Methods and systems for generating dose estimation models for radiotherapy treatment planning
Example methods and systems for generating dose estimation models for radiotherapy treatment planning are provided. One example method may comprise obtaining model configuration data that specifies multiple anatomical structures based on which dose estimation is performed by a dose estimation model. The method may also comprise obtaining training data that includes a first treatment plan associated with a first past patient and multiple second treatment plans associated with respective second past patients. The method may further comprise: in response to determination that automatic segmentation is required for the first treatment plan, performing automatic segmentation on image data associated with the first past patient to generate an improved first treatment plan, and generating the dose estimation model based on the improved first treatment plan and the multiple second treatment plans.
1. A method for a computer system to generate a dose estimation model for radiotherapy treatment planning, wherein the method comprises:
obtaining model configuration data associated with the dose estimation model, wherein the model configuration data specifies multiple anatomical structures based on which dose estimation is performed by the dose estimation model;
obtaining training data that includes a first treatment plan associated with a first past patient and multiple second treatment plans associated with respective multiple second past patients;
determining that automatic segmentation is required for the first treatment plan, and in response to said determining,
performing automatic segmentation on image data associated with the first past patient to generate an improved first treatment plan that includes structure data identifying at least one of the multiple anatomical structures; and
generating the dose estimation model based on the improved first treatment plan and the multiple second treatment plans.
2. The method of claim 1 , wherein determining that automatic segmentation is required comprises:
processing the first treatment plan to extract first structure data and first dose data associated with the first past patient; and
determining whether the first treatment plan is an outlier based on at least one of the following: the model configuration data, the first structure data and the first dose data.
3. The method of claim 2 , wherein determining that automatic segmentation is required comprises:
determining the first treatment plan to be a geometric outlier based on a sub-optimal characteristic associated with the first structure data.
4. The method of claim 3 , wherein determining that automatic segmentation is required comprises:
comparing the first structure data with the model configuration data to identify the sub-optimal characteristic in the form of missing data in the first structure data.
5. The method of claim 2 , wherein determining that automatic segmentation is required comprises:
determining the first treatment plan to be a dosimetric outlier based on a sub-optimal characteristic associated with the first dose data.
6. The method of claim 1 , wherein generating the improved first treatment plan comprises:
performing automatic segmentation on the image data to generate the improved structure data using a deep learning engine.
7. The method of claim 1 , wherein generating the improved first treatment plan comprises:
performing automatic segmentation on the image data to generate the improved structure data using one of the following: a machine learning algorithm, a threshold-based algorithm, and an atlas-based algorithm.
8. A non-transitory computer-readable storage medium that includes a set of instructions which, in response to execution by a processor of a computer system, cause the processor to perform a method of generating a dose estimation model for radiotherapy treatment planning, wherein the method comprises:
obtaining model configuration data associated with the dose estimation model, wherein the model configuration data specifies multiple anatomical structures based on which dose estimation is performed by the dose estimation model;
obtaining training data that includes a first treatment plan associated with a first past patient and multiple second treatment plans associated with respective multiple second past patients;
determining that automatic segmentation is required for the first treatment plan, and in response to said determining,
performing automatic segmentation on image data associated with the first past patient to generate an improved first treatment plan that includes structure data identifying at least one of the multiple anatomical structures; and
generating the dose estimation model based on the improved first treatment plan and the multiple second treatment plans.
9. The non-transitory computer-readable storage medium of claim 8 , wherein determining that automatic segmentation is required comprises:
processing the first treatment plan to extract first structure data and first dose data associated with the first past patient; and
determining whether the first treatment plan is an outlier based on at least one of the following: the model configuration data, the first structure data and the first dose data.
10. The non-transitory computer-readable storage medium of claim 9 , wherein determining that automatic segmentation is required comprises:
determining the first treatment plan to be a geometric outlier based on a sub-optimal characteristic associated with the first structure data.
11. The non-transitory computer-readable storage medium of claim 10 , wherein determining that automatic segmentation is required comprises:
comparing the first structure data with the model configuration data to identify the sub-optimal characteristic in the form of missing data in the first structure data.
12. The non-transitory computer-readable storage medium of claim 9 , wherein determining that automatic segmentation is required comprises:
determining the first treatment plan to be a dosimetric outlier based on a sub-optimal characteristic associated with the first dose data.
13. The non-transitory computer-readable storage medium of claim 8 , wherein generating the improved first treatment plan comprises:
performing automatic segmentation on the image data to generate the improved structure data using a deep learning engine.
14. The non-transitory computer-readable storage medium of claim 8 , wherein generating the improved first treatment plan comprises:
performing automatic segmentation on the image data to generate the improved structure data using one of the following: a machine learning algorithm, a threshold-based algorithm, and an atlas-based algorithm.
15. A computer system configured to generate a dose estimation model for radiotherapy treatment planning, the computer system comprising: a processor and a non-transitory computer-readable medium having stored thereon instructions that, when executed by the processor, cause the processor to:
obtain model configuration data associated with the dose estimation model, wherein the model configuration data specifies multiple anatomical structures based on which dose estimation is performed by the dose estimation model;
obtain training data that includes a first treatment plan associated with a first past patient and multiple second treatment plans associated with respective second past patients;
determine that automatic segmentation is required for the first treatment plan, and in response to said determining,
perform automatic segmentation on image data associated with the first past patient to generate an improved first treatment plan that includes structure data identifying at least one of the multiple anatomical structures; and
generate the dose estimation model based on the improved first treatment plan and the multiple second treatment plans.
16. The computer system of claim 15 , wherein the instructions for determining that automatic segmentation is required cause the processor to:
process the first treatment plan to extract first structure data and first dose data associated with the first past patient; and
determine whether the first treatment plan is an outlier based on at least one of the following: the model configuration data, the first structure data and the first dose data.
17. The computer system of claim 16 , wherein the instructions for determining that automatic segmentation is required cause the processor to:
determine the first treatment plan to be a geometric outlier based on a sub-optimal characteristic associated with the first structure data.
18. The computer system of claim 17 , wherein the instructions for determining that automatic segmentation is required cause the processor to:
compare the first structure data with the model configuration data to identify the sub-optimal characteristic in the form of missing data in the first structure data.
19. The computer system of claim 16 , wherein the instructions for determining that automatic segmentation is required cause the processor to:
determine the first treatment plan to be a dosimetric outlier based on a sub-optimal characteristic associated with the first dose data.
20. The computer system of claim 15 , wherein the instructions for generating the improved first treatment plan cause the processor to:
perform automatic segmentation on the image data to generate the improved structure data using a deep learning engine.
21. The computer system of claim 15 , wherein the instructions for generating the improved first treatment plan cause the processor to:
perform automatic segmentation on the image data to generate the improved structure data using one of the following: a machine learning algorithm, a threshold-based algorithm, and an atlas-based algorithm.