SYSTEMS AND METHODS FOR AUTOMATIC CREATION OF DOSE PREDICTION MODELS AND THERAPY TREATMENT PLANS AS A CLOUD SERVICE

• US 20140350863A1
• Filed: 09/27/2013
• Published: 11/27/2014
• Est. Priority Date: 05/21/2013
• Status: Active Grant

• Current Assignee: Varian Medical Systems International AG (Siemens AG)
• Original Assignee: Varian Medical Systems International AG (Siemens AG)
• Inventors: HARTMAN, Joona, CORDERO MARCOS, Maria, KUUSELA, Esa, PELTOLA, Jarkko, NORD, Janne
• Granted Patent: US 9,409,039 B2
• Time in Patent Office: Days
• US Class Current: 702/19
• CPC Class Codes: A61N 2005/1041 using a library of previously administered radiation treatment applied to other patients
• CPC Class Codes: A61N 2005/1074 Details of the control system, e.g. user interfaces
• CPC Class Codes: A61N 2005/1087 Ions; Protons
• CPC Class Codes: A61N 5/103 Treatment planning systems
• CPC Class Codes: A61N 5/1031 using a specific method of dose optimization
• CPC Class Codes: A61N 5/1043 Scanning the radiation beam, e.g. spot scanning or raster scanning
• CPC Class Codes: G16H 20/40 relating to mechanical, radiation or invasive therapies, e.g. surgery, laser therapy, dialysis or acupuncture
• CPC Class Codes: G16H 40/63 for local operation
• CPC Class Codes: G16H 40/67 for remote operation

Abstract

The present invention proposes a method for automatically creating a dose prediction model based on existing clinical knowledge that is accumulated from multiple sources without collaborators establishing communication links between each other. According to embodiments of the claimed subject matter, clinics can collaborate in creating a dose prediction model by submitting their treatment plans into a remote computer system (such as a cloud-based system) which aggregates information from various collaborators and produces a model that captures clinical information from all submitted treatment plans. According to further embodiments, the method may contain a step where all patient data submitted by a clinic is made anonymous or the relevant parameters are extracted and condensed prior to submitting them over the communications link in order to comply with local regulations.

Claims

1. A method for selecting a set of data for a dose prediction model, the method comprising:

receiving a set of patient anatomy data and a set of training data, the training data comprising a plurality of treatment plans selected from a treatment plan database based on a similarity of the plurality of treatment plans with the set of patient anatomy data;

performing an analysis on the set of patient anatomy data and the set of training data;

training a dose prediction model based on the analysis; and

predicting a plurality of dose parameters for a radiation therapy treatment plan by applying the dose prediction model to the set of patient anatomy data;

wherein the receiving, the performing, the training, and the predicting are performed in an application executed by a processor of a computing device.

2. The method according to claim 1, wherein the plurality of treatment plans comprises a radiation treatment plan.

3. The method according to claim 1, wherein training the dose prediction model comprises performing a DVH estimation with a DVH estimation model comprising a plurality of DVH values.

4. The method according to claim 3, wherein performing the analysis comprises:

identifying a plurality of organs at risk (OAR) comprised in the plurality of treatment plans;

analyzing anatomical information for the plurality of OAR; and

modeling an effect of an OAR on the plurality of DVH values.

5. The method according to claim 1, further comprising performing a plurality of uniformity checks on the set of training data.

6. The method according to claim 1, wherein the set of training data comprises a structure set and an absolute dose data.

7. The method according to claim 1, further comprising storing the dose in a knowledge base comprising a plurality of pre-generated dose prediction models.

8. The method according to claim 7, wherein the plurality of pre-dose prediction models comprises a plurality of DVH estimation models generated at a plurality of remote sites.

9. The method according to claim 7, wherein the knowledge base comprises a central storage device, the central storage device being accessible to a plurality of computing devices at the plurality of remote sites.

10. The method according to claim 9, wherein the central device comprises a cloud infrastructure component.

11. The method according to claim 1, wherein training the dose prediction model comprises:

analyzing the set of training data;

converting the set of training data into a plurality of data curves;

determining a plurality of principal components based on the plurality of data curves;

performing a parameterization on the plurality of data curves; and

determining a regression model for the plurality of data curves.

12. The method according to claim 11, wherein predicting the plurality of dose parameters for a radiation therapy treatment plan comprises predicting the plurality of dose parameters based on the regression model.

13. The method according to claim 11, wherein the set of training data comprises at least one of the group consisting of:

a plurality of structure sets comprising spatial information of a plurality of organs and a target of the treatment plan;

a plurality of pre-optimized treatment plans, and

a corresponding 3D dose distribution from a plurality of patients of the plurality of previously-optimized treatment plans.

14. The method according to claim 13, wherein converting the set of training data into a plurality of data curves comprises converting geometric information in the plurality of structure sets into a plurality of organ-specific distance-to-target histogram curves.

15. The method according to claim 13, wherein converting the set of training data into a plurality of data curves comprises converting a plurality of dose matrices comprised in the plurality of structure sets into a plurality of organ-specific dose-volume histogram curves.

16. The method according to claim 12, wherein performing the parameterization on the plurality of data curves comprises:

determining a plurality of structures from the plurality of structure sets; and

determining a plurality of dose parameters from a plurality of dose matrices.

17. The method according to claim 11, wherein performing the parameterization on the plurality of principal components comprises generating a parameter set corresponding to a structure of the plurality of structure sets, a parameter set being comprised from at least one from the group consisting of:

a plurality of principal components of the DTH;

a relative overlap volume;

a relative out-of-field volume;

an absolute organ at risk volume; and

an absolute target volume.

18. The method according to claim 11, wherein training the dose prediction model comprises generating a trained dose prediction model from the dose prediction model, the trained dose prediction model comprising at least one of the following;

a list of treatment plans comprised in the set of training of data;

an identified plurality of principal components for a DVH and a DTH for a plurality of OAR structures comprised in the dose prediction model;

a plurality of coefficients for a regression model based on trained dose prediction model; and

a mean and standard deviation for a plurality of anatomical features comprised in the training set of data.

19. A method for automatically predicting a dose distribution in a radiotherapy treatment plan, the method comprising:

selecting a dose prediction model from a plurality of dose prediction models based on a plurality of characteristics;

predicting a dose distribution based on the dose prediction model;

producing a plurality of optimization objectives based on the dose distribution;

creating a patient treatment plan based on the optimization objectives, and

verifying the patient treatment plan with the dose prediction model,

wherein the predicting, the producing, and the creating is performed in a first computing system,

further wherein the selecting is performed by a user through a user interface executing on a second computing system remote from the first computing system.

20. A system for storing a plurality of dose prediction models, the system comprising:

a shared library of treatment data, the library of treatment data comprising a plurality of pre-generated treatment plans;

a plurality of computing devices, a computing device of the plurality of computing devices comprising;

a memory configured to store a patient anatomical data;

a processor configured to execute an application, the application being configured to apply a dose prediction model to the patient anatomical data to predict a dose distribution to a patient corresponding to the patient anatomical data,

wherein the dose prediction model is trained by analyzing a selection of treatment plans from the plurality of pre-generated treatment plans,

further wherein the shared library of treatment data comprises a cloud storage component communicatively coupled to the plurality of computing devices.