Endoscope Insertion Direction Detection Using Scene Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional endoscope insertion direction detecting devices rely on the order of extracted feature values, which can lead to inaccurate insertion direction detection, especially in complex body cavies like the large intestine, requiring skilled operators and increasing examination time.
Innovation Solution
An endoscope insertion direction detecting device and method that classify endoscope image features into essential classes, using scene feature values computed through higher-order local autocorrelation coefficients and discrete wavelet transforms, to accurately determine the insertion direction by displaying markers indicating the optimal insertion path.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional feature extraction methods are used in order, then the detection process is simple, but the insertion direction detection accuracy deteriorates in complex body cavities
Solution Approach 1:
The patent segments the detection process into two independent stages: first extracting multiple types of feature values (luminal structure, density gradient, light-dark direction) in parallel without strict ordering, then classifying and synthesizing these features. This segmentation allows each feature to be extracted independently and contributes to the final determination, improving accuracy in complex cavities while maintaining manageable complexity through modular processing.
Solution Approach 2:
The patent changes the parameter of feature extraction by simultaneously computing multiple feature types (luminal structure feature, density gradient feature, light-dark direction feature) rather than relying on a single feature or strict sequential extraction. This multi-parameter approach enables the system to adapt to various scene conditions and achieve higher detection accuracy in diverse anatomical structures.
2Measurement precision
If multiple feature values are extracted and classified, then the detection accuracy improves, but the processing time increases
Solution Approach 1:
The patent performs preliminary classification of extracted feature values into different scene types (e.g., fold scenes, lumen dark section scenes, others) before computing the final insertion direction. This preliminary action organizes the multiple feature values into categories, allowing the system to apply appropriate determination methods for each scene type, thereby reducing unnecessary processing and minimizing overall examination time while maintaining high accuracy.
3Productivity
If feature extraction order is fixed, then the processing is efficient, but the detection accuracy deteriorates when multiple structures are present
Solution Approach 1:
The patent implements a dynamic determination process where the system classifies the current scene type based on extracted features and dynamically selects the appropriate determination method. Rather than following a fixed extraction order, the system adapts its processing based on the detected scene characteristics (folds, lumen dark sections, or other structures), ensuring both processing efficiency and high detection accuracy for multiple coexisting structures.
Data Source
AI summary
An endoscope insertion direction detecting device includes a classification section for performing, on a scene of an endoscope image of a moving image picked up by an endoscope inserted into a body cavity, classification into classes for a plurality of different feature values relating to detection of an endoscope insertion direction in the body cavity; and an insertion direction computing section, provided for each of the classes for the feature values, into which the classification is performed, for computing an insertion direction of the endoscope.