetic resonance (MR) contrast image, the method comprising:
performing a quantification scan, wherein the quantification scan is the measurement of MR signals reflecting absolute values of physical parameters of tissues being scanned;
receiving a target contrast type;
using a trained deep neural network to synthesize the MR contrast image from a quantitative acquisition obtained by the quantification scan based on the target contrast type;
using a trained discriminator to determine whether the synthesized MR contrast image is of realistic image quality;
in response to determining that the synthesized MR contrast image is of realistic image quality, using a trained classifier to determine the contrast type of the synthesized MR contrast image;
determining whether the determined contrast type matches the target contrast type;
in response to determining that the determined contrast type matches the target contrast type, outputting the synthesized image; and
wherein the quantification scan is the input to the trained deep neural network.
15. The method of claim 14 , further comprising:
in response to determining that the synthesized MR contrast image is not of realistic image quality, outputting an indication to indicate that.
16. The method of claim 14 , further comprising:
in response to determining that the determined contrast type does not match the target contrast type, outputting an indication to indicate that.
17. The method of claim 14 , wherein the target contrast type is any of T1-weighted, T2-weighted, T2*-weighted image, T1-FLAIR, T2-FLAIR, proton density weighted, and STIR.
18. The method of claim 17 , wherein the trained deep neural network uses one set of parameters for all contrast types.