Loss function weight determination and network training method and system and x-ray camera

By calculating the ratio and product of the areas of the masked segmented regions to determine the weights of the loss function, a multi-segmentation region weight loss function is constructed. This solves the problem of insufficient training caused by the uneven area of ​​the regions in the knee joint X-ray image, and improves the segmentation accuracy of the patella, femur, tibia and patellar tendon.

CN122115474APending Publication Date: 2026-05-29SHENZHEN BLUE SHADOW MEDICAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHENZHEN BLUE SHADOW MEDICAL TECH CO LTD
Filing Date
2026-01-30
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

In the medical image training set, the area ratio of the N mask segmentation regions corresponding to the mask label image is unbalanced, resulting in the smaller mask segmentation regions not being trained sufficiently. This is especially true in knee X-ray images, where the segmentation performance of the patella, femur, tibia, and patellar tendon is insufficient.

Method used

By calculating the ratio and product of the areas of the masked segmentation regions, the weight of the loss function for each region is determined, and a multi-segmentation region weight loss function is constructed. This function is then used to train the segmentation network, thereby improving the segmentation performance of smaller regions.

Benefits of technology

It improves the segmentation accuracy of regions such as the patella, femur, tibia and patellar tendon in knee X-ray images, and solves the problem of insufficient training caused by uneven region area.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122115474A_ABST
    Figure CN122115474A_ABST
Patent Text Reader

Abstract

The disclosure provides a loss function weight determination and network training method and system and an X-ray camera, and relates to the technical field of loss function weight determination. The loss function weight determination comprises the following steps: determining N mask segmentation area areas corresponding to mask label images in a set medical image training set corresponding to a set segmentation network; calculating N ratios of each mask segmentation area area to a total area of the N mask segmentation areas corresponding to the N mask segmentation area areas; and determining a loss function weight of each mask segmentation area corresponding to the set segmentation network based on N different products corresponding to any N-1 mask segmentation area areas in the N mask segmentation area areas. The embodiment of the disclosure can realize loss function weight determination.
Need to check novelty before this filing date? Find Prior Art