Devices, systems, and methods for medical imaging

1. A method comprising:

obtaining scan data that were generated by scanning a scanned region,

wherein the scan data include groups of scan data,

wherein each group, of the groups of scan data, was captured at a respective angle, and

wherein each group's respective angle is different from every other group's respective angle;

performing a reconstruction process on a first subset of the groups of scan data, thereby generating a first partial reconstruction of at least part of the scanned region;

performing a reconstruction process on a second subset of the groups of scan data, thereby generating a second partial reconstruction of the at least part of the scanned region, wherein the first subset is different from the second subset;

inputting the first partial reconstruction and the second partial reconstruction into a machine-learning model, which generates one or more motion-compensated reconstructions of the at least part of the scanned region based at least on the first partial reconstruction and the second partial reconstruction;

calculating a respective edge entropy of each of the one or more motion-compensated reconstructions of the at least part of the scanned region; and

adjusting the machine-learning model based on the respective edge entropy of each of the one or more motion-compensated reconstructions.

2. The method of claim 1 , wherein the machine-learning model includes a neural network.

3. The method of claim 2 , wherein edge entropy of motion-compensated reconstructions is a loss function of the neural network.

4. The method of claim 1 , wherein the one or more motion-compensated reconstructions include a plurality of motion-compensated reconstructions, and

wherein the machine-learning model is further configured to calculate a respective motion of each pixel in the plurality of motion-compensated reconstructions based on the plurality of motion-compensated reconstructions.

5. The method of claim 4 , wherein the machine-learning model is further configured to generate a smoothed motion-compensated reconstruction of the scanned region based on the respective motion of each pixel and on the plurality of motion-compensated reconstructions.

6. The method of claim 5 , wherein generating the smoothed motion-compensated reconstruction includes warping the plurality of motion-compensated reconstructions.

7. The method of claim 5 , wherein the respective edge entropies are calculated based on the smoothed motion-compensated reconstruction of the scanned region.

8. The method of claim 1 , wherein the first subset and the second subset do not have any groups in common.

9. The method of claim 1 , wherein the first subset and the second subset have at least one group in common.

10. The method of claim 1 , further comprising:

performing a reconstruction process on a third subset of the groups of scan data, thereby generating a third partial reconstruction of the at least part of the scanned region;

performing a reconstruction process on a fourth subset of the groups of scan data, thereby generating a fourth partial reconstruction of the at least part of the scanned region,

wherein the first subset, the second subset, the third subset, and the fourth subset are all different from each other; and

inputting the third partial reconstruction and the fourth partial reconstruction into the machine-learning model, which generates the one or more motion-compensated reconstructions of the at least part of the scanned region further based at least on the third partial reconstruction and the fourth partial reconstruction.

11. A system comprising:

one or more processors; and

one or more computer-readable storage media in communication with the one or more processors, wherein the one or more processors and the one or more computer-readable storage media cooperate to perform operations including:

obtaining scan data that were generated by scanning a scanned region,

wherein the scan data include groups of scan data,

wherein each group, of the groups of scan data, was captured at a respective angle, and

wherein each group's respective angle is different from every other group's respective angle;

performing a reconstruction process on a first subset of the groups of scan data, thereby generating a first partial reconstruction of at least part of the scanned region;

performing a reconstruction process on a second subset of the groups of scan data, thereby generating a second partial reconstruction of the at least part of the scanned region, wherein the first subset is different from the second subset;

generating a motion-corrected reconstruction of the at least part of the scanned region, wherein generating the motion-correct reconstruction includes inputting the first partial reconstruction and the second partial reconstruction into a machine-learning model.

12. The system of claim 11 , wherein the machine-learning model outputs the motion-corrected reconstruction of the at least part of the scanned region.

13. The system of claim 11 , wherein the operations further include:

calculating an edge entropy of the motion-corrected reconstruction of the at least part of the scanned region; and

modifying the machine-learning model based on the edge entropy.

14. The system of claim 11 , wherein the machine-learning model outputs a respective motion-corrected partial reconstruction of each of the partial reconstructions, and

wherein the motion-corrected reconstruction is generated from the motion-corrected partial reconstructions.

15. The system of claim 11 , wherein the at least part of the scanned region includes part of an object of interest.

16. The system of claim 15 , wherein the object of interest is a cardiac vessel.

17. One or more computer-readable storage media storing instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations comprising:

obtaining scan data that were generated by scanning a scanned region,

wherein the scan data include groups of scan data,

wherein each group, of the groups of scan data, was captured at a respective angle, and

wherein each group's respective angle is different from every other group's respective angle;

performing a reconstruction process on a first subset of the groups of scan data, thereby generating a first partial reconstruction of at least part of the scanned region;

performing a reconstruction process on a second subset of the groups of scan data, thereby generating a second partial reconstruction of the at least part of the scanned region, wherein the first subset is different from the second subset;

inputting the first partial reconstruction and the second partial reconstruction into a machine-learning model, which generates one or more motion-compensated reconstructions of the at least part of the scanned region based at least on the first partial reconstruction and the second partial reconstruction;

calculating a respective edge entropy of each of the one or more motion-compensated reconstructions of the at least part of the scanned region; and

adjusting the machine-learning model based on the respective edge entropy of each of the one or more motion-compensated reconstructions.

18. The one or more computer-readable storage media of claim 17 , wherein the scan data were obtained by computed-tomography scanning.

19. The one or more computer-readable storage media of claim 17 , wherein the machine-learning model is a neural network.

20. The one or more computer-readable storage media of claim 19 , wherein edge entropy of motion-compensated reconstructions is a loss function of the neural network.