A dual-energy CT image fusion method based on a generative adversarial network

By using a dual-energy CT image fusion method based on generative adversarial networks, the problems of traditional CT detectors being unable to distinguish material attenuation information under different energy levels and the complexity of image fusion methods are solved. The generated fused CT images retain the features of the source images, improve image details and accuracy, and reduce artifact interference.

CN115345807BActive Publication Date: 2026-03-17SHANDONG NON METALLIC MATERIAL RESEARCH INSTITUTE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210989919.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-17
Publication Date
2026-03-17
Estimated Expiration
2042-08-17

AI Technical Summary

Technical Problem

Traditional CT detectors cannot obtain information on material decay at different energies, resulting in beam hardening artifacts and loss of detail in the images. Existing image fusion methods are complex and suffer from severe information loss.

Method used

A dual-energy CT image fusion method based on generative adversarial networks is adopted. The generator G and two discriminators Dh and Dl are trained adversarially to generate a fused CT image containing information from both high-energy and low-energy CT images, thus avoiding complex fusion rule design and information loss.

Benefits of technology

The generated fused CT images retain the texture and brightness features of the source images, improve the detail and accuracy of the images, reduce artifact interference, and provide more complex and detailed image information.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115345807B_ABST
    Figure CN115345807B_ABST
Patent Text Reader

Abstract

This invention belongs to the field of CT technology and proposes a dual-energy CT image fusion method based on generative adversarial networks (GANs) for fusing low-energy and high-energy tomographic images under different X-ray exposure conditions. The GAN comprises a generator and two discriminators. The generator extracts detailed information from the CT images and generates random samples resembling real samples based on a preset loss to deceive the two discriminators. The discriminators distinguish the structural differences between the fused CT image and the two source images, determining whether the data is real or fake. Through continuous adversarial training between the generator and discriminators, and through end-to-end model training, the fusion model is constructed, finally generating a fused CT image containing both high and low energy information. This invention's dual-energy CT image fusion method based on GANs produces a fused dual-energy CT image with richer details, which is beneficial for further processing of the CT image.
Need to check novelty before this filing date? Find Prior Art