Endoscope Viewpoint Tracking with Autonomous Failure Stop
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing endoscopy systems face challenges in accurately determining the viewpoint of an endoscope when unknown structure regions are present in the endoscopic image, leading to potential incorrect tracking directions for medical workers.
Innovation Solution
An information processing apparatus that estimates the trackability of the endoscope's viewpoint using a three-dimensional model and machine learning models to autonomously stop tracking when failure is detected, and provides virtual endoscopic images for guidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If viewpoint tracking is performed continuously using endoscopic images, then navigation support is provided to medical workers, but incorrect tracking direction may be notified when unknown structure regions are present in the endoscopic image
Solution Approach 1:
The estimation unit performs preliminary assessment of viewpoint trackability before the tracking unit executes viewpoint tracking. By evaluating whether the endoscopic image contains sufficient recognizable structures in advance, the system determines whether tracking should proceed, thereby preventing incorrect tracking directions caused by unknown structure regions while maintaining efficient navigation support.
2Reliability
If viewpoint tracking is stopped when trackability is low, then incorrect tracking direction is prevented, but navigation support is interrupted
Solution Approach 1:
The system implements a feedback mechanism where the estimation unit continuously evaluates viewpoint trackability based on endoscopic image analysis, and the controller adjusts tracking operations accordingly. When trackability is sufficient, tracking proceeds to provide navigation support; when trackability is insufficient, tracking is stopped to prevent errors. This feedback loop ensures both reliability and efficient operation by adapting to real-time image quality.
3Measurement precision
If three-dimensional model and machine learning models are used to estimate trackability, then tracking accuracy is improved, but processing time and computational load increase
Solution Approach 1:
The estimation unit performs partial analysis of endoscopic images to assess viewpoint trackability, focusing on identifying whether sufficient recognizable structures are present rather than performing complete three-dimensional reconstruction. This partial action approach provides sufficient accuracy for tracking decisions while significantly reducing processing time and computational load compared to full model processing.
Data Source
Figure 1
Figure 2~3
Figure 4~6
AI summary
Tracking of a viewpoint of an endoscope is autonomously stopped based on an estimation result of the viewpoint of the endoscope. An information processing apparatus estimates whether or not a viewpoint of an endoscope is trackable from an endoscopic image, tracks the viewpoint of the endoscope using a virtual endoscopic image and a pseudo virtual endoscopic image generated from the endoscopic image in a case where it is estimated that the viewpoint of the endoscope is trackable, determining a tracking result of the viewpoint of the endoscope using the endoscopic image, and a tracking destination virtual endoscopic image and a tracking destination virtual depth image generated from a bronchial model based on the tracked viewpoint of the endoscope, and stops the tracking of the viewpoint of the endoscope until a tracking start instruction is notified in a case where it is determined that the tracking has failed.