In-vivo Image Processing for Capsule Endoscopy Lesion Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The large number of images captured by capsule endoscopes during digestive tract examinations leads to inefficiencies in identifying lesion areas, as medical professionals spend significant time reviewing numerous images for abnormalities.
Innovation Solution
An image processing apparatus that acquires in-vivo images, calculates feature data, extracts body tissues based on predetermined thresholds, creates detection criteria, and detects lesions using hue and saturation values to improve the efficiency of identifying abnormal areas within the images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If clustering is performed by mapping pixel values to feature space based on color information, then lesion areas can be detected automatically, but the detection precision is insufficient due to inability to distinguish lesion areas from normal mucous membranes with similar colors
Solution Approach 1:
The patent segments the image processing into multiple stages: first extracting mucous membrane areas, then creating adaptive criteria based on their color distribution, and finally detecting lesions. This multi-stage segmentation allows the system to first identify the background tissue and then detect deviations from it, improving both precision and reliability by separating the detection of normal tissue from abnormal tissue.
Solution Approach 2:
The patent changes the parameter space by transforming RGB pixel values into hue and saturation components, and further normalizing them. This parameter transformation allows better separation of lesion areas from normal mucous membranes in the transformed space, even when they appear similar in original color space, thereby improving detection precision and reliability.
2Reliability
If a large number of in-vivo images are captured to ensure complete coverage of the digestive tract, then diagnostic completeness is improved, but the time required for observation increases significantly
Solution Approach 1:
The patent extracts and processes only the mucous membrane areas from the large number of captured images, rather than manually reviewing every single image. By automatically extracting relevant regions and applying detection criteria, the system maintains diagnostic completeness while dramatically reducing the time required for observation.
Solution Approach 2:
The patent performs preliminary processing of images by extracting mucous membrane areas and creating detection criteria before the actual lesion detection. This preliminary action prepares the data in advance, allowing for rapid and accurate lesion detection across all captured images, thus reducing overall observation time while maintaining completeness.
3Productivity
If fixed threshold values are used for lesion detection, then the detection process is simple and fast, but the detection precision varies across different body sites with different mucous membrane colors
Solution Approach 1:
The patent implements dynamic threshold values that are automatically adjusted based on the actual color distribution of mucous membranes in each image or region. Instead of using fixed thresholds, the system adapts the detection criteria to match the local tissue characteristics, maintaining high detection precision across different body sites while preserving computational efficiency.
Solution Approach 2:
The patent transforms the detection approach by changing from fixed parameter thresholds to adaptive parameters based on local color distribution statistics. By calculating hue and saturation ranges from actual mucous membrane data and using these as dynamic thresholds, the system achieves both speed and precision across diverse anatomical locations.
Data Source
AI summary
An image processing apparatus includes an image acquiring unit that acquires an in-vivo image being a captured image of an inside of a body cavity; a feature-data calculating unit that calculates feature data corresponding to a pixel or an area in the in-vivo image; a body-tissue extracting unit that extracts, as a body tissue, a pixel or an area whose feature data corresponds to a predetermined threshold; a criterion creating unit that creates a criterion for detecting a detecting object based on the feature data of the body tissue; and a detecting unit that detects a body tissue corresponding to the criterion as the detecting object.