Plane-by-plane iterative reconstruction for digital breast tomosynthesis

1. A method for digitally reconstructing a three-dimensional tomosynthesis image of a patient's body part, the method which comprises:

receiving signals of a tomosynthesis image system, wherein x-ray radiation is sent by an x-ray source to be received at a detector and the radiation is attenuated by the patient's body part;

splitting the reconstruction volume into a plurality of patches, wherein the patches are parallel to a surface of the detector;

providing a model of image acquisition, where the model may be different for each patch;

applying the model to a cost function and determining an update formula for each patch using an iterative optimization algorithm to maximize the cost function; and

updating each patch separately and sequentially with the update formula to provide an updated patch.

2. The method according to claim 1 , wherein the model is based on positional data referring to a physical movement of the x-ray source during x-ray emission.

3. The method according to claim 1 , wherein the cost function is the Poisson likelihood of the measured signals.

4. The method according to claim 1 , which comprises applying only a certain fraction of the calculated update step to the patches during a first few iterations.

5. The method according to claim 1 , wherein updating comprises applying a prior function with a weight β.

6. The method according to claim 5 , which comprises adjusting the weight β interactively.

7. The method according to claim 5 , which comprises manually adjusting the weight β by receiving a user's input with respect to a training set of patient images, in order to automatically calculate delta and β so that contrast is sufficient and edges are preserved.

8. The method according to claim 5 , which comprises adjusting the weight β automatically.

9. The method according to claim 5 , wherein a noise level for each patch and iteration is estimated in function of breast composition and/or acquisition parameters in order to adjust weight β so that a preset target noise level is reached after a predetermined amount of iterations.

10. The method according to claim 5 , which comprises calculating the weight β separately for each of the patches.

11. The method according to claim 1 , which comprises implementing the method steps in a graphics processing unit.

12. A system for digitally reconstructing a three-dimensional tomosynthesis image of a patient's body part, the system comprising:

a receiver for receiving signals of a tomosynthesis image system, wherein x-ray radiation is emitted from an x-ray source to be received at a detector and the radiation is attenuated by structures of the patient body part;

a model generator adapted to provide a model for image acquisition;

a reconstructor adapted for splitting a reconstruction volume into a plurality of patches, wherein the patches are parallel to a surface of the detector;

a storage unit for storing a dataset, the dataset being an updated patch, wherein the updated patch is generated by carrying out the method according to claim 1 .

13. The system according to claim 12 , which comprises a monitor for displaying the three-dimensional image.

14. The system according to claim 12 , wherein said reconstructor is executed on a graphics processing unit.

15. A non-transitory computer-readable storage medium storing a computer program form and having code for execution of the method according to claim 1 , when the instructions are executed on a computer, wherein the computer program is loaded or loadable into a memory or into a graphics processing unit of the computer.