A method and related device for knee osteoarthritis kl grading

By extracting global and local features from knee X-ray images using a dual-backbone neural network and combining them with a multi-task prediction head, this method addresses the shortcomings of existing KL grading methods for knee osteoarthritis. It enables accurate identification of early lesions and risk alerts, reducing the workload of doctors and improving diagnostic efficiency.

CN122115945APending Publication Date: 2026-05-29THE THIRD AFFILIATED HOSPITAL OF SOUTHERN MEDICAL UNIV (ACAD OF ORTHOPEDICS GUANGDONG PROVINCE)
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
THE THIRD AFFILIATED HOSPITAL OF SOUTHERN MEDICAL UNIV (ACAD OF ORTHOPEDICS GUANGDONG PROVINCE)
Filing Date
2026-02-02
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing KL grading methods for knee osteoarthritis have shortcomings in structural design and assessment strategies. They fail to effectively distinguish between global anatomical information and local lesions, have inaccurate data segmentation, insufficient generalization ability, ignore the ordinal attribute of KL grades, have a single output format, and lack probability distribution information on the ambiguity between grades.

Method used

A dual-backbone neural network is used to extract features from local images of the entire knee joint and the left and right knee joints, respectively. Through generalized average pooling and feature concatenation, combined with the output of the multi-task prediction head, the KL level probability distribution, the flexible label distribution, and the KL≥2 probability, risk warning information is provided.

Benefits of technology

It improves the ability to identify early osteoarthritis lesions, reduces the workload of doctors, provides rich diagnostic references, is suitable for large-scale screening scenarios, and enhances the robustness and portability of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122115945A_ABST
    Figure CN122115945A_ABST
Patent Text Reader

Abstract

The present application belongs to the technical field of knee osteoarthritis grading, and relates to a knee osteoarthritis KL grading method and related device. The method comprises the following steps: acquiring a full knee X-ray image; automatically segmenting the full knee X-ray image into a left knee local image and a right knee local image; extracting global features of the full knee X-ray image, local features of the left knee local image and the right knee local image by using a double-main-stem neural network; respectively performing generalized average pooling on the three parts of features, and then sequentially performing feature splicing and L2 normalization processing to obtain fused features; based on the fused features, obtaining KL grade probability distribution, flexible label distribution and KL >= 2 probability through a multi-task prediction head; generating a final KL grading result and giving risk prompt information of KL >= 2. The problems of the prior art in structure design and evaluation strategy are solved.
Need to check novelty before this filing date? Find Prior Art