Determining a two-dimensional mammography dataset
A method is for determining a two-dimensional mammography dataset. The method includes the receipt of a three-dimensional mammography dataset of an examination region via an interface. The method furthermore includes the first determination of a two-dimensional mammography dataset of the examination region by application of a trained generator function to the three-dimensional mammography dataset via a processing unit, wherein the trained generator function is based on a trained GA network. Through this method, it is possible efficiently to create two-dimensional mammography datasets, which are visually similar to real two-dimensional mammography datasets and can therefore be appraised with standardized methods.
1. A method for determining a two-dimensional mammography dataset, comprising:
receiving a three-dimensional mammography dataset of an examination region via an interface; and
determining a two-dimensional mammography dataset of the examination region by application of a trained generator function to the three-dimensional mammography dataset received via a processing unit, wherein the trained generator function is based on a trained GA algorithm.
2. The method of claim 1 , wherein the trained GA algorithm includes a generative function and a classifying function, and wherein the trained generator function is identical to a generative subnetwork of the trained GA algorithm.
3. The method of claim 2 , wherein the trained generator function is an artificial neural network.
4. The method of claim 2 , wherein the trained generator function includes a concatenation of a first subfunction, a first projection function and a second subfunction, and wherein the determining of the two-dimensional mammography dataset of the examination region includes
mapping, via the first subfunction, the three-dimensional mammography dataset to a first feature vector,
mapping, via the first projection function, the first feature vector to a second feature vector, and
mapping, via the second subfunction, the second feature vector to the two-dimensional mammography dataset.
5. The method of claim 2 , wherein the two-dimensional mammography dataset is extended in relation to a first direction and is extended in relation to a second direction, the method further comprising:
removing a two-dimensional layer of the three-dimensional mammography dataset via the processing unit, wherein the two-dimensional layer is extended in relation to the first direction and is extended to the second direction.
6. The method of claim 5 , wherein the two-dimensional layer is determined based on a position of a nipple in the three-dimensional mammography dataset.
7. The method of claim 5 , wherein the two-dimensional layer is determined based on weighting factors of voxels of the three-dimensional mammography dataset.
8. The method of claim 2 , wherein the two-dimensional mammography dataset comprises one or more pixels, and wherein the method further comprises:
determining, in a second determination, a second two-dimensional probability dataset via the processing unit,
wherein the second two-dimensional probability dataset assigns, to each respective pixel of the one or more pixels of the two-dimensional mammography dataset, a respective probability that the respective pixel is mapping a lesion.
9. A non-transitory computer program product storing a computer program, directly loadable into a memory of a determination system, including program sections for carrying out the method of claim 2 , when the program sections are executed by the determination system.
10. A non-transitory computer-readable storage medium, storing program sections readable and executable by a determination system, for carrying out the method of claim 2 , when the program sections are executed by the determination system.
11. The method of claim 1 , wherein the trained generator function is an artificial neural network.
12. The method of claim 1 , wherein the trained generator function includes a concatenation of a first subfunction, a first projection function and a second subfunction, and wherein the determining of the two-dimensional mammography dataset of the examination region includes
mapping, via the first subfunction, the three-dimensional mammography dataset to a first feature vector,
mapping, via the first projection function, the first feature vector to a second feature vector, and
mapping, via the second subfunction, the second feature vector to the two-dimensional mammography dataset.
13. The method of claim 12 , wherein the first subfunction includes at least one three-dimensional convolutional operator.
14. The method of claim 13 , wherein the second subfunction includes at least one two-dimensional convolutional operator.
15. The method of claim 13 , wherein the projection function is a three-dimensional convolutional operator, wherein the two-dimensional mammography dataset is extended in relation to a first direction and in relation to a second direction, wherein the three-dimensional mammography dataset is extended in relation to a third direction, and wherein an extent of a convolution kernel of the three-dimensional convolutional operator, in relation to the third direction, is based on an extent of the three-dimensional mammography dataset in relation to the third direction.
16. The method of claim 12 , wherein the second subfunction includes at least one two-dimensional convolutional operator.
17. The method of claim 16 , wherein the projection function is a three-dimensional convolutional operator, wherein the two-dimensional mammography dataset is extended in relation to a first direction and in relation to a second direction, wherein the three-dimensional mammography dataset is extended in relation to a third direction, and wherein an extent of a convolution kernel of the three-dimensional convolutional operator, in relation to the third direction, is based on an extent of the three-dimensional mammography dataset in relation to the third direction.
18. The method of claim 12 , wherein the projection function is a three-dimensional convolutional operator, wherein the two-dimensional mammography dataset is extended in relation to a first direction and in relation to a second direction,
wherein the three-dimensional mammography dataset is extended in relation to a third direction, and wherein an extent of a convolution kernel of the three-dimensional convolutional operator, in relation to the third direction, is identical to an extent of the three-dimensional mammography dataset, corrected based on effects of at least one of a pooling operator and an unpooling operator of the first subfunction, in relation to the third direction.
19. The method of claim 1 , wherein the two-dimensional mammography dataset is extended in relation to a first direction and is extended in relation to a second direction, the method further comprising:
removing a two-dimensional layer of the three-dimensional mammography dataset via the processing unit, wherein the two-dimensional layer is extended in relation to the first direction and is extended to the second direction.
20. The method of claim 19 , wherein the two-dimensional layer is determined based on a position of a nipple in the three-dimensional mammography dataset.
21. The method of claim 19 , wherein the two-dimensional layer is determined based on weighting factors of voxels of the three-dimensional mammography dataset and wherein a weighting factor of a respective voxel, of the voxels, corresponds to a probability that the respective voxel is mapping a lesion.
22. The method of claim 1 , wherein the two-dimensional mammography dataset comprises one or more pixels, and wherein the method further comprises:
determining, in a second determination, a second two-dimensional probability dataset via the processing unit,
wherein the second two-dimensional probability dataset assigns, to each respective pixel of the one or more pixels of the two-dimensional mammography dataset, a respective probability that the respective pixel is mapping a lesion.
23. A non-transitory computer program product storing a computer program, directly loadable into a memory of a determination system, including program sections for carrying out the method of claim 1 , when the program sections are executed by the determination system.
24. A non-transitory computer-readable storage medium, storing program sections readable and executable by a determination system, for carrying out the method of claim 1 , when the program sections are executed by the determination system.
25. A determination system, comprising:
an interface, embodied to receive a three-dimensional mammography dataset of an examination region;
at least one processor, embodied to determine a two-dimensional mammography dataset of the examination region by application of a trained generator function to the three-dimensional mammography dataset received, wherein the trained generator function is based on a trained GA algorithm.
26. A determination system of claim 25 , wherein the trained GA algorithm includes a generative function and a classifying function, and wherein the trained generator function is identical to a generative subnetwork of the trained GA algorithm.
27. A mammography unit, comprising the determination system of claim 25 .