Selecting a dose prediction model based on clinical goals

1. A computer-implemented method, comprising:

accessing a clinical goal for radiation treatment of a patient, wherein the clinical goal defines a constraint for a quality metric associated with the radiation treatment of the patient, wherein the quality metric comprises a predicted numerical value for a result generated according to the radiation treatment plan;

selecting a first dose prediction model from a plurality of dose prediction models according to the clinical goal; and

generating a radiation treatment plan for the patient, said generating comprising executing the first dose prediction model selected according to the clinical goal.

2. The method of claim 1 , wherein the first dose prediction model is generated using training data based on a sample of other radiation treatment plans that also have the clinical goal.

3. The method of claim 1 , wherein each dose prediction model of the plurality of dose prediction models is indexed by a respective clinical goal and wherein said each dose prediction model is generated using training data based on a respective sample of radiation treatment plans that have the same respective clinical goal.

4. The method of claim 1 , wherein said generating further comprises:

accessing radiation treatment plans in a knowledge base; and

using the radiation treatment plans to generate the radiation treatment plan.

5. The method of claim 1 , wherein said generating further comprises:

accessing radiation treatment plans in a knowledge base;

using the radiation treatment plans to generate a balanced plan, wherein the balanced plan has associated therewith a number of quality metrics and respective quality metric values; and

varying the quality metric values one at a time to generate anchor plans that define a Pareto surface.

6. The method of claim 5 , further comprising navigating the Pareto surface in a graphical user interface that also comprises a number of sliders corresponding to a subset of less than the number of quality metrics.

7. The method of claim 6 , wherein the subset is selected according to a criterion selected from the group consisting of: a ranking of the quality metrics; a user input; a correlation between two or more of the quality metrics; and a knee point in the Pareto surface.

8. The method of claim 1 , wherein the clinical goal is in a format that is parsable by an application that performs said selecting.

9. A computing system comprising:

a central processing unit (CPU); and

memory coupled to the CPU and having stored therein instructions that, if executed by the computing system, cause the computing system to execute operations comprising:

accessing a knowledge base comprising a listing of a plurality of dose prediction models, wherein each dose prediction model of the plurality of dose prediction models is indexed by a respective clinical goal;

accessing a clinical goal for radiation treatment of a patient;

selecting a first dose prediction model from the plurality of dose prediction models, wherein the first dose prediction model is indexed by the clinical goal for the radiation treatment of the patient, wherein the clinical goal defines constraints for quality metrics for the radiation treatment of the patient and is in a format that is parsable by an application that performs said selecting; and

generating a radiation treatment plan for the patient, said generating comprising executing the first dose prediction model selected according to the clinical goal.

10. The system of claim 9 , wherein said each dose prediction model is generated using training data based on a respective sample of radiation treatment plans that have the same respective clinical goal.

11. The system of claim 9 , wherein the radiation treatment plan is a balanced plan, wherein said generating further comprises:

accessing radiation treatment plans in the knowledge base; and

using the radiation treatment plans to generate the balanced plan.

12. The system of claim 9 , wherein said generating further comprises:

accessing radiation treatment plans in the knowledge base;

using the radiation treatment plans to generate a balanced plan, wherein the balanced plan has associated therewith a number of the quality metrics and respective quality metric values; and

varying the quality metric values one at a time to generate anchor plans that define a Pareto surface.

13. The system of claim 12 , further comprising an output device operable for displaying a graphical user interface comprising the Pareto surface and a number of sliders corresponding to a subset of less than the number of the quality metrics, wherein the operations further comprise navigating the Pareto surface in the graphical user interface responsive to movement of the sliders.

14. The system of claim 13 , wherein the subset is selected according to a criterion selected from the group consisting of: a ranking of the quality metrics; a user input; a correlation between two or more of the quality metrics; and a knee point in the Pareto surface.

15. A computing system, comprising:

a central processing unit (CPU);

an output device comprising a display device coupled to the CPU;

memory coupled to the CPU and having stored therein instructions that, when executed by the computing system, cause the computing system to execute a method comprising:

accessing a knowledge base comprising a listing of a plurality of dose prediction models, wherein each dose prediction model of the plurality of dose prediction models is indexed by a respective clinical goal;

selecting a dose prediction model from the plurality of dose prediction models according to a clinical goal specified for a plurality of radiation treatment plans, wherein the clinical goal defines constraints for quality metrics for radiation treatment of a patient and is in a format that is parsable by an application that performs said selecting, and wherein the dose prediction model is indexed by the clinical goal specified for the plurality of radiation treatment plans;

displaying, on the display device, a first element of a graphical user interface, the first element comprising a Pareto surface representing the plurality of radiation treatment plans, wherein there are a number of the quality metrics and respective quality metric values associated with the radiation treatment plans and wherein the radiation treatment plans are generated by varying the quality metric values one at a time in the dose prediction model; and

displaying, on the display device, a second element of the graphical user interface, the second element comprising a number of sliders corresponding to a subset of less than the number of the quality metrics.

16. The system of claim 15 , wherein the method further comprises:

receiving an input indicating movement of at least one of the sliders; and

navigating the Pareto surface in the graphical user interface responsive to the input.

17. The system of claim 15 , wherein the subset is selected according to a criterion selected from the group consisting of: a ranking of the quality metrics; a user input; a correlation between two or more of the quality metrics; and a knee point in the Pareto surface.

18. The system of claim 15 , wherein the dose prediction model is generated using training data based on other radiation treatment plans that also have the clinical goal.