Radiation therapy planning using integrated model
System and method for automatically generate therapy plan parameters by use of an integrate model with extended applicable regions. The integrated model integrates multiple predictive models from which a suitable predictive model can be selected automatically to perform prediction for a new patient case. The integrated model may operate to evaluate prediction results generated by each predictive model and the associated prediction reliabilities and selectively output a satisfactory prediction. Alternatively, the integrated model may select a suitable predictive model by a decision hierarchy in which each level corresponds to divisions of a patient data feature set and divisions on a subordinate level are nested with divisions on a superordinate level.
1. A computer implemented method of automatically generating a radiation treatment plan for a patient, said method comprising:
accessing patient information;
accessing an integrated dose prediction model that integrates a plurality of predictive models, wherein said integrated dose prediction model is a hierarchical model comprising said plurality of predictive models arranged in a hierarchy, wherein a respective predictive model of said plurality of predictive models correlates input variables of patient information features with output variables pertinent to radiation treatment planning to predict an achievable dose distribution in said patient;
automatically selecting one or more predictive models based on said patient information in accordance with said integrated dose prediction model;
processing said patient information in accordance with said one or more predictive models; and
outputting, by said one or more predictive models, a resultant radiation treatment prediction comprising a predicted achievable dose distribution in said patient.
2. The computer implemented method of claim 1 , wherein said automatically selecting and said processing comprise:
generating radiation treatment predictions based on said patient information using said plurality of predictive models;
evaluating said radiation treatment predictions; and
selecting said one or more predictive models based on said evaluating.
3. The computer implemented method of claim 2 , wherein said evaluating comprises evaluating parameters representing reliability, complexity, and probability with respect to said radiation treatment predictions.
4. The computer implemented method of claim 1 ,
wherein said patient information comprises data of a plurality of sets of features, each set of said plurality of sets comprising one or more features,
wherein each of said plurality of predictive models is associated with a respective division with respect to each set of features of said plurality of sets of features, and
wherein said automatically selecting comprises:
identifying a corresponding division in each set of features based on said patient information; and
selecting said one or more predictive models based on identified divisions of said plurality of sets of features.
5. The computer implemented method of claim 4 , wherein said hierarchy comprises a plurality of intermediate levels and a lowest level, wherein said lowest level comprises said plurality of predictive models, wherein each intermediate level corresponds to divisions with respect to a respective set of features, and wherein divisions corresponding to a subordinate intermediate level are nested with divisions corresponding to a superordinate intermediate level.
6. The computer implemented method of claim 4 , wherein said plurality of sets of features are selected from a group consisting of organ type, organ dimension descriptions, target location, target size, geometric characterizations of one or more organs at risk proximate to a target volume.
7. The computer implemented method of claim 1 , wherein said integrated dose prediction model is generated by automatically classifying said plurality of predictive models based on previous clinical data in accordance with a clustering algorithm.
8. The computer implemented method of claim 1 , wherein each predictive model of said plurality of predictive models is automatically established through a machine training process in accordance with a regression method and based on previous clinical data.
9. A non-transitory computer-readable storage medium embodying instructions that, when executed by a processing device, cause the processing device to perform a method of automatically generating an integrated dose prediction model for planning radiation therapy, said method comprising:
accessing a plurality of predictive models, wherein each of said plurality of predictive models correlates input variables representing patient information features with variables pertinent to radiation treatment planning to predict an achievable dose distribution in a patient; and
integrating said plurality of predictive models into said integrated dose prediction model, wherein said integrated dose prediction model is a hierarchical model comprising said plurality of predictive models arranged in a hierarchy, wherein said integrated dose prediction model is configured to:
receive data of said patient information features;
automatically select one or more predictive models from said plurality of predictive models based on said data of said patient information features; and
process said data of said patient information features in accordance with said one or more predictive models to output a resultant radiation treatment prediction comprising a predicted achievable dose distribution in said patient.
10. The non-transitory computer-readable storage medium of claim 9 , wherein said method further comprises:
configuring an input interface operable to receive said data of said patient information features; and
configuring an output interface operable to output said resultant radiation treatment prediction.
11. The non-transitory computer-readable storage medium of claim 10 , wherein said patient information features are selected from a group consisting of organ identification, organ dimension descriptions, target location, target size, and geometric characterizations of one or more organs at risk proximate to a target volume.
12. The non-transitory computer-readable storage medium of claim 9 , wherein said integrating comprises associating an applicable category with each of said patient information features for each predictive model of said plurality of predictive models.
13. The non-transitory computer-readable storage medium of claim 12 , wherein said hierarchy comprises a bottom level and one or more decision levels, wherein said bottom level comprises said plurality of predictive models, and wherein said one or more decision levels are operable to automatically select said one or more predictive models from said bottom level based on applicable categories associated with said data of said patient information features.
14. The non-transitory computer-readable storage medium of claim 9 , wherein said method further comprises:
accessing training data; and
generating said plurality of predictive models based on said training data in a machine training process, wherein a respective predictive model of said plurality of predictive models is generated in accordance with an algorithm selected from a group consisting of a linear regression algorithm, a classification algorithm, a decision tree algorithm, a segmentation algorithm, an association algorithm, a sequence clustering algorithm, and a combination thereof.
15. The non-transitory computer-readable storage medium of claim 14 , wherein said method further comprises classifying said training data into subsets of training data in accordance with a clustering algorithm, and wherein said generating said plurality of predictive models comprises configuring a predictive model based on a subset of training data.
16. A system comprising:
a processor;
a memory coupled to said processor and comprising instructions that, when executed by said processor, cause the processor to perform a method of automatically generating a radiation therapy plan, said method comprising:
accessing patient information;
accessing an integrated dose prediction model that integrates a plurality of predictive models, wherein said integrated dose prediction model is a hierarchical model comprising said plurality of predictive models arranged in a hierarchy, wherein a respective predictive model of said plurality of predictive models correlates input variables of patient information features with output variables pertinent to radiation treatment planning to predict an achievable dose distribution in a patient;
automatically selecting one or more predictive models based on said patient information in accordance with said integrated dose prediction model;
processing said patient information in accordance with said one or more predictive models; and
outputting, by said one or more predictive models, a resultant radiation treatment prediction comprising a predicted achievable dose distribution in said patient.
17. The system of claim 16 , wherein said automatically selecting and said processing comprise:
generating radiation treatment predictions based on said patient information using said plurality of predictive models;
evaluating said radiation treatment predictions; and
selecting said one or more predictive models based on said evaluating.
18. The system of claim 16 , wherein said patient information comprises data of a plurality of sets of features, each set of said plurality of sets comprising one or more features,
wherein each predictive model of said plurality of predictive models is associated with a respective division with respect to each set of features of said plurality of sets of features, and
wherein said automatically selecting comprises:
identifying a corresponding division in each set of features based on said patient information; and
selecting said one or more predictive models based on identified divisions of said plurality of sets of features.
19. The system of claim 18 , wherein said plurality of sets of features are selected from a group consisting of organ type, organ dimension descriptions, target location, target size, geometric characterizations of one or more organs at risk proximate to a target volume, and wherein said resultant radiation treatment prediction comprises a predicted achievable dose distribution.