Neural network-based corrector for photon counting detectors

1. A method for x-ray photon count data correction, the method comprising:

receiving, by an artificial neural network (ANN), a detected photon count from a photon counting detector, corresponding to an attenuated energy spectrum, which is related to characteristics of an x-ray source used for imaging and an object being imaged; and

correcting, by the ANN, after training, the detected photon count to produce a corrected photon count in an energy window,

wherein the training comprises training the ANN using parameters including a trigger threshold, and

wherein the trigger threshold is greater than or equal to a maximum energy of an x-ray energy spectrum output fromm the x-ray source.

2. The method of claim 1 , wherein the correcting of the detected photon count in the energy window comprises correcting the detected photon count in the energy window based on respective photon counts in a plurality of energy windows.

3. The method of claim 1 , further comprising reconstructing, by image reconstruction circuitry, an image based on the corrected photon count.

4. The method of claim 1 , wherein the correcting of the detected photon count comprises correcting for at least one of pulse pileup or a charge sharing effect.

5. The method according to claim 1 , wherein at least one of the ANN, the photon counting detector, and the x-ray source is configured based on an x-ray imaging data formation model.

6. The method according to claim 1 , wherein the training comprises training the ANN based on simulated training data.

7. The method according to claim 1 , wherein the attenuated energy spectrum is based on an incident energy spectrum, which corresponds to radiation energy, x-ray energy, gamma ray or infrared light.

8. The method according to claim 1 , wherein the photon counting detector is paralyzable or nonparalyzable.

9. The method according to claim 1 , wherein the ANN is trained based on at most five epochs or using a back propagation technique.

10. The method according to claim 1 , wherein the ANN corresponds to a multilayer perceptron, a residual neural network or an adversarial neural network.

11. The method of claim 1 , wherein the training comprises training the ANN based on training pairs including a detected photon count and a corresponding actual photon count.

12. The method of claim 1 , wherein

the training further comprises training the ANN using training circuitry, which includes an x-ray source module, an attenuator module, a photon generator module, a pulse generator module, a signal generator module, and a signal detector module, and

the pulse generator module is configured to receive energies and corresponding arrival times of photons from the photon generator module, and output a corresponding pulse height of each of the photons to the signal generator.

13. The method of claim 12 , wherein

the signal generator module is configured to receive the pulse heights, determine a signal height as a function of time by computing a summation of the pulse heights, and output the signal height as a function of time,

the signal detector module is configured to receive the signal height, and output detected counts in respective energy windows, and

the signal detector module is further configured to transition from an active state to an inactive state, in response to a photon arriving.

14. The method of claim 13 , wherein the signal detector module is further configured to return to the active state, in response to a time interval elapsing and an absence of another photon arriving within the time interval, wherein the time interval starts at a time of the photon arriving.

15. The method of claim 1 , wherein

the parameters further include spectrum thresholds, and

the spectrum thresholds are less than the maximum energy of the x-ray energy spectrum output from the x-ray source.

16. The method of claim 15 , wherein the maximum energy of the x-ray energy spectrum output from the x-ray source is 120 keV.

17. A method for photon count correction, the method comprising:

receiving, by a trained artificial neural network (ANN), a detected photon count from a photon counting detector, the detected photon count corresponding to an attenuated energy spectrum, the attenuated energy spectrum being related to characteristics of an imaging object and based on an incident energy spectrum; and

correcting, by the trained ANN, the detected photon count to produce a corrected photon count in an energy window,

wherein the training comprises training the ANN using parameters including a trigger threshold, and

wherein the trigger threshold is greater than or equal to a maximum energy of an x-ray energy spectrum output from an x-ray source.

18. The method of claim 17 , further comprising reconstructing, by image reconstruction circuitry, an image, based on the corrected photon count.

19. The method of claim 17 , wherein the correcting of the detected photon count comprises correcting for pulse pileup.

20. The method of claim 17 , wherein the trained ANN is trained based on simulated training data.

21. The method of claim 17 , wherein the incident energy spectrum corresponds to x-ray energy.

22. The method of claim 17 , wherein the trained ANN is configured to receive a plurality of detected photon counts, and each of the detected photon counts correspond to a respective energy spectrum.

23. The method of claim 22 , wherein the trained ANN is configured to correct the plurality of detected photon counts to produce a corresponding plurality of corrected photon counts.

24. The method of claim 17 , wherein the photon counting detector is paralyzable or nonparalyzable.

25. The method of claim 17 , wherein the ANN is trained based on at most five epochs.

26. The method of claim 17 , wherein the trained ANN corresponds to a multilayer perceptron.

27. A device comprising means for performing the method according to claim 17 .

28. A computer readable storage device having stored thereon instructions that, when executed by one or more processors, cause the one or more processors to perform the method according to claim 17 .

29. A photon count correction system comprising:

a trained artificial neural network (ANN) configured to

receive a detected photon count from a photon counting detector, and

correct the detected photon count to produce a corrected photon count in an energy window,

wherein the detected photon count corresponds to an attenuated energy spectrum,

wherein the attenuated energy spectrum is related to characteristics of an imaging object and is based on an incident energy spectrum,

wherein the ANN is trained using parameters including a trigger threshold, and

wherein the trigger threshold is greater than or equal to a maximum energy of an x-ray energy spectrum output from an x-ray source.

30. The system of claim 29 , further comprising image reconstruction circuitry configured to reconstruct an image based on the corrected photon count.

31. The system of claim 29 , wherein the ANN is further configured to correct the detected photon count in the energy window for pulse pileup.

32. The system of claim 29 , wherein the trained ANN is trained based on simulated training data.

33. The system according to claim 29 , wherein the incident energy spectrum corresponds to x-ray energy.

34. The system according to claim 29 , wherein the trained ANN is further configured to capture a plurality of detected photon counts, each photon count corresponding to a respective energy spectrum.

35. The system of claim 34 , wherein the trained ANN is further configured to correct the plurality of photon counts to produce a corresponding plurality of corrected photon counts.

36. The system according to claim 29 , wherein the photon counting detector is paralyzable or nonparalyzable.

37. The system according to claim 29 , wherein the ANN is trained based on at most five epochs.

38. The system according to claim 29 , wherein the trained ANN corresponds to a multilayer perception.