Systems and methods for determining radiation therapy machine parameter settings

Patent metadata

Field	Value
Publication number	US-11517768-B2
Application number	US-201715658484-A
Country	US
Kind code	B2
Filing date	Jul 25, 2017
Priority date	Jul 25, 2017
Publication date	Dec 6, 2022
Grant date	Dec 6, 2022

Abstract

Systems and methods can include a method for training a deep convolutional neural network to provide a patient radiation treatment plan, the method comprising collecting patient data from a group of patients, the patient data including at least one image of patient anatomy and a prior treatment plan, wherein the treatment plan includes predetermined machine parameters, and training a deep convolution neural network for regression by using the prior treatment plans and the corresponding collected patient data to determine a new treatment plan. Systems and methods can also include a method of using a deep convolutional neural network to provide a radiation treatment plan, the method comprising retrieving a trained deep convolution neural network previously trained on patient data from a group of patients, collecting new patient data, wherein the new patient data includes at least one image of patient anatomy, and determining a new treatment plan for the new patient using the trained deep convolutional neural network for regression, wherein the new treatment plan has a new set of machine parameters.

First claim

What is claimed is:

1. A method for training a deep convolutional neural network to provide a patient radiation treatment plan, the method comprising:
collecting patient data from a group of patients, the patient data including at least one image of patient anatomy and a prior treatment plan, wherein the treatment plan includes predetermined machine parameters;
associating a collection of the patient data with a respective set of ground-truth radiotherapy treatment machine parameters; and
training a deep convolution neural network (DCNN) for regression by receiving the collected patient data and the associated respective set of ground-truth radiotherapy treatment machine parameters to determine a new treatment plan, the DCNN being trained to store one or more parameters that establish a relationship between the collection of training medical images, corresponding to the group of patients, and the respective set of ground-truth radiotherapy treatment machine parameters, the DCNN being trained to process a given one of the collection of training medical images to predict an output comprising two or more estimated radiotherapy treatment machine parameters, and the training comprising comparing the two or more estimated radiotherapy treatment machine parameters, that were predicted by the DCNN processing the given one of the collection of training medical images, with a given group of two or more of the set of ground-truth radiotherapy treatment machine parameters corresponding to the given one of the collection of training medical images.

2. The method of claim 1 , wherein the new treatment plan comprises predicted machine parameters, wherein training the DCNN comprises:
receiving training data comprising the collection of training medical images and the respective set of ground-truth radiotherapy treatment machine parameters, the set of ground-truth radiotherapy treatment machine parameters being generated using a treatment planning process prior to training the DCNN and being received together with the training medical images;
for each batch of training data comprising at least one training medical image of the collection of training medical images and a given set of ground-truth radiotherapy treatment machine parameters:
applying the DCNN to the at least one training medical image to generate a set of estimated radiotherapy treatment machine parameters;
comparing the set of estimated radiotherapy treatment machine parameters with the given set of ground-truth radiotherapy treatment machine parameters; and
updating the one or more parameters of the DCNN based on a result of comparing the set of estimated radiotherapy treatment machine parameters with the given set of ground-truth radiotherapy treatment machine parameters; and
applying the DCNN with the updated one or more parameters to another batch of the training data.

3. The method of claim 1 , further comprising:
prior to training the DCNN:
obtaining a first training medical image from the collection of training medical images;
applying a radiotherapy treatment planning process to generate a first set of ground-truth radiotherapy treatment machine parameters based on the first training medical image; and
forming a training data pair comprising the first training medical image and the generated first set of ground-truth radiotherapy treatment machine parameters;
receiving, by the DCNN, the first training medical image and the generated first set of ground-truth radiotherapy treatment machine parameters together as the formed training data pair; and
training the deep convolutional neural network, based on training data pair comprising the first training medical image and the generated first set of ground-truth radiotherapy treatment machine parameters, by adjusting the one or more parameters of the deep convolutional neural network to minimize a cost function that includes a difference between the predetermined sets of machine parameters comprising the first set of ground-truth radiotherapy treatment machine parameters and predicted sets of machine parameters generated based on the first training medical image.

4. The method of claim 2 , wherein the predetermined machine parameters include at least one of a gantry angle, a multi-leaf collimator leaf position, or a radiation therapy beam intensity, wherein the treatment planning process used to generate the set of ground-truth radiotherapy treatment machine parameters excludes the DCNN, a result of generating the set of ground-truth radiotherapy treatment machine parameters using the treatment planning process that excludes the DCNN being used to train the DCNN to predict the radiotherapy treatment machine parameters.

5. The method of claim 3 wherein the predicted machine parameters include at least one of a gantry angle, a multi-leaf collimator leaf position, or a radiation therapy beam intensity.

6. The method of claim 1 , further comprising collecting patient data including at least one signed distance map from each patient in a group of patients.

7. The method of claim 1 , wherein the at least one image of patient anatomy includes at least one of a planning CT image, an anatomy label map, a determined object distance such as a signed distance map from the patient.

8. A method of using a deep convolutional neural network to provide a radiation treatment plan, the method comprising:
retrieving a trained deep convolution neural network previously trained on patient data from a group of patients;
collecting new patient data, wherein the new patient data includes at least one image of patient anatomy; and
determining a new treatment plan for the new patient using the trained deep convolutional neural network (DCNN) for regression, wherein the new treatment plan has a new set of machine parameters, the DCNN being trained to receive a collection of training medical images and associated respective set of ground-truth radiotherapy treatment machine parameters and to store one or more parameters that establish a relationship between the collection of training medical images, corresponding to the group of patients, and the respective set of ground-truth radiotherapy treatment machine parameters, the DCNN being trained to process a given one of the collection of training medical images to predict an output comprising two or more estimated radiotherapy treatment machine parameters, and the training comprising comparing the two or more estimated radiotherapy treatment machine parameters, that were predicted by the DCNN processing the given one of the collection of training medical images, with a given group of two or more of the set of ground-truth radiotherapy treatment machine parameters corresponding to the given one of the collection of training medical images.

9. The method of claim 8 , wherein the trained deep convolutional neural network can provide the new treatment plan including the set of machine parameters, wherein the set of machine parameters includes at least one of a gantry angle, a multi-leaf collimator leaf position, or a radiation therapy beam intensity.

10. The method of claim 9 , wherein the new treatment plan is created in real-time.

11. The method of claim 9 , wherein the new treatment plan is created in real-time during a radiation therapy treatment, the DCNN being trained by:
receiving training data comprising the collection of training medical images and the respective set of ground-truth radiotherapy treatment machine parameters, the set of ground-truth radiotherapy treatment machine parameters being generated using a treatment planning process prior to training the DCNN and being received together with the training medical images;
for each batch of training data compr

Assignees

• Elekta Inc

Inventors

• Hibbard Lyndon S

Classifications

• A61N5/1031 Primary

  using a specific method of dose optimization · CPC title

• G16H20/40

  relating to mechanical, radiation or invasive therapies, e.g. surgery, laser therapy, dialysis or acupuncture · CPC title

• A61N2005/1041

  using a library of previously administered radiation treatment applied to other patients · CPC title

• A61N5/1067

  in real time, i.e. during treatment · CPC title

• G16H30/40

  for processing medical images, e.g. editing · CPC title