Automatic creation and selection of dose prediction models for treatment plans

Patent metadata

Field	Value
Publication number	US-9827445-B2
Application number	US-201314040468-A
Country	US
Kind code	B2
Filing date	Sep 27, 2013
Priority date	Sep 27, 2013
Publication date	Nov 28, 2017
Grant date	Nov 28, 2017

Abstract

A dose prediction model can be determined for generating a dose distribution of a treatment plan for irradiating a target structure within a patient. Treatment plans from previous patients can be analyzed to determine D characteristic values to obtain a D dimensional point for each treatment plan. The treatment plans can be clustered based on the D dimensional points. The treatment plans of a cluster can then be used to determine a dose prediction model. A dose prediction model for patient can be selected from among multiple models. Characteristics about the patient can be used to determine a D dimensional point corresponding to the patient. The D-dimensional point can be used to select a model in comparison to D dimensional points of the models.

First claim

What is claimed is:

1. A method of determining a dose prediction model for generating a dose distribution of a radiation treatment plan for irradiating a target structure within a patient, the method comprising:
for each of N radiation treatment plans corresponding to a plurality of former patients treated with a same type of radiation treatment:
receiving information about the radiation treatment plan, including information about a patient for whom the radiation treatment plan was designed and a respective dose distribution;
analyzing, by a computer system, the information to identify D characteristic values of the radiation treatment plan, each characteristic value corresponding to a different characteristic of the radiation treatment plan, the D characteristic values including at least one geometric characteristic value of a tumor of the patient;
assigning, by the computer system, a coordinate to each of the D characteristic values to obtain a D-dimensional point corresponding to the radiation treatment plan, wherein each characteristic corresponds to a different dimension in a D-dimensional space;

identifying, by the computer system, a cluster of M radiation treatment plans based on the D-dimensional points, where M is less than N, and wherein at least two of the M radiation treatments plans in the cluster have a non-zero distance between the respective D-dimensional points;
using the M radiation treatment plans of the cluster to determine a dose prediction model for generating a new radiation treatment plan, wherein when a new patient is assigned to the cluster from among a plurality of clusters of the N radiation treatment plans, the dose prediction model is configured to output a new dose distribution for the new patient using the dose distributions of the M radiation treatment plans in the cluster based on a new D-dimensional point corresponding to the new patient, the new patient being not one of the former patients;
generating the new radiation treatment plan based on the new dose distribution, the new radiation treatment plan including one or more angles of a treatment head and sequence of movements of a multileaf collimator for achieving the new dose distribution; and
providing, by the treatment head coupled with a radiation source, radiation at the one or more angles and using the sequence of movements of the multileaf collimator to specific portions of a treatment area of the new patient according to the new radiation treatment plan.

2. The method of claim 1 , wherein identifying a cluster of M radiation treatment plans based on the D-dimensional points includes:
determining distances between the D-dimensional points; and
identifying the cluster of M radiation treatment plans based on the distances.

3. The method of claim 1 , further comprising:
identifying K clusters of radiation treatment plans;
determining a respective dose prediction model for each cluster;
receiving information about the new patient;
analyzing the information to determine the new D-dimensional point for the new patient;
using the new D-dimensional point to determine one or more clusters corresponding to the new patient; and
using one or more dose prediction models corresponding to the one or more clusters to determine the new radiation treatment plan for the new patient.

4. The method of claim 3 , wherein the one or more clusters are a plurality of clusters, the method further comprising:
for each of the plurality of clusters:
identifying the corresponding dose prediction model;
determining a reliability score of the dose prediction model;
comparing the reliability scores to identify the dose prediction model for determining the new radiation treatment plan.

5. The method of claim 1 , wherein identifying a cluster of M radiation treatment plans based on the D-dimensional points includes:
receiving new information about the new patient;
analyzing the new information to determine the new D-dimensional point for the new patient;
determining distances between the new D-dimensional point and each of the D-dimensional points corresponding to the N radiation treatment plans; and
identifying the cluster of M radiation treatment plans based on the distances.

6. The method of claim 5 , wherein identifying the cluster of M radiation treatment plans based on the distances includes:
determining the M closest D-dimensional points to the new D-dimensional point; or
determining the M D-dimensional points that are within a specified boundary around the new D-dimensional point.

7. The method of claim 1 , wherein the D characteristic values of a radiation treatment plan include a dose distribution of the radiation treatment plan.

8. The method of claim 1 , wherein using the M radiation treatment plans of the cluster to determine the dose prediction model includes:
using the M radiation treatment plans to determine an initial dose prediction model;
identifying one or more of the M radiation treatment plans that are not representative of the initial dose prediction model;
removing the identified one or more radiation treatment plans from the cluster;
using the remaining radiation treatment plans of the cluster to determine the dose prediction model.

9. The method of claim 1 , wherein using the M radiation treatment plans to determine the dose prediction model includes:
comparing dose distributions of the dose prediction model to dose distributions of the M radiation treatment plans; and
updating the dose prediction model based on the comparison.

10. A computer product comprising a non-transitory computer readable medium storing a plurality of instructions that when executed control a computer system to determine a dose prediction model for generating a dose distribution of a radiation treatment plan for irradiating a target structure within a patient, the instructions comprising:
for each of N radiation treatment plans corresponding to a plurality of former patients treated with a same type of radiation treatment:
receiving information about the radiation treatment plan, including information about a patient for whom the treatment plan was designed and a respective dose distribution;
analyzing the information to identify D characteristic values of the radiation treatment plan, each characteristic value corresponding to a different characteristic of the radiation treatment plan, the D characteristic values including at least one geometric characteristic value of a tumor of the patient; and
assigning a coordinate to each of the D characteristic values to obtain a D-dimensional point corresponding to the radiation treatment plan, wherein each characteristic corresponds to a different dimension in a D-dimensional space;

identifying a cluster of M radiation treatment plans based on the D-dimensional points, where M is less than N, and wherein at least two of the M radiation treatments plans in the cluster have a non-zero distance between the respective D-dimensional points;
using the M radiation treatment plans of the cluster to determine a dose prediction model for generating one or more new radiation treatment plans, wherein when a new patient is assigned to the cluster from among a plurality of clusters of the N radiation treatment plans, the dose prediction model is configured to output a new dose distribution for the new patient using the dose distributions of the M radiation treatment plans in the cluster based on a new D-dimensional point corresponding to the new patient;
generating the new radiation treatment plan based on the new dose distribution, the new radiation treatment plan including angles of a treatment head and sequence of movements of a multileaf collimator for achieving the new dose d

Assignees

• Varian Medical Systems Int Ag

Inventors

• Cordero Marcos María Isabel
• Hartman Joona
• Kuusela Esa
• Peltola Jarkko Yrjana
• Nord Janne Ilmari

Classifications

• A61N2005/1041

  using a library of previously administered radiation treatment applied to other patients · CPC title

• A61N5/1031 Primary

  using a specific method of dose optimization · CPC title

• G16H20/17

  delivered via infusion or injection · CPC title

• G16H50/20

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title

• G16H20/40

  relating to mechanical, radiation or invasive therapies, e.g. surgery, laser therapy, dialysis or acupuncture · CPC title