Magnetic resonance imaging with deep neutral networks

1. A method of magnetic resonance imaging, the method comprising:

scanning a subject with a first undersampling factor and a first sampling trajectory to obtain first imaging information;

processing the first imaging information with a deep neural network (DNN) to obtain second imaging information corresponding to a second undersampling factor, the first undersampling factor being larger than the second undersampling factor; and

reconstructing a magnetic resonance image of the subject from the second imaging information,

wherein the DNN is trained with a plurality of training samples before processing the first imaging information with the DNN, and

wherein each training sample of the plurality of training samples is obtained by

scanning a subject sample with the second undersampling factor to obtain third imaging information,

selecting a corresponding part of information from the third imaging information according to the first undersampling factor and a second sampling trajectory to obtain fourth imaging information, and

taking the third imaging information as an output sample and the fourth imaging information as an input sample to form the training sample.

2. The method of claim 1 , wherein reconstructing the magnetic resonance image of the subject from the second imaging information comprises:

reconstructing the magnetic resonance image from the second imaging information with one of parallel imaging accelerated imaging, compressed sensing accelerated imaging, and half-Fourier accelerated imaging.

3. The method of claim 1 , wherein, when the magnetic resonance image of the subject is reconstructed from the second imaging information with parallel imaging accelerated imaging, the first sampling trajectory is same as the second sampling trajectory.

4. The method of claim 1 , wherein, when the magnetic resonance image of the subject is reconstructed from the second imaging information with half-Fourier accelerated imaging, the first sampling trajectory is same as the second sampling trajectory.

5. The method according to claim 1 , wherein, when the magnetic resonance image of the subject is reconstructed from the second imaging information with compressed sensing accelerated imaging, both the first sampling trajectory and the second sampling trajectory satisfy a sparse constraint condition of the compressed sensing accelerated imaging.

6. The method of claim 1 , wherein the first imaging information and the second imaging information are parts of k-space data in a k-space domain.

7. The method of claim 1 , wherein the first imaging information and the second imaging information are images obtained by performing a Fourier transform on parts of k-space data in a k-space domain.

8. The method of claim 1 , wherein the second undersampling factor is no more than an undersampling factor for accelerated imaging.

9. An apparatus for magnetic resonance imaging, the apparatus comprising:

at least one processor; and

at least one non-transitory machine readable storage medium coupled to the at least one processor having machine-executable instructions stored thereon that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:

controlling a coil of a magnetic resonance imaging apparatus to scan a subject with a first undersampling factor and a first sampling trajectory to obtain first imaging information;

processing the first imaging information with a deep neural network (DNN) to obtain second imaging information corresponding to a second undersampling factor, the first undersampling factor being larger than the second undersampling factor; and

reconstructing a magnetic resonance image of the subject from the second imaging information,

wherein the DNN is trained with a plurality of training samples before processing the first imaging information with the DNN, and

wherein each training sample of the plurality of training samples is obtained by

scanning a subject sample with the second undersampling factor to obtain third imaging information,

selecting a corresponding part of information from the third imaging information according to the first undersampling factor and a second sampling trajectory to obtain fourth imaging information, and

taking the third imaging information as an output sample and the fourth imaging information as an input sample to form the training sample.

10. The apparatus of claim 9 , wherein reconstructing the magnetic resonance image of the subject from the second imaging information comprises:

reconstructing the magnetic resonance image from the second imaging information with one of parallel imaging accelerated imaging, compressed sensing accelerated imaging, and half-Fourier accelerated imaging.

11. The apparatus of claim 9 , wherein, when the magnetic resonance image of the subject is reconstructed from the second imaging information with parallel imaging accelerated imaging, the first sampling trajectory is same as the second sampling trajectory.

12. The apparatus of claim 9 , wherein, when the magnetic resonance image of the subject is reconstructed from the second imaging information with half-Fourier accelerated imaging, the first sampling trajectory is same as the second sampling trajectory.

13. The apparatus of claim 9 , wherein, when the magnetic resonance image of the subject is reconstructed from the second imaging information with compressed sensing accelerated imaging, both the first sampling trajectory and the second sampling trajectory satisfy a sparse constraint condition of the compressed sensing accelerated imaging.

14. The apparatus of claim 9 , wherein the first imaging information and the second imaging information are parts of k-space data in a k-space domain.

15. The apparatus of claim 9 , wherein the first imaging information and the second imaging information are images obtained by performing a Fourier transform on parts of k-space data in a k-space domain.

16. The apparatus of claim 9 , wherein the second undersampling factor is no more than an undersampling factor for accelerated imaging.

17. A non-transitory machine readable storage medium having machine-executable instructions stored thereon that, when executed by at least one processor, cause the at least one processor to perform operations comprising:

controlling a coil of a magnetic resonance imaging apparatus to scan a subject with a first undersampling factor and a first sampling trajectory to obtain first imaging information;

processing the first imaging information with a deep neural network (DNN) to obtain second imaging information corresponding to a second undersampling factor, the first undersampling factor being larger than the second undersampling factor; and

reconstructing a magnetic resonance image of the subject from the second imaging information,

wherein the DNN is trained with a plurality of training samples before processing the first imaging information with the DNN, and

wherein each training sample of the plurality of training samples is obtained by

scanning a subject sample with the second undersampling factor to obtain third imaging information,

selecting a corresponding part of information from the third imaging information according to the first undersampling factor and a second sampling trajectory to obtain fourth imaging information, and

taking the third imaging information as an output sample and the fourth imaging information as an input sample to form the training sample.

18. The non-transitory machine readable storage medium of claim 17 , wherein reconstructing the magnetic resonance image of the subject from the second imaging information comprises:

reconstructing the magnetic resonance image from the second imaging information with one of parallel imaging accelerated imaging, compressed sensing accelerated imaging, and half-Fourier accelerated imaging.

19. The non-transitory machine readable storage medium of claim 17 , wherein:

when the magnetic resonance image of the subject is reconstructed from the second imaging information with parallel imaging accelerated imaging or half-Fourier accelerated imaging, the first sampling trajectory is same as the second sampling trajectory, or

when the magnetic resonance image of the subject is reconstructed from the second imaging information with compressed sensing accelerated imaging, both the first sampling trajectory and the second sampling trajectory satisfy a sparse constraint condition of the compressed sensing accelerated imaging.

20. The non-transitory machine readable storage medium of claim 17 , wherein the first imaging information and the second imaging information are one of:

parts of k-space data in a k-space domain, or

images obtained by performing a Fourier transform on parts of k-space data in a k-space domain.