CBCT (Cone Beam Computed Tomography) head shadow survey mark point automatic detection method based on multi-geometric guidance and specific perceptual coding

By employing a multi-geometric guidance and specific perceptual coding approach, combined with a shared encoder and a low-rank adapter to extract visual features, constructing a multi-anatomical heterogeneous map and performing iterative optimization, the accuracy and consistency issues of landmark detection in CBCT cephalometrics were resolved, achieving efficient and automated landmark detection.

CN122335671APending Publication Date: 2026-07-03ZHEJIANG UNIV OF TECH
0 Cites 0 Cited by

Patent Information

Application Number
CN202610307013.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-13
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing technologies lack explicit geometric modeling in CBCT cephalometrics, resulting in insufficient accuracy and consistency in landmark detection. Traditional methods are time-consuming, labor-intensive, and highly subjective, while deep learning methods suffer from error propagation problems when dealing with complex anatomical structures.

Method used

We employ a multi-geometric guidance and specific perceptual coding approach, extracting local visual features through a shared basic encoder and a low-rank adaptive adapter, combining multi-anatomical heterogeneous maps and Transformer for geometric constraints, and utilizing iterative closed-loop optimization to improve the accuracy of marker detection.

Benefits of technology

It significantly improves the accuracy and robustness of landmark detection, realizes high-precision automatic detection of three-dimensional landmarks, and solves the problems of time-consuming, labor-intensive and subjective methods in traditional methods, providing a high-precision automated solution for orthodontic diagnosis and treatment planning.

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

Abstract

A CBCT (cone beam computed tomography) head shadow survey mark point automatic detection method based on multi-geometric guidance and specific perceptual coding comprises the following steps: step S1, down-sampling a three-dimensional CBCT image, and obtaining a preliminary coordinate set through a coarse positioning network; s2, cutting an image block by taking the coordinate as a center, and extracting local features through a visual encoder comprising a shared basic encoder and a low-rank adapter; s3, calculating a relative position matrix of the mark points, coding a spatial relationship by using a radial basis function, and constructing a multi-anatomical heterogeneous graph; s4, inputting the vision and edge features into a multi-geometric guidance Transform, fusing global constraints by using an attention mechanism, and updating the features; s5, extracting a directional geometrical relationship by using a spherical harmonic function to construct a high-order update item, and dynamically updating the heterogeneous graph in combination with a gating residual mechanism; and S6, predicting a coordinate offset through a multi-layer network, executing iterative optimization, and outputting a high-precision three-dimensional coordinate. According to the method, the detection accuracy and robustness are remarkably improved.
Need to check novelty before this filing date? Find Prior Art