Convolutional Neural Network Lesion Detection in Endoscopy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current endoscopic diagnosis for gastric cancer is hindered by high false-negative rates and the need for extensive experience and time, as well as subjective determination, which can lead to fatigue-related accuracy reduction in endoscopists.
Innovation Solution
A diagnostic imaging support apparatus utilizing a convolutional neural network to estimate lesion names and locations in endoscopic images, providing a certainty score and overlaying analysis results onto the images, trained on features like atrophy, intestinal metaplasia, mucosal swelling, and mucosal color tones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If endoscopists perform double-checking of endoscopic images to manage diagnostic accuracy, then diagnostic reliability is improved, but time consumption and operational load increase significantly
Solution Approach 1:
An AI-based diagnostic support system acts as an intermediary between the endoscopic image and the endoscopist. The system automatically analyzes endoscopic images to detect gastric cancer lesions, providing diagnostic assistance that maintains high accuracy while eliminating the need for time-consuming manual double-checking by endoscopists.
Solution Approach 2:
The patent replaces the mechanical process of manual double-checking by endoscopists with an automated AI-based image analysis system. This substitution uses machine learning algorithms to process and interpret endoscopic images, significantly reducing time consumption while maintaining or improving diagnostic reliability.
2Measurement precision
If endoscopists rely on subjective determination based on experience and observation, then diagnostic capability can be developed through training, but false-positive and false-negative decisions increase due to human limitations
Solution Approach 1:
The AI system creates a digital model or copy of the diagnostic reasoning process by training on large datasets of endoscopic images with known outcomes. This digital copy can consistently apply learned patterns to new images without the variability and fatigue that affect human endoscopists, improving both precision and reliability.
Solution Approach 2:
The system transforms subjective diagnostic parameters into objective, quantifiable metrics through automated image analysis. By converting visual assessment into measurable data points and applying consistent algorithmic thresholds, the system eliminates the subjectivity and inconsistency inherent in human-based diagnostic parameters.
3Measurement precision
If extensive training with 10,000 images and 10 years of experience is provided to endoscopists, then diagnostic capability is improved, but the time and resource investment required is extremely high
Solution Approach 1:
The AI system performs self-learning by automatically training on large datasets of labeled endoscopic images without requiring human instructors or structured training programs. This self-service approach to acquiring diagnostic capability eliminates the extremely long training period required for human endoscopists while achieving comparable or superior detection accuracy.
Solution Approach 2:
The diagnostic support system performs preliminary analysis of endoscopic images automatically, providing pre-assessment results before the endoscopist makes the final diagnosis. This preliminary action reduces the burden on endoscopists and allows them to focus on cases that require human judgment, effectively reducing the practical training time needed.
Data Source
AI summary
Provided are: an image diagnosis assistance apparatus capable of assisting diagnosis of an endoscopic image captured by an endoscopist; a data collection method; an image diagnosis assistance method; and an image diagnosis assistance program. The image diagnosis assistance apparatus is provided with: a lesion assessment unit that assesses, by a convolutional neural network, the denomination and the position of a lesion which is present in a digestive system endoscopic image of a patient captured by a digestive system endoscopic imaging device and information about accuracies thereof; and a display control unit that performs control for generating an analysis result image in which the denomination and the position of the lesion and the accuracies thereof are displayed and for displaying the image on the digestive system endoscopic image.