Medical image segmentation method, device and equipment based on generative large model, medium and program product

By generating label-free medical images based on a generative large model, and combining real label images in the semi-supervised training framework for training, the dependence on large-scale training sets and the implementation of semi-supervised methods in the existing technology is solved, and efficient medical image segmentation effect is achieved.

CN120219406APending Publication Date: 2025-06-27PEKING UNIV SHENZHEN GRADUATE SCHOOL
0 Cites 0 Cited by

Patent Information

Application Number
CN202510343107.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-21
Publication Date
2025-06-27

AI Technical Summary

Technical Problem

The existing medical image segmentation technology requires large-scale and high-quality training sets, and the implementation of semi-supervised methods lacks a unified framework, and cannot effectively utilize label-free data, resulting in insufficient model performance.

Method used

Using a generative big model-based method, a large model is generated through pre-trained targets to generate label-free medical images, and in a highly automated semi-supervised training framework, semi-supervised training is carried out in combination with real label images to improve the performance of medical image segmentation models.

Benefits of technology

By generating high-quality label-free medical images, the segmentation capability of the medical image segmentation model can be quickly improved, the robustness and generalization of the model can be enhanced, and the application effect on small-scale data sets can be improved.

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

Abstract

The invention relates to a medical image segmentation method and device based on a generative large model, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a real label medical image and a target semi-supervised demand; reading a label-free medical image according to the target semi-supervised training framework; wherein the label-free medical image is generated by a pre-trained target generation large model; performing semi-supervised training on the initial medical image segmentation model according to the target semi-supervised demand, the no-label medical image, the real-label medical image and the target semi-supervised training framework to obtain a target medical image segmentation model; and performing medical image segmentation on the real label medical image according to the target medical image segmentation model to obtain a medical image segmentation result. By adopting the method, the medical image generated by the target generation large model can be used as label-free data, and the performance of the segmentation model is improved under a highly automatic and robust semi-supervised training framework.
Need to check novelty before this filing date? Find Prior Art