# US 2024/0203100 A1 · Training device, prediction device, training method, and recording medium

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2024/0203100 A1
- Document ID: US20240203100A1
- Status: Published application
- Kind: A1
- Application number: 18/287,082
- Filed: 2021-04-22
- Published: 2024-06-20
- Earliest claimed: 2021-04-22
- Assignees: NEC Corporation
- Applicants: NEC Corporation
- Inventors: Tomoharu KIYUNA
- Web page: https://patentgrep.com/patent/US20240203100A1

## CPC

- G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06N 3/09: Supervised learning
- G06V 10/25: Determination of region of interest [ROI] or a volume of interest [VOI]
- G06V 10/44: Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V 10/764: using classification, e.g. of video objects
- G06V 10/776: Validation; Performance evaluation
- G06V 10/82: using neural networks
- G06V 20/70: Labelling scene content, e.g. deriving syntactic or semantic representations
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The partial image generation means generates partial images from the input image, and the partial image selection means selects a plurality of training partial images to be used as training data from all the generated partial images. The training means trains a prediction model for predicting a probability that a predetermined feature is included in the selected training partial images. The prediction means performs prediction for all the partial images using the trained prediction model. Then, the partial image selection means selects the plurality of training partial images to be used in a next training based on predicted values for all the partial images. The training of the prediction model by the training means is repeated while the training partial images are updated.

## Claims

### Claim 1

A training device comprising:
a memory configured to store instructions; and
a processor configured to execute the instructions to:
generate partial images smaller than an input image from the input image;
select a plurality of training partial images to be used as training data from all the generated partial images;
train a prediction model for predicting a probability that a predetermined feature is included in the selected training partial images; and
perform prediction for all the partial images using the trained prediction model,
wherein the processor selects the plurality of training partial images to be used in a next training based on predicted values for all the partial images.

### Claim 2 (depends on 1)

The training device according to claim 1, wherein the processor selects the partial images in which the predicted value is equal to or higher than a predetermined threshold value as the training partial images in a second and subsequent training.

### Claim 3 (depends on 2)

The training device according to claim 2, wherein the processor determines a reliability of the predicted value of each of the partial images based on the predicted values obtained through multiple predictions, and excludes the partial image whose reliability is lower than a predetermined reference value from the training partial images to optimize next training.

### Claim 4 (depends on 3)

The training device according to claim 3, wherein the processor determines the reliability of each of the partial images based on a variation in the predicted values of each of the partial images.

### Claim 5 (depends on 2)

The training device according to claim 2, wherein, in a first training, the processor randomly selects the training partial images from all the partial images for the input image of positive example which includes the predetermined feature, and selects all the partial images as the training partial images for the input image of negative example which does not include the predetermined feature.

### Claim 6 (depends on 5)

The training device according to claim 5, wherein, in the first training, the processor uses a label given in advance to an entirety of the input image as a label for each of the training partial images included in the input image.

### Claim 7 (depends on 1)

A prediction device comprising:
an additional memory configured to store instructions; and
an additional processor configured to execute the instructions to:
generate partial images smaller than the input image from the input image;
predict a probability that the predetermined feature is included in the partial image for all the generated partial images using the prediction model trained by the training device according to claim 1; and
integrate a prediction result for all the partial images and outputs a prediction score indicating a probability that the predetermined feature is included in the input image.

### Claim 8 (depends on 7)

The prediction device according to claim 7, wherein the additional processor determines an area in which the predetermined feature is included in the input image based on the prediction result for all the partial images, and displays the area on a display device.

### Claim 9

A training method comprising:
generating partial images smaller than an input image from the input image;
selecting a plurality of training partial images to be used as training data from all the generated partial images;
training a prediction model for predicting a probability that a predetermined feature is included in the selected training partial images;
performing prediction for all the partial images using the trained prediction model; and
selecting the plurality of training partial images to be used in a next training based on predicted values for all the partial images.

### Claim 10

A non-transitory computer-readable recording medium recording a program, the program causing a computer to execute processing of:
generating partial images smaller than an input image from the input image;
selecting a plurality of training partial images to be used as training data from all the generated partial images;
training a prediction model for predicting a probability that a predetermined feature is included in the selected training partial images;
performing prediction for all the partial images using the trained prediction model; and
selecting the plurality of training partial images to be used in a next training based on predicted values for all the partial images.

### Claim 11 (depends on 5)

The training device according to claim 5, wherein the processor performs deep learning of the prediction model to predict medicinal effect to cancer based on pathological tissue images.
