A method and system for real-time correction of PET-CT brain metabolism image motion artifacts based on a large language model

By employing a real-time motion artifact correction method for PET-CT brain metabolic images based on a large language model, we have achieved automatic identification of single-frame motion, multimodal feature fusion, and closed-loop iterative optimization. This solves the problems of existing technologies that cannot identify single-frame motion in real time and rely on manual correction, thereby improving image quality and diagnostic efficiency.

CN122176005APending Publication Date: 2026-06-09SICHUAN CANCER HOSPITAL
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SICHUAN CANCER HOSPITAL
Filing Date
2026-05-06
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Current PET-CT brain metabolic imaging suffers from several problems, including the inability to identify single-frame motion in real time, reliance on manual correction leading to low efficiency, the risk of metabolic information loss due to single-modal registration, and a lack of closed-loop quality assessment and parameter iterative optimization.

Method used

A large language model-based approach is adopted, which uses frame-level motion recognition, multimodal feature fusion and hierarchical registration, combined with Dice similarity coefficient, gray value variance ratio and artifact retention index for quality assessment, and establishes a closed-loop iterative optimization mechanism to achieve automatic and accurate motion artifact correction.

Benefits of technology

It achieves real-time single-frame motion recognition, improves the spatial resolution and quantitative accuracy of images, reduces the loss of metabolic information, enhances correction efficiency and consistency, and adapts to different subjects and equipment scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122176005A_ABST
    Figure CN122176005A_ABST
Patent Text Reader

Abstract

This invention discloses a real-time correction method and system for motion artifacts in PET-CT brain metabolic images based on a large language model, belonging to the field of medical image processing technology. It acquires sequential images at a fixed frame rate, extracts key CT anatomical points and PET metabolic hotspot features; calculates the multimodal feature matching degree between the current frame and the reference frame based on a large language model, and identifies translational, rotational, or locally deformed motion frames; uses ICP registration for rigid motion and rigid-elastic hybrid registration for local deformation; calculates the Dice similarity coefficient, grayscale variance ratio, and artifact persistence index for quality assessment; and updates the identification threshold and registration parameters in a closed loop based on historical correction data. This invention achieves automatic real-time identification and hierarchical correction of single-frame motion, integrates multimodal features to avoid metabolic information loss, and has parameter iterative optimization capabilities, significantly improving correction accuracy and clinical examination efficiency.
Need to check novelty before this filing date? Find Prior Art