Computing radiotherapy dose distribution

1. A computer implemented method for calculating radiotherapy dose distribution, the method comprising:

receiving data representing at least one of particle trajectories or a dose deposition pattern in a simulated delivery of a radiotherapy plan;

applying a dose calculation process to the received data to generate a first radiotherapy dose distribution having a first level of detail; and

processing the first radiotherapy dose distribution using a trained machine learning technique to generate a second radiotherapy dose distribution having a second level of detail that enhances the first level of detail; and

wherein the dose calculation process includes one or more operations that are prematurely stopped or skipped over, before convergence to the second level of detail, to generate the first radiotherapy dose distribution.

2. The method of claim 1 , further comprising generating a radiation dose distribution or treatment plan to perform an operation associated with a radiotherapy treatment device based on the second radiotherapy dose distribution, wherein the operation includes at least one of delivering radiotherapy using the radiotherapy treatment device or verifying proper operation of the radiotherapy treatment device in real-time during treatment.

3. The method of claim 1 , wherein the particle includes at least one of a photon, proton, electron or ion, and wherein the dose calculation process applied to generate the first radiotherapy dose distribution includes at least one or a combination of a Monte Carlo simulation or a deterministic calculation using a point kernel convolution algorithm, a pencil kernel algorithm, or a Boltzmann equation solver.

4. The method of claim 1 , wherein the data includes multiple single particle simulations, wherein the data includes at least one or a combination of a mean-free path parameter, interaction type parameter, deposited energy parameter, and wherein the particle trajectories represent post direction of travel sampled from one or more specified distributions or one or more distributions specified up to a specified value.

5. The method of claim 1 , wherein the dose calculation process includes a Monte Carlo simulation that is prematurely stopped before convergence to the second level of detail, where convergence is measured using a measure of statistical dispersion, and wherein the first dose distribution has a higher measure of statistical dispersion than the second dose distribution.

6. The method of claim 1 , wherein the dose calculation process comprises a point kernel convolution process that includes a ray trace step and a convolution step, further comprising:

applying a dose calculation process to the received data comprises computing an amount representing a total energy released per unit mass in a patient based on the received data using a ray trace process; and

without requiring performing the convolution step modeling transport and dose deposition of photons and electrons generated by the incident photons, based on the trained machine learning technique to generate the second radiotherapy dose distribution.

7. The method of claim 1 , wherein the second radiotherapy distribution represents a dose from individual beamlets or shots.

8. The method of claim 1 , wherein prematurely stopping or skipping over the one or more operations of the dose calculation process results in generating the first radiotherapy dose distribution for the absence of a magnetic field, wherein the second radiotherapy dose distribution is generated, based on the first radiotherapy dose distribution, for the presence of the magnetic field, further comprising generating a complete dose plan based on the second radiotherapy dose distribution for a given radiotherapy machine configuration.

9. The method of claim 1 , wherein the second radiotherapy dose distribution represents an external radiation field associated with a radiotherapy treatment device, further comprising:

determining an amount of radiation associated with the radiotherapy treatment device or treatment room; and

based on the amount of radiation, performing at least one of estimating patient motion, configuring a radiation shield, or computing a beam model associated with the radiotherapy treatment device.

10. The method of claim 1 , wherein the machine learning technique comprises a neural network.

11. The method of claim 10 , wherein the neural network comprises at least one of a convolutional neural network and a denoising autoencoder, wherein the first radiotherapy dose distribution is represented as a three-dimensional image, and wherein processing the first radiotherapy dose distribution using the trained machine learning technique further comprises providing the first radiotherapy dose distribution, one or more additional radiotherapy dose distributions generated using the dose calculation process, patient geometry information, and beam setup information to the trained machine learning technique.

12. The method of claim 1 , wherein the dose calculation process comprises a point kernel convolution process, and wherein the trained machine learning technique is trained based on training data to establish a relationship between a computation representing total energy released per unit mass in a patient and a ground truth dose calculation.

13. The method of claim 12 , wherein the training data comprises sets of input-output data pairs, wherein the ground truth dose calculation is generated by applying the point kernel convolution process including a raytracing step and a convolution step to the training data set, and wherein the input portion of the input-output data pairs comprises a patient model.

14. The method of claim 1 , wherein the machine learning technique is trained based on a single beam and is subsequently further trained based on multiple beams.

15. The method of claim 1 , wherein the dose calculation process comprises a Monte Carlo process, and wherein the trained machine learning technique is trained based on training data to establish a relationship between a partial simulation result of the Monte Carlo process and a full simulation result of the Monte Carlo process.

16. The method of claim 1 , wherein the dose calculation process is performed for the absence of a magnetic field, and wherein the trained machine learning technique is trained based on training data to establish a relationship between the dose calculation process performed for the absence of a magnetic field and the dose calculation process performed for the presence of the magnetic field.

17. The method of claim 1 , wherein the dose calculation process includes a differentiable process, and wherein the machine learning technique comprises a differentiable machine learning technique.

18. The method of claim 1 , wherein the first level of detail represents at least one of or a combination of a first level of noise, a suboptimal input image, or a dose distribution in the absence of a magnetic field, and wherein the second level of detail represents at least one or a combination of a second level of noise less than the first level of detail, an improved input image or the dose distribution in the presence of a magnetic field.

19. A system for calculating radiotherapy dose distribution, the system comprising:

one or more processors configured to perform operations comprising:

receiving data representing at least one of particle trajectories or a dose deposition pattern in a simulated delivery of a radiotherapy plan;

applying a dose calculation process to the received data to generate a first radiotherapy dose distribution having a first level of detail; and

processing the first radiotherapy dose distribution using a trained machine learning technique to generate a second radiotherapy dose distribution having a second level of detail that enhances the first level of detail, wherein the dose calculation process includes one or more operations that are prematurely stopped or skipped over, before convergence to the second level of detail, to generate the first radiotherapy dose distribution.

20. The system of claim 19 , wherein the operations further comprise generating a radiation dose distribution or treatment plan to perform an operation associated with a radiotherapy treatment device based on the second radiotherapy dose distribution, wherein the operation includes at least one of delivering radiotherapy using the radiotherapy treatment device or verifying proper operation of the radiotherapy treatment device in real-time during treatment.

21. The system of claim 19 , wherein the trained machine learning technique is trained based on training data to establish a relationship between one or a combination of a computation representing total energy released per unit mass in a patient and a ground truth dose calculation; a partial simulation result of a Monte Carlo process and a full simulation result of the Monte Carlo process; and the dose calculation process performed for the absence of a magnetic field and the dose calculation process performed for the presence of the magnetic field.

22. A computer implemented method for training a machine learning technique, the method comprising:

receiving a first radiotherapy dose distribution having a first level of detail, the first radiotherapy dose distribution being computed based on application of a dose calculation process to data representing at least one of particle trajectories or a dose deposition pattern in a simulated delivery of a radiotherapy plan; and

training a machine learning technique to establish a relationship between the first radiotherapy dose distribution having the first level of detail and a radiotherapy dose distribution having a second level of detail that enhances the first level of detail, wherein the dose calculation process includes one or more operations that are prematurely stopped or skipped over, before convergence to the second level of detail, to generate the first radiotherapy dose distribution.

23. The method of claim 22 , wherein the machine learning technique is trained based on training data to establish a relationship between one or a combination of a computation representing total energy released per unit mass in a patient and a ground truth dose calculation; a partial simulation result of a Monte Carlo process and a full simulation result of the Monte Carlo process; and the dose calculation process performed for the absence of a magnetic field and the dose calculation process performed for the presence of the magnetic field.

24. The method of claim 22 , wherein the first level of detail represents at least one of or a combination of a first level of noise, a suboptimal input image, or a dose distribution in the absence of a magnetic field, and wherein the second level of detail represents at least one or a combination of a second level of noise less than the first level of detail, an improved input image or the dose distribution in the presence of a magnetic field.

25. The method of claim 22 , wherein the machine learning technique is trained by:

obtaining a training data pair comprising a training radiotherapy dose distribution having the first level of detail and a corresponding ground truth radiotherapy dose distribution having the second level of detail;

processing the training radiotherapy dose distribution having the first level of detail with the machine learning technique to generate an estimate of the training radiotherapy dose distribution having the second level of detail;

computing a deviation between the estimate of the training radiotherapy dose distribution having the second level of detail and the ground truth radiotherapy dose distribution having the second level of detail; and

updating parameters of the machine learning model based on the computed deviation.

26. A system for training a machine learning technique, the system comprising:

one or more processors configured to perform operations comprising:

receiving a first radiotherapy dose distribution having a first level of detail, the first radiotherapy dose distribution being computed based on application of a dose calculation process to data representing at least one of particle trajectories or a dose deposition pattern in a simulated delivery of a radiotherapy plan; and

training a machine learning technique to establish a relationship between the first radiotherapy dose distribution having the first level of detail and a second radiotherapy dose distribution having a second level of detail that enhances the first level of detail, wherein the dose calculation process includes one or more operations that are prematurely stopped or skipped over, before convergence to the second level of detail, to generate the first radiotherapy dose distribution.

27. The system of claim 26 , wherein the machine learning technique is trained based on training data to establish a relationship between one or a combination of a computation representing total energy released per unit mass in a patient and a ground truth dose calculation; a partial simulation result of a Monte Carlo process and a full simulation result of the Monte Carlo process; and the dose calculation process performed for the absence of a magnetic field and the dose calculation process performed for the presence of the magnetic field.

28. The method of claim 26 , wherein the first level of detail represents at least one of or a combination of a first level of noise, a suboptimal input image, or a dose distribution in the absence of a magnetic field, and wherein the second level of detail represents at least one or a combination of a second level of noise less than the first level of detail, an improved input image or the dose distribution in the presence of a magnetic field.

29. The method of claim 26 , wherein the machine learning technique is trained by:

obtaining a training data pair comprising a training radiotherapy dose distribution having the first level of detail and a corresponding ground truth radiotherapy dose distribution having the second level of detail;

processing the training radiotherapy dose distribution having the first level of detail with the machine learning technique to generate an estimate of the training radiotherapy dose distribution having the second level of detail;

computing a deviation between the estimate of the training radiotherapy dose distribution having the second level of detail and the ground truth radiotherapy dose distribution having the second level of detail; and

updating parameters of the machine learning model based on the computed deviation.