Decision support system for individualizing radiotherapy dose
For decision support in a medical therapy, machine learning provides a machine-learned generator for generating a prediction of outcome for therapy personalized to a patient. The outcome prediction may be used to determine dose. To assist in decision support, a regression analysis of the cohort used for machine training relates the outcome from the machine-learned generator to the dose and an actual control time (e.g., time-to-event). The dose that minimizes side effects while minimizing risk of failure to a time for any given patient is determined from the outcome for that patient and a calibration from the regression analysis.
1. A method for decision support in a medical therapy system, the method comprising:
acquiring a medical scan of a patient;
generating a prediction of outcome from therapy for the patient, the outcome generated by a machine-learned multi-task generator having been trained based with both image feature error and outcome error;
determining a dose for the patient based on a calibration relating the outcome, the dose, and a time-to-event; and
displaying a report including the dose.
2. The method of claim 1 wherein determining comprises determining based on the calibration being a regression from a cohort used to train the machine-learned multi-task generator.
3. The method of claim 2 wherein determining based on the calibration being a regression comprises determining where the regression is a Fine and Gray regression.
4. The method of claim 1 wherein determining comprises determining the dose wherein the outcome has a probability of failure of less than a configurable percentage.
5. The method of claim 1 wherein determining comprises determining with the calibration being for a histological subtype for the patient.
6. The method of claim 1 wherein determining comprises determining with the dose modeled as a continuous variable in the calibration.
7. The method of claim 1 wherein determining comprises identifying the dose as providing the outcome in a given value for the time-to-event.
8. The method of claim 1 wherein determining comprises determining with the calibration comprising a nomogram.
9. The method of claim 1 wherein determining comprises determining with the calibration, the calibration based on estimation of a cumulative incidence function.
10. The method of claim 1 wherein displaying the report comprises displaying the dose as a suggested dose with an estimated failure probability for the suggested dose and further comprises displaying a physician prescribed dose and an estimated failure probability for the prescribed dose.
11. A medical imaging system for therapy decision support, the medical imaging system comprising:
a medical imager configured to scan a patient;
an image processor configured to predict a result of therapy for the patient in response to input of scan data from the scan to a multi-task trained network, and the image processor configured to estimate a dose for the therapy from a regression relating the dose, a time-to-event, and the result, the dose estimated from the regression so that the result is below a threshold probability of failure at a given value of the time-to-event; and
a display configured to display the dose.
12. The medical imaging system of claim 11 wherein the medical imager comprises a computed tomography imager, and wherein the multi-task trained network was trained using a first loss for image features based on handcrafted radiomics and using a second loss for outcome.
13. The medical imaging system of claim 11 wherein the regression comprises a calibration from a cohort used to train the multi-task trained network.
14. The medical imaging system of claim 11 wherein the regression comprises a nomogram relating the dose, the time-to-event, and the result.
15. The medical imaging system of claim 11 wherein the threshold probability comprises a clinician configurable percentage.
16. The medical imaging system of claim 11 wherein the regression is for a histological subtype for the patient.
17. The medical imaging system of claim 11 wherein the dose is modeled as a continuous variable in the regression.
18. The medical imaging system of claim 11 wherein the image processor is configured to estimate the dose as providing the result in the given value for the time-to-event.
19. The medical imaging system of claim 11 wherein the regression is based on estimation of a cumulative incidence function.
20. The medical imaging system of claim 11 wherein the display is configured to display the dose and a physician selected dose with respective estimates of local failure probabilities.