A method for judging the concave-convex shape of colon polyps based on physical light consistency

What is AI technical title?
AI technical title is built by Patsnap AI team. It summarizes the technical point description of the patent document.
By combining deep learning and physical illumination consistency methods, a depth inversion hypothesis testing mechanism was constructed, which solved the problem of misjudgment of concave and convex shapes in monocular colonoscopy imaging. This enabled accurate discrimination of concave and convex shapes and surgical navigation warnings without increasing hardware costs, thereby improving the safety and accuracy of endoscopic surgery.

CN122289148APending Publication Date: 2026-06-26HANGZHOU DIANZI UNIV +1

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: HANGZHOU DIANZI UNIV
Filing Date: 2026-03-13
Publication Date: 2026-06-26

AI Technical Summary

Technical Problem

Existing monocular colonoscopy imaging technology suffers from misjudgment in the identification of concave and convex shapes, resulting in a high misdiagnosis rate and making it difficult to achieve accurate identification of concave and convex shapes without increasing hardware costs.

Method used

A method based on physical illumination consistency is adopted. Polyp regions are extracted through a deep learning segmentation model and a monocular depth estimation model. A depth inversion hypothesis testing mechanism is constructed using a physical model of near-field point light sources. Concave and convex shapes are determined by combining photometric consistency loss, and auxiliary information is provided by augmented reality technology.

Benefits of technology

It improves the safety and accuracy of minimally invasive endoscopic surgery, reduces the risk of misdiagnosis, and enables accurate identification of concave and convex shapes and surgical navigation warnings without increasing hardware costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122289148A_ABST

Patent Text Reader

Abstract

This invention proposes a method for determining the convexity / concave morphology of colonic polyps based on physical illumination consistency. First, a real-time video stream is processed. Using a deep learning segmentation model and a monocular depth estimation model, binary masks of the polyp region in each frame are extracted, and the initial depth map of that frame is predicted. Then, a depth inversion algorithm is used to generate a reversed depth map with the opposite geometry to the initial depth map. Based on an endoscope coaxial light source illumination imaging model, two virtual brightness images are rendered using the initial and reversed depth maps respectively. Finally, the convexity / concave morphology is determined by calculating the photometric consistency loss between these two virtual images and the real colonoscope grayscale image: if the re-rendering loss corresponding to the original depth is lower, the original prediction is considered correct; if the loss corresponding to the reversed depth is lower, the original prediction is considered to have a convexity / concave inversion error. This invention can solve the ambiguity problem of convexity / concave morphology in monocular endoscopes without increasing hardware costs.

Need to check novelty before this filing date? Find Prior Art