Cell endoscopic image lesion classification method and system

Through deep learning algorithms and deep segmentation technology, the lesion area of ​​the endoscopic image is divided into central and marginal areas, and combined with text information for analysis, solving the problem of ignoring regional differences and scarcity of data in the existing technology, and improving the accuracy of lesion classification and the generalization ability of the model.

CN119693944AInactive Publication Date: 2025-03-25JILIN UNIV FIRST HOSPITAL
0 Cites 0 Cited by

Patent Information

Application Number
CN202510192367.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-02-21
Publication Date
2025-03-25
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing endoscopic image lesion classification strategy ignores the differences between the central and marginal regions, resulting in missed classification problems, and due to scarce data, the classification model prediction accuracy is low.

Method used

The classification recognition model is constructed through deep learning algorithms, current and historical diagnostic data are collected, the lesion area is divided into the center and edge areas, and feature analysis is performed separately, and category and degree analysis is performed based on text information to generate simulated disease images to expand the training data.

Benefits of technology

It improves the accuracy of lesion classification, reduces the risk of misclassification, enhances the discriminant ability of the model, makes up for the shortcomings of single image data, and alleviates the problem of data scarcity.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a cell endoscopic image lesion classification method and system, and relates to the field of data classification, and the method comprises the following steps: 1, collecting a current diagnosis text, a current diagnosis image, a historical diagnosis text, a historical diagnosis image and a historical treatment text of a target patient, cleaning and standardizing the current and historical diagnosis texts, extracting a plurality of lesion tags to form a reference text, and synchronously storing the reference text to a historical database; 2, constructing a classification recognition model through a deep learning algorithm, and obtaining historical diagnosis text and image data for training; a lesion area is actively segmented into a central area and an edge area, influence coefficients of different areas are extracted respectively, fusion analysis is performed by combining the respective influence coefficients of the central area and the edge area, overall information of the lesion is further comprehensively reflected, and the defect of single image data is made up by integrating text information and image data, so that the accuracy of fusion analysis is improved. And classification performance is improved.
Need to check novelty before this filing date? Find Prior Art