Intestinal polyp size measurement method and system based on monocular depth vision

By using a multi-model collaborative pipeline based on monocular depth vision, the subjectivity and accuracy issues in polyp size measurement are resolved, enabling automated, real-time, and high-precision measurement of polyp physical dimensions, which is suitable for endoscopic video streams.

CN122368153APending Publication Date: 2026-07-10CHENGDU TIANXING HUICHUANG TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHENGDU TIANXING HUICHUANG TECHNOLOGY CO LTD
Filing Date
2026-04-16
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing methods for measuring polyp size are highly subjective, have low accuracy, and lack automation. Monocular endoscopic images lack direct depth information, leading to inaccurate pixel size conversion.

Method used

A multi-model collaborative pipeline based on monocular depth vision is adopted, including depth estimation, object detection, instance segmentation and machine learning regression. Through geometric fitting and distance transformation strategies, the physical size of polyps can be measured automatically and with high precision.

Benefits of technology

It enables automatic, real-time, and high-precision measurement of polyp size in endoscopic video streams, reducing human error and improving measurement stability and adaptability, making it suitable for complex endoscopic scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122368153A_ABST
    Figure CN122368153A_ABST
Patent Text Reader

Abstract

This invention belongs to the interdisciplinary field of medical devices and artificial intelligence, and discloses a method and system for measuring the size of intestinal polyps based on monocular depth vision. The method includes: preprocessing endoscopic keyframe images; sequentially generating a depth map using a monocular depth estimation model; locating the polyp using a target detection model; and generating a pixel-level segmentation mask using an instance segmentation model. Next, the polyp body and tail are separated based on the distance distribution analysis from the contour points to the centroid, and the pixel diameter is obtained by geometric fitting of the body contour. Then, edge depth values ​​are sampled in the target's internal region using a contour inward contraction method based on distance transformation. Finally, the edge depth and pixel diameter are used as features input to a pre-trained machine learning regression model to predict the millimeter / pixel ratio and calculate the physical size of the polyp. This invention achieves fully automated and high-precision measurement of polyp size, effectively solving the problems of high subjectivity and low accuracy in existing technologies.
Need to check novelty before this filing date? Find Prior Art