Techniques for comparing image contours of different human participants using automated tool

Patent metadata

Field	Value
Publication number	US-12586193-B2
Application number	US-202118001163-A
Country	US
Kind code	B2
Filing date	Jan 29, 2021
Priority date	Jun 11, 2020
Publication date	Mar 24, 2026
Grant date	Mar 24, 2026

Abstract

Using a computer-implemented intermediary by which contouring performed by two participants, such as two physicians, can be compared. First, contouring performed by each participant can be compared to contouring performed by the intermediary. Then, by way of the common intermediary and a transitive analysis, contouring performed by each participant can be compared.

First claim

The claimed invention is:

1 . A computer-implemented method of comparing image contouring by different human participants without requiring that the different human participants contour a shared image dataset, the method comprising:
receiving a first image dataset collection, including one or more first image datasets of one or more human or animal subjects produced by an imaging modality, and including first contouring data generated by a first human participant on the one or more first image datasets; receiving a second image dataset collection, including one or more second image datasets of one or more human or animal subjects produced by the imaging modality, including second contouring data generated by a second human participant on the one or more second image datasets, the one or more second image datasets not overlapping with images in the one or more first image datasets; auto-contouring the first one or more first image datasets and the second one or more image datasets, using a computer-implemented autosegmentation engine to generate participant-specific computational reference contours third contouring data generated by the autosegmentation engine without requiring human contouring and without requiring identical or overlapping first and second image datasets, wherein the computational reference contours for each participant are generated from a participant's respective image datasets; comparing contouring by the first human participant to contouring by the second human participant by:
comparing the first contouring data generated by the first human participant to the participant-specific computational reference contours of the third contouring data for the first human participant generated by the autosegmentation engine to generate first comparison data:
comparing the second contouring data generated by the second human participant to the participant-specific computational reference contours of the third contouring data for the second human participant generated by the autosegmentation engine to generate second comparison data; and
performing a transitive analysis comparison of the first human participant to the second human participant by separately analyzing the first comparison data and the second comparison data relative to a common reference provided by the third contouring data generated by the autosegmentation engine;
generating an indication of one or more systemic differences between contouring by the first and second human participants based on the transitive analysis; and displaying, on a user interface, the indication of the one or more systemic differences, wherein the indication includes visually distinguishable representations of missing, extra, and common volumes.

2 . The method of claim 1 , wherein the indication of one or more systematic differences includes at least one of:
an indication of a systematic difference in a contoured volume; an indication of a systematic difference in a lateral contoured dimension; an indication of a systematic difference in an anterior or posterior contoured dimension; or an indication of a systematic difference in a superior or inferior contoured dimension.

3 . The method of claim 2 , wherein the contoured volume or dimension includes at least one of:
a contoured organ volume or dimension; a contoured organ substructure volume or dimension; a contoured anatomical structure volume or dimension; or an injured or diseased region structure volume or dimension.

4 . The method of claim 1 , wherein the indication of the one or more systemic differences are normalized with respect to a parameter based upon a contour generated by the autosegmentation engine.

5 . The method of claim 1 , wherein at least one of the first image dataset collection and the second image dataset collection includes metadata, in addition to the human first contouring data or second contouring data, wherein the metadata is provided as an input to the autosegmentation engine.

6 . The method of claim 5 , wherein the metadata includes at least one of:
an indication of at least one imaging modality type or other imaging modality parameter used to generate images in a corresponding image dataset; an indication of a characteristic of an organ or other target structure to be targeted for treatment; an indication of a characteristic of an organ or other structure at risk to be avoided for treatment; an indication of a characteristic of a patient demographic of the human or animal subject corresponding to images in a corresponding image dataset; an indication of a disease characteristic associated with the human or animal subject corresponding to images in a corresponding image dataset; an indication of a treatment characteristic associated with the human or animal subject corresponding to images in a corresponding image dataset; and an indication of a desired treatment outcome associated with the human or animal subject corresponding to images in a corresponding image dataset.

7 . The method of claim 1 , wherein the autosegmentation engine is configured to perform the auto-contouring by including or using an atlas-based model.

8 . The method of claim 1 , wherein the autosegmentation engine is configured to perform the auto-contouring by including or using a trained model, the trained model trained using at least one or more of statistical learning, artificial intelligence, machine learning, neural networks, generative adversarial network, or deep learning.

9 . The method of claim 8 , wherein the model is trained independently using a different and independent learning image dataset.

10 . The method of claim 9 , wherein:
the different and independent learning image dataset includes images from a population of human or animal subjects overlapping with at least one the first and second image datasets.

11 . The method of claim 1 , wherein at least one of the comparing the first contouring data generated by the first human participant to the third contouring data generated by the autosegmentation engine, or the comparing the second contouring data generated by the second human participant to the third contouring data generated by the autosegmentation engine comprises:
voxel analysis of a human-contoured region compared to an auto-contoured region.

12 . The method of claim 11 , further comprising:
analyzing a distance between (1) one or more unmatched voxels, between the human-contoured region and the auto-contoured region, and (2) a closest voxel location matched between the human-contoured region and the auto-contoured region.

13 . The method of claim 12 , further comprising:
analyzing a direction between (1) one or more unmatched voxels, between the human-contoured region and the auto-contoured region, and (2) a closest voxel location matched between the human-contoured region and the auto-contoured region.

14 . The method of claim 11 , further comprising:
generating a statistical representation of difference vectors between (1) one or more unmatched voxels, between the human-contoured region and the auto-contoured region, and (2) a closest voxel location matched between the human-contoured region and the auto-contoured region.

15 . The method of claim 1 , wherein comparing contouring by the first human participant to contouring by the second human participant comprises comparing contouring by the first human participant to individual or aggregated contouring by a group or population of second human participants.

16 . The method of claim 1 , wherein comparing contouring by the first human participant to contour

Assignees

• Elekta Inc

Inventors

• Nelms Benjamin Edward
• Christodouleas John

Classifications

• G16H30/20

  for handling medical images, e.g. DICOM, HL7 or PACS · CPC title

• G16H20/40

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

• G06T2207/30096

  Tumor; Lesion · CPC title

• G06T2207/30081

  Prostate · CPC title

• G06T2207/30061

  Lung · CPC title