Systems and methods for radiotherapy

Patent metadata

Field	Value
Publication number	US-12370381-B2
Application number	US-202318171396-A
Country	US
Kind code	B2
Filing date	Feb 20, 2023
Priority date	Aug 20, 2020
Publication date	Jul 29, 2025
Grant date	Jul 29, 2025

Abstract

A method may include obtaining input data relating to a target treatment plan for performing radiotherapy on a lesion using a radiation device. The input data may include a first target image of the lesion. The method may also include obtaining a segment shape estimation model. The method may also include estimating, based on the segment shape estimation model and the input data, a plurality of target location combinations of the target treatment plan and a plurality of target segment shapes of a collimator of the radiation device. One of the plurality of target location combinations may indicate a location of the collimator relative to the lesion. Each of the plurality of target segment shapes may correspond to one of the plurality of target location combinations.

First claim

What is claimed is:

1. A system, comprising:
at least one storage device including a set of instructions; and
at least one processor configured to communicate with the at least one storage device, wherein when executing the set of instructions, the at least one processor is configured to direct the system to perform operations including:
obtaining input data relating to a target treatment plan for performing radiotherapy on a lesion using a radiation device, the input data including a first target image of the lesion;
obtaining a segment shape estimation model; and
estimating, based on the segment shape estimation model and the input data, a plurality of target location combinations of the target treatment plan and a plurality of target segment shapes of a collimator of the radiation device, one of the plurality of target location combinations indicating a location of the collimator relative to the lesion, each of the plurality of target segment shapes corresponding to one of the plurality of target location combinations.

2. The system of claim 1 , wherein the plurality of target location combinations are within a plurality of discrete candidate location combinations of a location universal set.

3. The system of claim 2 , wherein the segment shape estimation model is obtained by performing a training process including:
obtaining the location universal set including the plurality of candidate location combinations; and
determining the segment shape estimation model by iteratively training a preliminary model based on the location universal set.

4. The system of claim 3 , wherein obtaining the location universal set including the plurality of candidate location combinations includes:
obtaining a plurality of candidate gantry angles, a plurality of candidate collimator angles, or a plurality of candidate couch locations; and
obtaining the location universal set based on the plurality of candidate gantry angles, the plurality of candidate collimator angles, or the plurality of candidate couch locations.

5. The system of claim 3 , wherein the plurality of target segment shapes are within a distance universal set including a plurality of discrete candidate leaf locations.

6. The system of claim 5 , wherein the training process includes:
obtaining the distance universal set including the plurality of candidate leaf locations; and
determining the segment shape estimation model by iteratively training the preliminary model based on the distance universal set so that the candidate segment shape corresponding to each of the plurality of candidate location combinations output by the segment shape estimation model is within the distance universal set.

7. The system of claim 6 , wherein the plurality of candidate leaf locations include a plurality of candidate opening locations and a plurality of candidate opening widths.

8. The system of claim 6 , wherein the training process includes:
obtaining training data including a plurality of training sets.

9. The system of claim 8 , wherein obtaining the training data includes:
for one of the plurality of training sets, obtaining a historical treatment plan previously generated based on a sample lesion;
obtaining a first sample image of the sample lesion corresponding to the historical treatment plan;
obtaining sample location combinations and corresponding sample segment shapes in the historical treatment plan; and
obtaining the training set based on the first sample image, the sample location combinations, and the sample segment shapes of historical treatment plan.

10. The system of claim 9 , wherein obtaining the training set based on the first sample image, the sample location combinations, and the sample segment shapes of the historical treatment plan includes:
obtaining processed sample location combinations that are within the location universal set, the processed sample location combinations being obtained by processing the sample location combinations based on the location universal set;
obtaining processed sample segment shapes that are within the distance universal set, the processed sample segment shapes being obtained by processing the sample segment shapes based on the distance universal set;
obtaining a sample set including the processed sample segment shapes and closed segment shapes, the closed segment shapes corresponding to the candidate location combinations excluding the processed sample location combinations; and
obtaining the training set by including the first sample image, the processed sample location combinations, and the sample set of the historical treatment plan.

11. The system of claim 10 , wherein the training process includes:
initializing the preliminary model; and
obtaining the segment shape estimation model by updating the initialized preliminary model using an iteration process including a plurality of iterations, at least one of the plurality of iterations of the iteration process including:
obtaining one of the plurality of training sets;
generating estimated segment shapes corresponding to the plurality of candidate location combinations by inputting the first sample image of the training set into an intermediate model, the intermediate model being the initialized preliminary model in a first iteration of the plurality of iterations of the iteration process or a previously updated model generated in a previous iteration in the iteration process;
determining a value of a loss function based on the estimated segment shapes and the sample set in the training set;
determining whether a termination condition is satisfied;

in response to determining that the termination condition is not satisfied,
generating an updated model by updating the intermediate model based on the value of the loss function; and
initiating a next iteration; and

designating the intermediate model in a last iteration of the plurality of iterations of the iteration process as the segment shape estimation model.

12. The system of claim 11 , wherein the value of the loss function is determined based on sparsity of the sample set, the sparsity of the sample set relating to the closed segment shapes in the sample set.

13. The system of claim 11 , wherein the training set includes at least one of a second sample image of normal tissue surrounding the sample lesion, a third sample image of the sample lesion, or sample radiation information of the historical treatment plan, the sample radiation information including at least one of a sample output dose, a sample dose output rate, a sample dose per pulse, or a sample dose distribution in the sample lesion.

14. The system of claim 13 , wherein the sample radiation information is predicted based on the first sample image of the sample lesion, the second sample image of normal tissue surrounding the sample lesion, and the third sample image of the sample lesion.

15. The system of claim 13 , wherein the at least one of the plurality of iterations of the iteration process includes:
generating the estimated segment shapes by inputting at least one of the second sample image, the third sample image, or the sample radiation information of the training set into the intermediate model.

16. A system, comprising:
at least one storage device including a set of instructions; and
at least one processor configured to communicate with the at least one storage device, wherein when executing the set of instructions, the at least one processor is configured to direct the system to perform operations including:
obtaining a preliminary model;
obtaining a location universal set including a plur

Assignees

• Shanghai United Imaging Healthcare Co Ltd

Inventors

• Bose Supratik
• Maltz Jonathan
• Zhou Jingjie

Classifications

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• A61N5/1038

  taking into account previously administered plans applied to the same patient, i.e. adaptive radiotherapy · CPC title

• G06N3/045

  Combinations of networks · CPC title