Data generation apparatus, data generation method, and program

Patent metadata

Field	Value
Publication number	US-12387339-B2
Application number	US-202117917542-A
Country	US
Kind code	B2
Filing date	Apr 5, 2021
Priority date	Apr 6, 2020
Publication date	Aug 12, 2025
Grant date	Aug 12, 2025

Abstract

There is provided a data generation apparatus including a display control unit that displays, on a screen, an image and borders of a plurality of partial images generated by dividing the image into a plurality of pieces while the image and the borders are superimposed together, an input unit that receives input of a label to be given to each of the plurality of partial images, and a generation unit that generates learning data for causing a learning model to learn by associating each of the plurality of partial images with the label given to the corresponding partial image of the plurality of partial images.

First claim

The invention claimed is:

1. A data generation apparatus comprising:
a display control unit that displays, on a screen, a pathological image and borders of a plurality of partial pathological images generated by dividing the pathological image into a plurality of pieces while the pathological image and the borders are superimposed;
an input unit that receives input of a pathology-related label to be given to each of the plurality of partial pathological images; and
a generation unit that generates learning data for causing a learning model to learn by associating each of the plurality of partial pathological images with the pathology-related label given to a corresponding partial pathological image of the plurality of partial pathological images; wherein
the input unit stores a partial pathological image ID for unique identification of each of the plurality of partial pathological images and the pathology-related label given to the corresponding partial pathological image of the plurality of partial pathological images in a label database in association with each other,
the generation unit acquires the partial pathological image ID and the pathology-related label from the label database, extracts image data of a partial pathological image corresponding to the partial pathological image ID from among pathological images stored in a pathological image database and generates the learning data by combining the extracted image data of the partial pathological image and the pathology-related label corresponding to the partial pathological image ID,
the input unit rejects pathology-related label input for a partial pathological image that meets a predetermined condition among the plurality of partial pathological images, and
the generation unit excludes the partial pathological image, for which pathology-related label input is rejected, among the plurality of partial pathological images from the learning data.

2. The data generation apparatus according to claim 1 , wherein the pathological image includes a plurality of pathological images,
the input unit receives input of information on an image type corresponding to the plurality of pathological images, and
the data generation apparatus further comprises an image processing unit that changes color-related pixel values of the plurality of pathological images in accordance with a change method determined for the image type input to the input unit.

3. The data generation apparatus according to claim 2 , wherein
the change method includes a method that changes an RGB value of each of pixels in the plurality of pathological images to a standardized value and a method that randomly changes the RGB value of each of the pixels in the plurality of pathological images.

4. The data generation apparatus according to any one of claims 1 to 3 , wherein
the display control unit displays both information indicating a ratio of respective pathology-related labels to be given to each of the plurality of partial pathological images, the ratio being a target value, and information indicating an actual value of a ratio of the respective pathology-related labels that have been given to the plurality of partial pathological images.

5. The data generation apparatus according to any one of claims 1 to 3 , wherein
the plurality of partial pathological images are pathological images generated by dividing the pathological image at a predetermined interval in a right-left direction and in an up-down direction from a point shifted by a predetermined distance from a point serving as a reference in the pathological image, and
the input unit receives designation of the predetermined distance.

6. A data generation method to be executed by a data generation apparatus, comprising:
a step of displaying, on a screen, a pathological image and borders of a plurality of partial pathological images generated by dividing the pathological image into a plurality of pieces while the pathological image and the borders are superimposed;
a step of receiving input of a pathology-related label to be given to each of the plurality of partial pathological images;
a step of generating learning data for causing a learning model to learn by associating each of the plurality of partial pathological images with the pathology-related label given to a corresponding partial pathological image of the plurality of pathological partial images;
a step of storing a partial pathological image ID for unique identification of each of the plurality of partial pathological images and the pathology-related label given to the corresponding partial pathological image of the plurality of partial pathological images in a label database in association with each other;
a step of acquiring the partial pathological image ID and the pathology-related label from the label database;
a step of extracting image data of a partial pathological image corresponding to the partial pathological image ID from among pathological images stored in a pathological image database;
a step of generating the learning data by combining the extracted image data of the partial pathological image and the pathology-related label corresponding to the partial pathological image ID;
a step of rejecting pathology-related label input for a partial pathological image that meets a predetermined condition among the plurality of partial pathological images; and
a step of excluding the partial pathological image, for which pathology-related label input is rejected, among the plurality of partial pathological images from the learning data.

7. A computer-readable non-transitory storage medium storing a program for causing a computer to execute:
a step of displaying, on a screen, a pathological image and borders of a plurality of partial pathological images generated by dividing the pathological image into a plurality of pieces while the pathological image and the borders are superimposed;
a step of receiving input of a pathology-related label to be given to each of the plurality of partial pathological images;
a step of generating learning data for causing a learning model to learn by associating each of the plurality of partial pathological images with the pathology-related label given to a corresponding partial pathological image of the plurality of partial pathological images;
a step of storing a partial pathological image ID for unique identification of each of the plurality of partial pathological images and the pathology-related label given to the corresponding partial pathological image of the plurality of partial pathological images in a label database in association with each other;
a step of acquiring the partial pathological image ID and the pathology-related label from the label database;
a step of extracting image data of a partial pathological image corresponding to the partial pathological image ID from among pathological images stored in a pathological image database;
a step of generating the learning data by combining the extracted image data of the partial pathological image and the pathology-related label corresponding to the partial pathological image ID;
a step of rejecting pathology-related label input for a partial pathological image that meets a predetermined condition among the plurality of partial pathological images; and
a step of excluding the partial pathological image, for which pathology-related label input is rejected, among the plurality of partial pathological images from the learning data.

8. The computer-readable non-transitory storage medium according to claim 7 , wherein the pathological image includes a plurality of pathological images, and the computer further executes steps comprising:
receiving input of information on an image type corresponding to the pl

Assignees

• Univ Niigata
• Denka Company Ltd

Inventors

• Okuda Shujiro
• Watanabe Yu
• Izutsu Hiroshi
• Kodama Keisuke

Classifications

• G06T11/10

  Texturing; Colouring; Generation of textures or colours (retouching, inpainting or scratch removal G06T5/77) · CPC title

• G06V10/774 Primary

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06T2210/41

  Medical · CPC title

• G06T2207/30096

  Tumor; Lesion · CPC title

• G06T2207/20081

  Training; Learning · CPC title