A feature extraction method and device for medical images, a computer device and a medium

By constructing a dual-branch integrated multi-label contrastive learning classification model, and utilizing the complementary information from PET and CT images, the problem of insufficient feature extraction accuracy of PET/CT in diagnosing occult lymph node metastasis in non-small cell lung cancer was solved, achieving higher diagnostic accuracy and assisting in treatment plans.

CN120236088BActive Publication Date: 2026-06-09CHANGZHOU INST OF TECH +1
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Patents(China)
Current Assignee / Owner
CHANGZHOU INST OF TECH
Filing Date
2025-03-24
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing PET/CT methods lack sufficient accuracy in feature extraction when diagnosing occult lymph node metastasis in non-small cell lung cancer, leading to significant diagnostic errors.

Method used

A dual-branch ensemble multi-label contrastive learning classification model is constructed to extract features from PET and CT images respectively, and then perform contrastive classification through the contrastive learning model. By utilizing the complementary information of the two imaging technologies, multiple pre-trained models are used to reduce bias and improve the robustness and generalization ability of the model.

Benefits of technology

It improves the accuracy of lymph node metastasis feature extraction in non-small cell lung cancer, assisting doctors in making more accurate diagnoses and developing treatment plans.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120236088B_ABST
    Figure CN120236088B_ABST
Patent Text Reader

Abstract

The application provides a feature extraction method and device of medical images, computer equipment and a medium, and belongs to the field of image processing, and the method comprises the following steps: acquiring PET and CT image data and corresponding labels of non-small cell lung cancer; a double-branch integrated multi-label contrast learning classification model comprising two parallel branches and connected with the two branches is constructed; the PET image and the CT image are input into the two branches respectively to obtain classification results, the double-branch integrated multi-label contrast learning classification model is trained through the classification results and classification labels, and a feature extraction model is obtained; target PET and CT image data of a target patient are acquired, and the features of the non-small cell lung cancer lymph nodes of the target patient are extracted through the feature extraction model. In this way, the information integration and complementation of two different imaging technologies in the PET and CT image data are fully utilized, and then comparison and classification are performed, so that the accuracy of feature extraction is improved.
Need to check novelty before this filing date? Find Prior Art