AI Colonoscopy Detection via Vascular Bed Disconnection Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting colon polyps during colonoscopy are inefficient due to the difficulty in distinguishing abnormal tissue from normal tissue, especially for thin and flat polyps, leading to potential misdiagnosis and missed cases, even for skilled surgeons, due to the complexity of the colon's anatomy and limitations in image resolution and equipment costs.
Innovation Solution
An AI-based method and device using deep learning to recognize patterns in colonic mucosa and blood vessels, specifically identifying disconnected vascular beds to detect thin, flat, and thin-film planar polyps by displaying distinct visual effects, allowing for accurate detection even by less proficient colonoscopy testers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human experts manually examine colonoscopy images to detect polyps, then diagnostic accuracy may be high for visible abnormalities, but detection precision for thin and flat polyps deteriorates due to the difficulty of distinguishing them from normal tissue
Solution Approach 1:
The patent introduces an AI-based deep learning system as an intermediary between the colonoscopy image and the human expert. The system processes the image to generate a processed image with enhanced visual effects that highlight suspicious areas, making thin and flat polyps more detectable. This intermediary processing step bridges the gap between the original image and human detection capabilities.
Solution Approach 2:
The patent applies color changes and visual effect modifications to the colonoscopy image to enhance the visibility of polyps. The deep learning system generates processed images with altered color patterns and visual effects that make thin and flat polyps distinguishable from normal tissue, directly addressing the detection difficulty.
2Measurement precision
If high-resolution colonoscopy image reading monitors are used to improve detection accuracy, then measurement precision improves, but device complexity and cost increase
Solution Approach 1:
The patent replaces the mechanical/optical enhancement approach (high-resolution monitors and image enhancement equipment) with an intelligent software-based deep learning system. The AI processing unit performs image analysis and generates enhanced visual representations through computational algorithms, substituting complex hardware requirements with software intelligence.
Solution Approach 2:
The patent changes the parameters of the image data through deep learning processing. The system transforms the original image parameters (color, contrast, visual effects) to generate processed images that enhance polyp visibility, achieving high detection accuracy through parameter transformation rather than hardware upgrades.
3Reliability
If skilled surgeons manually analyze colonoscopy images, then diagnostic reliability may be maintained, but productivity deteriorates due to the time required to examine large amounts of information
Solution Approach 1:
The patent implements preliminary action by having the deep learning system process and analyze the colonoscopy image before the human expert examines it. The AI generates processed images with highlighted suspicious areas in advance, preparing the information in a format that facilitates faster and more reliable human review, thus improving both productivity and maintaining reliability.
Solution Approach 2:
The patent incorporates feedback mechanisms where the deep learning system's processed image output is fed back to the human expert for final diagnosis. The system provides computational analysis results as feedback to guide human decision-making, combining AI efficiency with human reliability in a collaborative diagnostic workflow.
Data Source
AI summary
A method of detecting colon polyps disclosed in the present disclosure includes: (a) receiving an image captured by an endoscope inserted into colon of a test subject; (b) recognizing each image section including colonic mucosa and colonic blood vessels in the image; (c) determining whether a colonic vascular bed is disconnected in each image section; (d) displaying a first visual effect representing a vascular bed in which the colonic blood vessels are disconnected; and (e) displaying a second visual effect representing a continuous vascular bed of the colonic blood vessels, wherein operation (b) is configured to recognize each image section through a deep learning model, which is machine learned based on blood vessel data in a plurality of colonic images of the test subject obtained from external annotators, and a degree of disconnection of the vascular bed and a blood vessel pattern.