Liver image synthesis method based on plain scan CT to generate three-phase enhanced images

By using the Liver-GAN framework to generate realistic three-phase enhanced images from plain CT scans, the limitations of multi-phase enhanced CT examinations of the liver are overcome, high-quality virtual enhanced image reconstruction is achieved, radiation burden is reduced, and diagnostic reliability is improved.

CN122115229APending Publication Date: 2026-05-29THE AFFILIATED HOSPITAL OF GUIZHOU MEDICAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
THE AFFILIATED HOSPITAL OF GUIZHOU MEDICAL UNIV
Filing Date
2026-01-20
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Current multiphase enhanced CT scans of the liver are limited by factors such as contrast agent allergy, renal insufficiency, radiation dose, and economic costs, making it difficult to reconstruct high-quality multiphase enhanced images in patients who cannot or are not suitable for enhanced scanning.

Method used

A liver-guided generative adversarial network framework, Liver-GAN, was constructed to generate multi-phase enhanced CT images covering different liver tumor types using plain CT scans. A geometric correction network guided by a liver mask was used for affine and dense registration. A predictive enhancement residual generator was designed by combining U-Net and Transformer structures and supervised by a dual discriminator structure to generate virtual enhanced images of the arterial phase, portal venous phase, and delayed phase.

Benefits of technology

It enables the generation of realistic three-phase enhanced images from a single plain CT scan, reducing radiation burden, improving image quality and diagnostic reliability, and is suitable for patients who cannot undergo multi-phase enhanced scans, thus enhancing the cross-center availability and stability of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122115229A_ABST
    Figure CN122115229A_ABST
Patent Text Reader

Abstract

The application belongs to the technical field of liver images, and provides a liver image synthesis method for generating three-phase enhanced images based on plain CT, S1, an upper abdominal plain CT image of a patient to be processed is acquired; S2, a pre-trained liver mask guided corrector module is used to respectively perform affine transformation and dense deformation registration on real arterial phase, portal phase and delayed phase enhanced CT images matched with the plain CT image, to obtain registered enhanced images and corresponding liver masks in the liver region which are aligned with the plain CT image at the voxel level; S3, the plain CT image, a 2.5D input tensor formed by the plain CT image and its adjacent upper and lower slices, and the current layer plain image as a residual base are input into a pre-trained Liver-GAN generator, and an attention mechanism based on the liver mask is used to weight the features, and a predicted enhanced residual is output; S4, the enhanced residual and the residual base are added to generate a virtual arterial phase, portal phase or delayed phase enhanced image corresponding to the plain CT image.
Need to check novelty before this filing date? Find Prior Art