A method and system for estimating depth in a monocular endoscope based on instrument segmentation

By introducing instrument segmentation and depth estimation models, combined with pose and brightness estimation modules, the challenge of relative depth estimation between instruments and tissues in narrow-cavity surgical environments using monocular endoscopes was solved, achieving high-precision depth and relative depth estimation.

CN118644538BActive Publication Date: 2026-06-30BEIJING INST OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
BEIJING INST OF TECH
Filing Date
2024-06-13
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing technologies lack methods for estimating the relative depth of instruments and tissues for multiple tasks. Monocular endoscopic depth estimation is affected by humidification and light reflection in narrow surgical environments, and the presence of instruments can affect the accuracy of the algorithm.

Method used

By introducing an instrument segmentation model and a depth estimation model, an encoder-decoder structure is used for instrument segmentation and depth estimation. Combined with pose estimation and brightness estimation modules, a supervision signal is constructed for model training to calculate the relative depth between the instrument and the tissue.

Benefits of technology

It improves the accuracy of instrument segmentation and depth estimation, solves the depth estimation challenge in narrow cavity surgical environments, and provides accurate instrument-tissue relative depth estimation results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118644538B_ABST
    Figure CN118644538B_ABST
Patent Text Reader

Abstract

This invention provides a method and system for monocular endoscope depth estimation based on instrument segmentation. It can assist surgical robots in performing safe operations by calculating the relative depth between instruments and tissues. The specific process is as follows: using a trained instrument segmentation model, instruments are segmented in the dataset to be estimated in terms of depth, and the segmentation results are overlaid as a colored transparent layer mask onto the dataset; using the trained depth estimation model, the depth of the data in the overlaid mask dataset is estimated to obtain the estimated depth value.
Need to check novelty before this filing date? Find Prior Art