Endoscopic Capsule Image Quality Classifier Using Digital Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating the visualization quality of intestinal mucosa during video capsule endoscopy lack efficiency and reproducibility due to the large number of images and variability in image quality, with no standardized 'ground truth' for assessing image quality, leading to inconsistent diagnostic outcomes.
Innovation Solution
A digital image classifier is developed using automatic statistical learning to evaluate the quality of endoscopic videocapsule images by extracting digital parameters such as colorimetric and textural features, allowing for the classification of images into 'adequate' or 'inadequate' visualization categories, improving reproducibility and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual evaluation methods are used to assess visualization quality, then flexibility in clinical judgment is maintained, but reproducibility and efficiency are poor due to the large number of images and subjectivity
Solution Approach 1:
The patent replaces manual visual evaluation with an automated digital image classifier that uses machine learning algorithms to assess visualization quality. The system extracts digital parameters (colorimetric, textural, brightness) and automatically classifies images, eliminating human subjectivity and improving both reproducibility and efficiency.
Solution Approach 2:
The patent creates a digital copy of the evaluation process through a trained machine learning model. The model learns from training images and reproduces the evaluation consistently on test images, ensuring reproducible results that match expert clinical judgment without requiring repeated manual assessment.
2Productivity
If automated digital parameters are extracted, then efficiency and reproducibility are improved, but the complexity of the system increases
Solution Approach 1:
The patent segments the evaluation process into distinct stages: image acquisition, extraction of digital parameters (colorimetric, textural, brightness), classification using machine learning, and quality assessment. This segmentation allows each component to be optimized independently and simplifies the overall system architecture.
Solution Approach 2:
The patent introduces digital parameter extraction as an intermediary between the raw images and the final quality assessment. This intermediary layer transforms complex visual data into quantifiable features that the machine learning model can process, simplifying the connection between image data and evaluation results.
3Reliability
If a standardized classification system is implemented, then reproducibility is improved, but the ability to capture nuanced clinical judgment is reduced
Solution Approach 1:
The patent changes the evaluation parameters from subjective visual assessment to objective digital measurements (colorimetric ratios, textural features, brightness values). These parameter transformations enable standardized, reproducible classification while maintaining the ability to capture clinically relevant nuances through multiple measured dimensions.
Solution Approach 2:
The patent combines multiple types of digital parameters (colorimetric, textural, brightness) into a composite assessment framework. This composite approach integrates diverse information sources to create a comprehensive evaluation that maintains clinical relevance while achieving standardized, reproducible results through multi-parameter analysis.
Data Source
Figure 1A~1J
Figure 2
Figure 3
AI summary
The invention relates to a method for producing a numerical classifier of images, so as to automatically determine the viewing quality of the endoscopic videocapsule images of segments of the digestive tube, comprising a step of video acquisition in the digestive tube by a videocapsule; a step of extracting images of the video; a so-called "ground truth" step of clinically evaluating the viewing quality of the images on the basis of medical criteria, a step of selecting an initial set of images with "suitable" viewing and of images with "unsuitable" viewing, a step of calculating at least one numerical parameter which deals with at least one of the medical criteria, a step of automatic statistical learning to produce the numerical classifier.