Method for generating a magnetic resonance image

1. A method for generating a magnetic resonance image (I) of an object region, the magnetic resonance image having N×M image voxels, wherein N×M denotes a number of sampling steps in a fully sampled k-space, said method comprising:

a) providing at least n magnetic resonance (MR) segment data records (S 1 , S 2 , . . . , Sn), wherein n≥1, wherein each MR segment data record has N×M frequency voxels in k-space, each MR segment data record contains MR signal data for q frequency voxels and wherein, in each case, q<N×M and different MR segment data records contain MR signal data for different combinations of frequency voxels;

b) calculating a first reconstruction kernel (K 1 ) from a reference spectrum (R) of the object region;

c) calculating n preliminarily reconstructed MR segment data records (S 1 . 1 , S 1 . 2 , . . . , S 1 . n ) by reconstructing data missing in a respective MR segment data record (S 1 , S 2 , . . . , Sn) within a k-space region specified for each of the MR segment data records (S 1 , S 2 , . . . , Sn) with the first reconstruction kernel (K 1 );

d) calculating respective phase images (P 1 , P 2 , . . . , Pn) from the preliminarily reconstructed MR segment data records (S 1 . 1 , S 1 . 2 , . . . , S 1 . n ) with a Fourier transform;

e) calculating a reference image (IR), which has an absolute value, from the reference spectrum (R) with a Fourier transform;

f) calculating n modified reference images (IR 1 , IR 2 , . . . , IRn) by combining the absolute value of the reference image (IR) respectively with respective phases of each of the phase images (P 1 , P 2 , . . . , Pn);

g) calculating n modified reference spectra (R 1 , R 2 , . . . , Rn) with a Fourier transform of each of the modified reference images (IR 1 , IR 2 , . . . , IRn);

h) calculating n second reconstruction kernels (K 2 . 1 , K 2 . 2 , . . . , K 2 . n ), wherein each second reconstruction kernel is ascertained from all modified reference spectra (R 1 , R 2 , . . . , Rn) for each of the MR segment data records;

i) ascertaining at least one finally reconstructed MR segment data record (S 2 ; S 2 . 1 , S 2 . 2 , . . . , S 2 . n ) by reconstructing missing MR signal data of the MR segment data records (S 1 , S 2 , . . . , Sn) with the second reconstruction kernels (K 2 . 1 , K 2 . 2 , . . . , K 2 . n ) and the MR signal data of the n MR segment data records (S 1 , S 2 , . . . , Sn); and

j) calculating the magnetic resonance image (I) from the at least one finally reconstructed MR segment data record (S 2 ; S 2 . 1 , S 2 . 2 , . . . , S 2 . n ).

2. The method as claimed in claim 1 , wherein

the calculation of the n second reconstruction kernels (K 2 . 1 , K 2 . 2 , . . . , K 2 . n ) is carried out with the modified reference spectra (R 1 , R 2 , . . . , Rn) and additionally with a selected reference spectrum (R*) in said step h);

one single finally reconstructed MR segment data record (S 2 ) is ascertained in said step i), and the magnetic resonance image (I) is generated in said step j) by a Fourier transform of the finally reconstructed MR segment data record (S 2 ).

3. The method as claimed in claim 2 , wherein

the selected reference spectrum (R*) is the reference spectrum (R) or one of the modified reference spectra (R 1 , R 2 , . . . , Rn).

4. The method as claimed in claim 1 , wherein

n finally reconstructed MR segment data records (S 2 ; S 2 . 1 , S 2 . 2 , . . . , S 2 . n ) are ascertained in said step i), and

a segment image (I 1 , I 2 , . . . , In) is initially generated in said step j) by a Fourier transform for each finally reconstructed MR segment data record (S 2 ; S 2 . 1 , S 2 . 2 , . . . , S 2 . n ) and the segment images (I 1 , I 2 , . . . , In) are subsequently combined to form the magnetic resonance image (I).

5. The method as claimed in claim 1 , wherein said providing of the n MR segment data records (S 1 , S 2 , . . . , Sn) in said step a) is implemented by recording at least one MR spectrum (S), wherein each MR spectrum (S) is recorded by scanning through the MR spectrum (S) with an excitation pulse in an MR sequence, in order to excite nuclear spins within the object region.

6. The method as claimed in claim 5 , wherein

the MR spectrum (S) is recorded in said step a), and

the recorded MR spectrum (S) is stored and is subdivided into and provided as the MR segment data records (S 1 , S 2 , . . . , Sn).

7. The method as claimed in claim 6 , wherein

an Echo-Planar Imaging (EPI) sequence is used as the MR sequence and

the subdivision of the MR segment data records is implemented by subdividing MR signals in accordance with a sampling direction in k-space of the MR signals.

8. The method as claimed in claim 5 , wherein a plurality of MR spectra are recorded with signal data for in each case less than M×N frequency voxels, and wherein different encodings are used for each of a plurality of excitation pulses.

9. The method as claimed in claim 1 , wherein the specified k-space region in said step c) comprises all N×M frequency voxels.

10. The method as claimed in claim 1 , wherein the specified k-space region in said step c) comprises a low-frequency portion of k-space.

11. The method as claimed in claim 1 , wherein the specified k-space region in said step c) comprises portions of k-space in which the MR segment data record corresponding to the specified k-space has peaks.

12. The method as claimed in claim 1 , wherein a movement of the object region between the recording of individual ones of the MR segment data records (S 1 , S 2 , . . . , Sn) is corrected with the preliminarily reconstructed MR segment data records (S 1 . 1 , S 1 . 2 , . . . , S 1 . n ).

13. The method as claimed in claim 1 , wherein the phase images (P 1 -Pn) are smoothed in said step d) with an x-space/image-space filter and/or wherein the preliminarily reconstructed MR segment data records (S 1 . 1 -S 1 . n ) are smoothed in said step d) with a k-space filter.

14. The method as claimed in claim 6 , wherein the MR spectrum (S) is recorded with signal data for N×M frequency voxels.

15. The method as claimed in claim 8 , wherein n MR spectra are recorded with signal data for in each case less than M×N frequency voxels.

16. A method for generating a magnetic resonance image (I) of an object region, the magnetic resonance image having N×M image voxels, wherein N×M denotes a number of sampling steps in a fully sampled k-space, said method comprising:

aa) providing at least n magnetic resonance (MR) segment data records (S 1 , S 2 , . . . , Sn), wherein n> 1 , wherein each MR segment data record has N×M frequency voxels in k-space, each MR segment data record contains MR signal data for q frequency voxels and wherein, in each case, q<N×M and different MR segment data records contain MR signal data for different combinations of frequency voxels;

bb) calculating a first reconstruction kernel (K 1 ) from a reference spectrum (R) of the object region;

cc) calculating n preliminarily reconstructed MR segment data records (S 1 . 1 , S 1 . 2 , . . . , S 1 .n) by reconstructing data missing in a respective MR segment data record (S 1 , S 2 , . . . , Sn) within a k-space region specified for each of the MR segment data records (S 1 , S 2 , . . . , Sn) with the first reconstruction kernel (K 1 );

dd) calculating respective phase images (P 1 , P 2 , . . . , Pn) from the preliminarily reconstructed MR segment data records (S 1 . 1 , S 1 . 2 , . . . , S 1 .n) with a Fourier transform;

ee) calculating a reference image (IR), which has an absolute value, from the reference spectrum (R) with a Fourier transform;

ff) calculating n modified reference images (IR 1 , IR 2 , . . . , IRn) by combining the absolute value of the reference image (IR) respectively with respective phases of each of the phase images (P 1 , P 2 , . . . , Pn);

gg) calculating n modified reference spectra (R 1 , R 2 , . . . , Rn) with a Fourier transform of each of the modified reference images (IR 1 , IR 2 , . . . , IRn);

hh) calculating n second reconstruction kernels (K 2 . 1 , K 2 . 2 , . . . , K 2 .n), wherein each second reconstruction kernel is ascertained from all modified reference spectra (R 1 , R 2 , . . . , Rn) for each of the MR segment data records;

ii) ascertaining at least one finally reconstructed MR segment data record (S 2 ; S 2 . 1 , S 2 . 2 , . . . , S 2 .n) by reconstructing missing MR signal data of the MR segment data records (S 1 , S 2 , . . . , Sn) with the second reconstruction kernels (K 2 . 1 , K 2 . 2 , . . . , K 2 .n) and the MR signal data of the n MR segment data records (S 1 , S 2 , . . . , Sn);

jj) calculating new respective phase images (P 1 . 2 , P 2 . 2 , . . . , Pn. 2 ) from the finally reconstructed MR segment data records (S 2 ; S 2 . 1 , S 2 . 2 , . . . , S 2 .n) with a Fourier transform;

kk) repeating said steps ee) to ii) factoring in the new respective phase images (P 1 . 2 , P 2 . 2 , . . . Pn. 2 );

ll) calculating the magnetic resonance image (I) from the at least one finally reconstructed MR segment data record (S 2 ; S 2 . 1 , S 2 . 2 , . . . , S 2 .n) factoring in the new respective phase images (P 1 . 2 , P 2 . 2 , . . . Pn. 2 ).