Automatic quality checks for radiotherapy contouring

Patent metadata

Field	Value
Publication number	US-12002216-B2
Application number	US-202318120552-A
Country	US
Kind code	B2
Filing date	Mar 13, 2023
Priority date	Sep 22, 2015
Publication date	Jun 4, 2024
Grant date	Jun 4, 2024

Abstract

Systems, devices, methods, and computer processing products for automatically checking for errors in segmentation (contouring) using heuristic and/or statistical evaluation methods.

First claim

The invention claimed is:

1. A method for evaluating quality of segmentation of an anatomical structure in a radiological image, comprising:
receiving a set of rules to be used to evaluate the segmented structure;
determining the rules that are appropriate to be used to evaluate the segmented structure;
selecting the appropriate rules for execution;
determining order of execution of the selected rules; and
applying the selected rules in the determined order to check the quality of the segmentation in the image,
wherein the segmentation is determined to fail quality check when one or more of the selected rules are violated,
wherein the rules include heuristic rules and statistical rules,
the heuristic rules include rules that require that the structure be inside the segmentation in an image, that the segmented structure not overlap with another structure in the image, that the segmentation in the image be a single connected contour, that a length, volume, or shape of the structure be in a predefined range, that gross tumor volume structures be completely covered by at least one planning target volume structures, and that clinical target volume structures be completely covered by at least one planning target volume structures, and
wherein Boolean operators and pixel/voxel counting or a 3D connectivity labeling process is applied to determine whether the heuristic rules are violated.

2. The method of claim 1 , wherein the determining and selection of the rules is based on predefined radiotherapy atlases.

3. The method of claim 1 , wherein the determining, selection, and order of execution of the selected rules are done automatically using a determination engine or manually by a user.

4. The method of claim 1 , wherein the determining whether the heuristic rules are violated comprises:
generating binary images for the segmented structure;
combine the binary images using at least one of a plurality of Boolean operators to obtain a resulting binary image; and
count pixels/voxels in the resulting binary image,
wherein a heuristic rule is determined to be violated if the number of pixels/voxels is not within a predetermined range.

5. The method of claim 1 , wherein the applying of the 3D connectivity labeling process includes identifying disconnections in the segmented structure by:
determining how many segment components does the segmented structure have; and
comparing the number of the segment components with a previously determined number representing the number of connected components delineating the structure,
wherein a heuristic rule is violated if the number of segment components is different from the previously determined number of connected components.

6. The method of claim 1 , wherein the statistical rules evaluate whether the shape of the segmented structure is within a predetermined range of known shapes for the structure by:
generating a statistical shape model for the segmented structure; and
determining whether the shape of the segmented structure is within a predetermined range of known shapes for the segmented structure,
wherein a segmentation is determined to fail the quality check when the shape of the segment is not within the predetermined range.

7. The method of claim 6 , further comprising:
generating the shape model based on a probabilistic distribution of shape variations found in the range of known shapes for the segmented structure; and
performing a statistical test to check if the shape of the segmented structure is within the predetermined range.

8. The method of claim 1 , further comprising:
generating an alert signal when it is determined that the segmentation failed the quality check;
determining and displaying where and why the segmentation failed the quality check; and
allowing a user to correct errors in the segmentation based on the determination.

9. The method of claim 8 , further comprising:
generating an evaluation report providing an overview of the failed quality check; and
allowing a user to respond to the failed quality check in real-time.

10. A system for evaluating quality of segmentation of an anatomical structure in a radiological image, comprising:
a computer processing system including:
a database configured to store a plurality of segmentation evaluation rules;
a user input device; and
a display device,

the computer processing device configured to:
determine the evaluation rules appropriate to be used to evaluate the segmented structure;
select the evaluation rules for execution;
determine order of execution of the selected evaluation rules; and
apply the selected evaluation rules in the determined order to evaluate a quality of the segmentation in the image,

wherein the segmentation is determined to fail quality check when one or more of the selected evaluation rules are violated,
wherein the rules include heuristic rules and statistical rules,
the heuristic rules include rules that require that the structure be inside the segmentation in an image, that the segmented structure not overlap with another structure in the image, that the segmentation in the image be a single connected contour, that a length, volume, or shape of the structure be in a predefined range, that gross tumor volume structures be completely covered by at least one planning target volume structures, and that clinical target volume structures be completely covered by at least one planning target volume structures, and
wherein Boolean operators and pixel/voxel counting or a 3D connectivity labeling process is applied to determine whether the heuristic rules are violated.

11. The system of claim 10 , wherein the computer processing system is configured to automatically determine and select the evaluation rules and the order of execution of the selected evaluation rules using a determination engine.

12. The system of claim 10 , wherein the determining, selection, and order of execution of the selected evaluation rules is done manually by a user via the user input device.

13. The system of claim 10 , wherein the computer processing system is further configured to:
generate an alert signal when it is determined that the segmentation failed the quality check;
determine and display where and why the segmentation failed the quality check; and
allow the user to correct errors in the segmentation based on the determination.

14. The system of claim 13 , wherein the computer processing system is further configured to:
generate an evaluation report providing an overview of the failed quality check; and
allow the user to respond to the failed quality check in real-time.

15. The system of claim 10 ,
wherein the statistical rules evaluate whether the shape of the segmented structure is within a predetermined range of known shapes for the structure by:
generating a statistical shape model for the segmented structure; and
determining whether the shape of the segmented structure is within a predetermined range of known shapes for the segmented structure,
wherein a segmentation is determined to fail the quality check when the shape of the segment is not within the predetermined range.

Assignees

• Siemens Healthineers Int Ag

Inventors

• Albrecht Thomas
• Hess Peter
• Houhou Nawal

Classifications

• G06T7/143 Primary

  involving probabilistic approaches, e.g. Markov random field [MRF] modelling · CPC title

• G06F16/54

  Browsing; Visualisation therefor · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06T7/0002

  Inspection of images, e.g. flaw detection · CPC title

• G06T7/001 Primary

  using an image reference approach · CPC title