Endoscope Control System for Stenosis Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current endoscope control systems face difficulties in automatically navigating through complex or stenosed regions of the gastrointestinal tract due to challenges in accurately determining the center of the lumen and avoiding obstacles, leading to increased insertion difficulty and potential harm to the patient.
Innovation Solution
An endoscope control system that utilizes machine learning to classify endoscopic images and determine appropriate operation details, combining an operation selection model for straightforward cases with algorithms for complex scenarios, such as bending portions, stenosis, and diverticula, to automatically control the endoscope's movement and safely pass through challenging areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning operation selection model is used to determine operation details, then automation capability is improved, but reliability deteriorates in complex scenarios such as bending portions, stenosis, and diverticula
Solution Approach 1:
The control system segments the endoscopic image analysis into two distinct pathways: a machine learning-based operation selection model for straightforward cases and a rule-based algorithm for complex scenarios. This segmentation allows each pathway to be optimized for its specific domain, with the ML model handling routine tasks and the algorithm providing reliable control in challenging anatomical regions such as bending portions, stenosis, and diverticula.
Solution Approach 2:
The system changes the control parameter selection based on the complexity of the endoscopic scene. When the image analysis detects complex structures (bending portions, stenosis, diverticula), the system switches from using ML-based operation probabilities to using deterministic algorithm-based operation details. This parameter change ensures reliable control by selecting the appropriate control method based on scene complexity.
2Reliability
If rule-based algorithm is used to determine operation details, then reliability is improved for complex scenarios, but productivity deteriorates due to increased computation time
Solution Approach 1:
The system segments the image processing workload based on scene complexity. Straightforward endoscopic scenes are processed quickly using the machine learning model, while only complex scenes trigger the more computationally intensive algorithm. This segmentation of processing tasks optimizes productivity by avoiding unnecessary heavy computation in routine cases while ensuring reliable analysis when needed.
Solution Approach 2:
The system applies partial action by using the simpler machine learning approach for the majority of straightforward cases, reserving the more comprehensive algorithm only for complex scenarios. This partial application of the heavy algorithm avoids excessive computation time while maintaining reliable control where required.
3Productivity
If automatic insertion control is implemented, then productivity is improved, but object-affected harmful factors increase due to potential harm to patients
Solution Approach 1:
The control system incorporates feedback mechanisms where the algorithm continuously monitors the endoscopic image and the distal end position of the endoscope. When the algorithm detects that the distal end is approaching a bending portion, stenosis, or diverticulum, it provides feedback to adjust the operation details, thereby preventing harmful contact with the patient's tissue while maintaining efficient automatic insertion control.
Solution Approach 2:
The system applies beforehand cushioning by using the algorithm to predict and prepare for potential harmful situations before they occur. When complex structures are detected in advance, the algorithm proactively adjusts the insertion path and operation details to avoid contact with the patient's tissue, cushioning against potential harm before it can occur during the insertion process.
Data Source
AI summary
An image acquisitor acquires an image of a lumen imaged by an endoscope. An image classifier classifies the acquired lumen image as one of a plurality of types. When the lumen image is classified as a first type, an operation detail determinator determines an insertion operation detail of the endoscope using an insertion operation selection model generated by machine learning. When the lumen image is classified as a second type, the operation detail determinator determines the insertion operation detail of the endoscope using an algorithm for determining the insertion operation detail.