KR-20260062113-A - APPARATUS AND METHOD FOR PROVIDING WHOLE SLIDE IMAGES-BASED TUMOR ANALYSIS SERVICE USING ARTIFICIAL INTELLIGENCE

KR20260062113AKR 20260062113 AKR20260062113 AKR 20260062113AKR-20260062113-A

Abstract

The present disclosure relates to an apparatus and method for providing a tumor analysis service based on a whole slide image using artificial intelligence. In particular, the apparatus comprises: a communication module that communicates with an external device; a memory in which at least one process for performing a tumor analysis operation based on a whole slide image using artificial intelligence is stored; and a processor that performs a tumor analysis operation based on a whole slide image using artificial intelligence according to the process. The processor may be configured to collect a whole slide image of a patient, classify the whole slide image into a first data set and a second data set based on the resolution of each of the whole slide images, input the first data set and the second data set into an artificial intelligence-based pre-training model, and, when a tumor segmentation result based on the first data set and the second data set is output from the pre-training model, generate analysis information using the tumor segmentation result.

Inventors

임광일
김만수
양성규

Assignees

가톨릭대학교 산학협력단
광주과학기술원

Dates

Publication Date: 20260507
Application Date: 20241025

Claims (15)

A communication module that communicates with an external device; A memory storing at least one process for performing a tumor analysis operation based on a whole slide image using artificial intelligence; and It includes a processor that performs a tumor analysis operation based on a whole slide image using artificial intelligence according to the above process, The above processor is, A method configured to collect a whole slide image of a patient, classify the whole slide image into a first data set and a second data set based on the resolution of each of the whole slide images, input the first data set and the second data set into an artificial intelligence-based pre-training model, and, when a tumor segmentation result based on the first data set and the second data set is output from the pre-training model, generate analysis information using the tumor segmentation result. Tumor segmentation device based on whole-slide images using artificial intelligence.
In paragraph 1, The above first data set is, The above entire slide image is composed of low-resolution images with a resolution below a preset threshold, and The above second data set is, The above entire slide image is composed of high-resolution images with a resolution greater than or equal to a preset threshold, Tumor segmentation device based on whole-slide images using artificial intelligence.
In paragraph 1, The above-mentioned prior training model is, Extracting global information through a global branch based on the first dataset, extracting local information through a local branch based on the second dataset, and generating the tumor analysis result by dividing the tumor region by fusing the global information and the local information through a dual branch. Tumor segmentation device based on whole-slide images using artificial intelligence.
In paragraph 3, The above-mentioned prior training model is, The operation of extracting the above global information and the operation of extracting the above local information are each repeated a preset number of times, Tumor segmentation device based on whole-slide images using artificial intelligence.
In paragraph 3, The above global branch is, Extracting global relationships between patches as global information based on the first data set through four steps consisting of a token and patch merging layer and a transformer block, Tumor segmentation device based on whole-slide images using artificial intelligence.
In paragraph 3, The above local branch is, Extracting local information based on the second dataset through four stages of a merge layer and ConvNeXt convex using a convolution operator, Tumor segmentation device based on whole-slide images using artificial intelligence.
In paragraph 4, The above-mentioned prior training model is, Generating fused features by fusing the above global information and the above local information according to the following <Mathematical Formula 1>, Tumor segmentation device based on whole-slide images using artificial intelligence. <Mathematical Formula 1> , Here, represents global features with attention applied at step t, and represents local features with attention applied at step t, and represents the final fused feature.
In a tumor segmentation method based on a whole slide image using artificial intelligence, performed by a device, Step of collecting full slide images of the patient; A step of classifying into a first data set and a second data set based on the resolution of each of the above-mentioned entire slide images; A step of inputting the first data set and the second data set into an artificial intelligence-based pre-training model; and A method comprising the step of generating analysis information using the tumor segmentation results when a tumor segmentation result based on the first data set and the second data set is output from the prior learning model. Tumor segmentation method based on whole-slide images using artificial intelligence.
In paragraph 8, The above first data set is, The above entire slide image is composed of low-resolution images with a resolution below a preset threshold, and The above second data set is, The above entire slide image is composed of high-resolution images with a resolution greater than or equal to a preset threshold, Tumor segmentation method based on whole-slide images using artificial intelligence.
In paragraph 8, The above-mentioned prior training model is, Extracting global information through a global branch based on the first dataset, extracting local information through a local branch based on the second dataset, and generating the tumor analysis result by dividing the tumor region by fusing the global information and the local information through a dual branch. Tumor segmentation method based on whole-slide images using artificial intelligence.
In Paragraph 10, The above-mentioned prior training model is, The operation of extracting the above global information and the operation of extracting the above local information are each repeated a preset number of times, Tumor segmentation method based on whole-slide images using artificial intelligence.
In Paragraph 10, The above global branch is, Extracting global relationships between patches as global information based on the first data set through four steps consisting of a token and patch merging layer and a transformer block, Tumor segmentation method based on whole-slide images using artificial intelligence.
In Paragraph 10, The above local branch is, Extracting local information based on the second dataset through four stages of a merge layer and ConvNeXt convex using a convolution operator, Tumor segmentation method based on whole-slide images using artificial intelligence.
In Paragraph 11, The above-mentioned prior training model is, Generating fused features by fusing the above global information and the above local information according to the following <Mathematical Formula 1>, Tumor segmentation method based on whole-slide images using artificial intelligence. <Mathematical Formula 1> , Here, represents global features with attention applied at step t, and represents local features with attention applied at step t, and represents the final fused feature.
A recording medium capable of being recorded on a computer, storing a program for executing a whole-slide image-based tumor segmentation method using artificial intelligence according to any one of claims 8 to 14.

Description

Apparatus and Method for Providing Whole Slide Image-Based Tumor Analysis Service Using Artificial Intelligence The present disclosure relates to an apparatus and method for providing a tumor analysis service based on whole-slide images using artificial intelligence. In cancer diagnosis and prognosis evaluation, tissue examination and analysis by pathologists based on whole slide images (WSI) are very important as they form the basis of disease diagnosis and research. However, existing manual analysis methods are time-consuming and wasteful, and interpretation may vary among pathologists. Recent advancements in artificial intelligence, particularly in the field of computer vision, have demonstrated the potential to automate medical image analysis with high accuracy. Convolutional Neural Networks (CNNs) excel at extracting local features but have limitations in grasping overall contextual information. On the other hand, Vision Transformers (ViTs) effectively understand the context of an entire image but are computationally expensive and require a large amount of training data. Existing dual-branch approaches have improved performance by combining CNNs and Transformers, but they have the problem of potentially missing details by focusing only on specific resolutions of the entire slice image. Therefore, there is a need to develop technology based on artificial intelligence that can automatically and more accurately segment tumors from whole-slice images. FIG. 1 is a diagram showing the network structure of a tumor analysis service provision system based on a whole slide image using artificial intelligence according to one embodiment of the present disclosure. FIG. 2 is a diagram schematically illustrating the configuration of a tumor analysis service providing device based on a full slide image using artificial intelligence according to one embodiment of the present disclosure. FIG. 3 is a diagram schematically illustrating a data processing procedure through an artificial intelligence-based pre-learning model according to one embodiment of the present disclosure. FIG. 4 is a diagram schematically illustrating the configuration of a user terminal using a tumor analysis service based on a full slide image using artificial intelligence according to one embodiment of the present disclosure. FIG. 5 is a diagram illustrating a method for providing a tumor analysis service based on a whole slide image using artificial intelligence according to one embodiment of the present disclosure. FIG. 6 is a diagram illustrating a specific operation for generating tumor analysis results in a tumor segmentation method based on a whole slide image using artificial intelligence according to one embodiment of the present disclosure. FIGS. 7 and 8 are drawings for comparing and explaining the tumor segmentation performance of an artificial intelligence-based pre-trained model according to one embodiment of the present disclosure. The advantages and features of the present disclosure and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to make the present disclosure complete and to fully inform those skilled in the art of the scope of the present disclosure, and the present disclosure is defined only by the scope of the claims. The terms used in this specification are for describing embodiments and are not intended to limit the disclosure. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. The terms “comprises” and/or “comprising” as used in this specification do not exclude the presence or addition of one or more other components in addition to the components mentioned. Throughout the specification, the same reference numerals refer to the same components, and “and/or” includes each of the mentioned components and all combinations of one or more. Although terms such as “first,” “second,” etc., are used to describe various components, these components are not limited by these terms. These terms are used merely to distinguish one component from another. Accordingly, the first component mentioned below may be the second component within the technical scope of this disclosure. Unless otherwise defined, all terms used herein (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which this disclosure pertains. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise. Throughout this disclosure, the same reference numerals denote the same components. This disclosure does not describe all elements of the embodiments, and general content in the ar