Method and system for automated quality assurance in radiation therapy

Patent metadata

Field	Value
Publication number	US-11735309-B2
Application number	US-201815927414-A
Country	US
Kind code	B2
Filing date	Mar 21, 2018
Priority date	Jun 12, 2013
Publication date	Aug 22, 2023
Grant date	Aug 22, 2023

Abstract

Methods and systems for evaluating a proposed treatment plan for radiation therapy, for evaluating one or more delineated regions of interest for radiation therapy, and/or for generating a proposed treatment plan for radiation therapy. Machine learning based on historical data may be used.

First claim

The invention claimed is:

1. A method for evaluating an aspect of a proposed treatment plan for radiation therapy the method comprising:
obtaining the aspect of a proposed radiation therapy treatment plan defining radiation therapy treatment for at least one treatment site, and a set of patient data for a patient using a processor that communicates with one or more memories storing the proposed treatment plan and the set of patient data, the set of patient data comprising at least one set of image data for the at least one treatment site or data derived from the at least one set of image data;
generating a quality assessment output of a calculated quality estimate for the aspect of the proposed radiation therapy treatment plan using the processor to access the proposed radiation therapy treatment plan and the set of patient data stored in the one or more memories and extract one or more plan features from the aspect of the proposed treatment plan and one or more patient features from the set of patient data to evaluate the aspect of the proposed treatment plan according to a quality assurance model of one or more machine-learned rules for automated quality assessment stored in the one or more memories, the machine-learned rules defining expected relationships between the one or more plan features, and the one or more patient features derived from the set of patient data;
generating the quality assurance model of the one or more machine-learned rules for automated quality assessment by one or more of methods selected from the group of: artificial neural networks, tree-based models, support vector machines, K-means, naïve Bayes, deep learning models, and non-linear, multivariate classification or regression models; and
wherein the quality assurance model of the one or more machine-learned rules for automated quality assessment stored in the one or more memories were developed or refined by machine learning trained on features extracted from a plurality of radiation therapy treatment plans.

2. The method of claim 1 , wherein generating the output further comprises:
evaluating the aspect of the proposed treatment plan according to the one or more rules defining expected relationships between a treatment plan characterization and one or more of: the one or more plan features, and the one or more patient features defined in the set of patient data.

3. The method of claim 2 , further comprising:
determining the treatment plan characterization by automatically characterizing the aspect of the proposed treatment plan according to the one or more plan features;
wherein characterizing the aspect of the proposed treatment plan comprises determining a treatment plan class for the proposed treatment plan according to one of a plurality of predefined treatment plan classes, using an automated classification process.

4. The method of claim 3 , wherein the automated classification process involves rules developed or refined by machine learning using plan features and patient features extracted by the processor from historical data derived from historical treatment plans, a mathematical function, or a general rule governing treatment plans irrespective of the proposed treatment plan and irrespective of the set of patient data.

5. The method of claim 3 , wherein the automated classification process is based on determining similarity of features of the proposed treatment plan to features of plans associated with a predefined treatment plan class.

6. The method of claim 1 , wherein the plurality of aspects of the proposed treatment plan comprises
a set of region of interest (ROI) data delineating at least one ROI in the set of image data; and
wherein the method further comprises automatically characterizing the at least one ROI according to one or more features to determine at least one ROI characterization;
wherein generating the output for the aspect of the proposed treatment plan includes evaluating the aspect of the proposed treatment plan according to one or more rules defining expected relationships between one or more of: the one or more plan features, one or more patient features, the treatment plan characterization, and the at least one ROI characterization.

7. The method of claim 6 , wherein characterizing the at least one ROI comprises determining at least one ROI class respectively for the at least one ROI, according to one of a plurality of predefined ROI classes, using an automated classification algorithm.

8. The method of claim 6 , further comprising, for each of the at least one ROI, calculating a quality estimate that a given ROI belongs to a given ROI characterization, based on predefined expected features of the given ROI characterization.

9. The method of claim 6 , wherein the at least one ROI characterization is determined based on shape and density value of a given ROI.

10. The method of claim 6 , wherein the at least one ROI is characterized according to ROI features including one or more of: anatomical correspondence, tumours, dosage, regions to avoid, regions for dose evaluation, reference structures and structures to facilitate treatment planning.

11. The method of claim 6 , wherein obtaining the set of ROI data comprises automatically segmenting the at least one ROI from the image data.

12. The method of claim 1 , wherein the one or more patient features comprise at least one of: a patient characteristic, a patient history, a patient diagnosis, and an imaged feature.

13. The method of claim 1 , wherein the output comprises one or more of:
a confidence measure for the treatment plan characterization;
a probability of error for the aspect of the proposed treatment plan;
a confidence level for the aspect of the proposed treatment plan;
an automatic quality estimate score;
one or more suggestions for modifying the aspect of the proposed treatment plan in order to improve the quality estimate;
and one or more features of the treatment plan characterization that is relevant to the quality estimate.

14. The method of claim 1 , wherein the one or more rules defining expected relationships include at least one rule generated by machine learning based on one or more of:
historical suitability of a given treatment plan characterization for historical patients;
historical treatment outcome of a given treatment plan characterization for historical patients;
historical treatment plans for a specific patient;
historical treatment outcomes for the specific patient;
a mathematical function; and
a rule governing treatment plans irrespective of the treatment plan characterization and irrespective of the patient data.

15. The method of claim 1 , wherein providing the output comprises displaying, on an output device, an indication of one or more plan features or one or more patient features giving rise to a quality estimate of a particular value or within a particular value range.

16. The method of claim 1 wherein the proposed treatment plan is characterized according to treatment plan features including one or more of: number of beams, beam parameters, an anatomical site, a tumour histology, a prescription dose, a treatment technique, and a treatment intent.

17. The method of claim 1 , wherein evaluating the proposed treatment plan according to one or more rules defining expected relationships between the one or more patient features comprises evaluating similarity of the one or more patient features for the patient to one or more patient features for another patient.

18. The method of claim 1 further comprising:
displaying a visualization of th

Assignees

• Univ Health Network

Inventors

• Purdie Thomas G
• Mcintosh Christopher James
• Svistoun Igor

Classifications

• G16H20/40 Primary

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

• A61N5/103

  Treatment planning systems · CPC title

• G16H10/60

  for patient-specific data, e.g. for electronic patient records · CPC title

• G16H20/10

  relating to drugs or medications, e.g. for ensuring correct administration to patients · CPC title

• G16H20/30

  relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising · CPC title