Endoscopic Lesion Classification With Adaptive Image Thresholding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing endoscopic image classification methods struggle with accurately identifying lesion parts, especially when using a single image, and determining the optimal number of images for classification is challenging.
Innovation Solution
An image processing device and method that acquires endoscopic images, calculates scores for candidate classes using a Sequential Probability Ratio Test (SPRT) based approach, and classifies lesion parts based on reaching a threshold value, allowing for a variable number of images to be used for classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single endoscopic image is used for classification, then the processing time is short, but the classification accuracy is insufficient for difficult-to-identify lesion parts
Solution Approach 1:
The system dynamically adjusts the number of images used for classification based on the accumulated score from the sequential probability ratio test. When the score reaches a threshold, classification is performed; otherwise, more images are acquired. This dynamic approach allows the system to use more images for difficult cases (improving accuracy) while maintaining quick classification for easier cases (reducing processing time).
Solution Approach 2:
The sequential probability ratio test provides continuous feedback through score accumulation as images are processed. This feedback mechanism allows the system to monitor classification confidence in real-time and determine when sufficient evidence has been gathered, enabling adaptive termination of the classification process based on accumulated information quality.
2Measurement precision
If multiple endoscopic images are used for classification, then the classification accuracy improves, but it is difficult to determine the appropriate number of images to use
Solution Approach 1:
The system uses the score from the sequential probability ratio test as a dynamic parameter to determine when to stop acquiring images. Instead of fixing the number of images in advance, the termination condition is based on the score reaching a predetermined threshold, which automatically adapts to the difficulty of each classification case without requiring complex manual configuration.
3Productivity
If a fixed number of images is used for classification, then the processing flow is simple, but the classification performance cannot be optimized for different lesion types
Solution Approach 1:
The system transitions from static fixed-image classification to dynamic variable-image classification using the sequential probability ratio test. The number of images used is dynamically determined by when the accumulated score reaches the threshold, allowing optimal performance for each case while maintaining a relatively simple processing flow through automated score-based termination.
Data Source
AI summary
The image processing device 1X includes an acquisition means 30X, a score calculation means 31X, and a classification means 32X. The acquisition means 30X is configured to acquire an endoscopic image in which an examination target is photographed by a photographing unit provided in an endoscope. The score calculation means 31X is configured to calculate scores on likelihoods of respective candidate classes corresponding to types of lesion, wherein the candidate classes are candidates for classification of an image group of the acquired endoscopic image. The classification means 32X is configured to perform the classification of the image group upon determining that at least one of the scores has reached a threshold value.