Radiation therapy treatment planning

Patent metadata

Field	Value
Publication number	US-11826560-B2
Application number	US-202117999708-A
Country	US
Kind code	B2
Filing date	May 19, 2021
Priority date	Jun 5, 2020
Publication date	Nov 28, 2023
Grant date	Nov 28, 2023

Abstract

A computer-implemented method for generating a radiation therapy treatment plan for a volume of a patient, the method comprising: receiving an image of the volume; receiving at least one dose-distribution-derived function configured to provide a value as an output based on, as input, at least part of a dose distribution defined relative to the image; receiving a first probability distribution and at least a second, different, probability distribution, the first and at least second probability distributions; defining a multi-criteria optimization problem comprising at least a first objective function based on the at least one dose-distribution-derived function, the first probability distribution and a loss function; and a second objective function based on the at least one dose-distribution-derived function, the second probability distribution and the loss function; and performing a multi-criteria optimization process based on the at least two objective functions to generate at least two output treatment plans.

First claim

The invention claimed is:

1. A computer-implemented method for generating a radiation therapy treatment plan for a volume of a patient, the method being performed in a treatment planning system and comprising the steps of:
receiving an image of the volume;
receiving at least one dose-distribution-derived function, the at least one dose-distribution-derived function configured to provide a value as an output based on, as an input, at least part of a dose distribution defined relative to the image;
receiving a first probability distribution and at least a second, different, probability distribution, the first and at least second probability distributions indicating an achievability or desirability of a range of values output from the at least one dose-distribution-derived function;
defining a multi-criteria optimization problem comprising at least two objective functions comprising:
a first objective function based on the at least one dose-distribution-derived function, the first probability distribution and a loss function; and
a second objective function based on the at least one dose-distribution-derived function, the at least, a second probability distribution and the loss function; and

performing a multi-criteria optimization process based on the at least two objective functions to generate at least two output treatment plans, wherein each treatment plan among the at least two output treatment plans is configured to deliver a radiation dose to the patient when the treatment plan is executed on a treatment machine.

2. The computer-implemented method of claim 1 , wherein the method comprises modifying the first probability distribution to form the at least a second probability distribution.

3. The computer-implemented method of claim 2 , wherein the modification of the first probability distribution to form the at least a second probability distribution comprises one or more of:
a change of a mean value of the first probability distribution;
a change in a standard deviation of the first probability distribution;
a change derived from exponential tilting of the first probability distribution; and
a change of a skewness of the first probability distribution.

4. The computer-implemented method of claim 1 , wherein the method includes modifying the first probability distribution to form a modified version thereof prior to defining the multi-criteria optimization problem and wherein the at least two objective functions comprise:
the first objective function based on the at least one dose-distribution-derived function, the modified version of the first probability distribution and the loss function; and
the second objective function based on the at least one dose-distribution-derived function, the at least a second probability distribution and the loss function.

5. The computer-implemented method of claim 1 , wherein the first probability distribution is determined from a database of previously delivered treatment plans, and indicates a likelihood of achieving a range of the values for the at least one dose-distribution-derived function, wherein the likelihood is determined based on dose distributions achieved in previously delivered treatment plans.

6. The computer-implemented method of claim 5 , wherein the step of receiving the first probability distribution comprises:
receiving a current patient image, x, comprising the image of the volume of the patient to be treated and information identifying at least one bodily structure in the image; and
based on the at least one dose-distribution-derived functions, {ψjj}jj, each comprising a function of a dose distribution, d, over the current patient image, estimating the conditional probability distribution:
p ({ψ j ( d )} j |x ,{( x n ,d n )} n

using a machine learning process trained using training data comprising pairs {(x n , d n )} n of historic patient images x n with information identifying the at least one bodily structure in the image and corresponding historic dose distributions d n achieved in previously delivered treatment plans, the conditional probability distribution thereby being indicative of a likelihood of a range of outputs from the dose-distribution-derived functions for the dose distribution, d, for the current patient based on the dose distributions achieved for historic patients.

7. The computer-implemented method of claim 5 , wherein the first probability distribution comprises a Gaussian mixture model wherein parameters of the Gaussian mixture model are determined based on the at least one dose-distribution-derived function and dose distributions derived from the database of previously delivered treatment plans.

8. The computer-implemented method of claim 1 , wherein the method comprises:
receiving a current patient image comprising the image of the volume of the patient to be treated and information identifying at least one bodily structure in the image;
accessing a database having a plurality of records of dose distributions for previously delivered treatment plans and respective patient images with information identifying the at least one bodily structure in the images;
determining a measure of similarity between the current patient image and each of the patient images of the records, at least with respect to the at least one bodily structure;
evaluating one or more of the at least one dose-distribution-derived functions received for the current patient image using the plurality of dose distributions of the records to obtain a dataset of values of the dose-distribution-derived function for each dose distribution in the plurality of dose distributions;
determining, from the dataset, the first probability distribution, corresponding to the evaluated one or more dose-distribution-derived functions, using a mapping function that gives a greater weighting to values of the dataset that correspond to a patient image having a greater measure of similarity with the current patient image and a lesser weighting to values of the dataset that correspond to a patient image having a lesser measure of similarity with the current patient image.

9. The computer-implemented method of claim 8 , wherein
when the dose-distribution-derived function is configured to provide a respective value based on a predefined region of the volume and an input dose distribution is a dose-volume histogram for the predefined region of the volume, the method comprises applying a weighting to the evaluated one or more dose-distribution-derived functions using the mapping function, wherein the mapping function comprises a monotone transformation of the measure of similarity; and
when the dose-distribution-derived function comprises a single-voxel function, wherein the patient image is formed of a plurality of voxels and the dose-distribution-derived function is configured to provide an output equal to a dose delivered to a particular single voxel, the method comprises using a dose prediction model trained to predict a dose distribution of the current patient based on the patient image and information identifying the at least one bodily structure in the image.

10. The computer-implemented method of claim 1 , wherein the method comprises, based on user input, interpolating between the at least two output treatment plans to define an interpolated treatment plan as a final treatment plan of the method.

11. The computer-implemented method of claim 1 , wherein one or more of the at least one dose-distribution-derived functions is defined such that an output value from the one or more of the at least one dose-distribution-derived functions cannot be determined solely by a dose-volume-histogram of the at least part of the dose dis

Assignees

• Raysearch Lab Ab
• Raysearch Laboratories Ab Publ

Inventors

• Zhang Tianfang

Classifications

• A61N5/1039 Primary

  using functional images, e.g. PET or MRI · CPC title

• A61N5/1031 Primary

  using a specific method of dose optimization · CPC title

• A61N5/1038

  taking into account previously administered plans applied to the same patient, i.e. adaptive radiotherapy · CPC title

• G16H20/40

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

• A61N2005/1034

  Monte Carlo type methods; particle tracking · CPC title