AI Endomicroscope Image Conversion for Reliable Digital Biopsy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biopsy methods, particularly those using endomicroscopes, are time-consuming and prone to inaccuracies due to the need for pathologic slides and unfamiliar tissue images, leading to incomplete judgments by medical technicians.
Innovation Solution
An AI-based system that processes tissue images from endomicroscopes, considering conditions of image acquisition, converts them to resemble H&E staining, and provides accurate medical diagnosis assistance using AI models trained on specific conditions and pseudo-coloring techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If AI models output medical diagnosis assistance information for all input tissue images, then the quantity of diagnostic information is increased, but the effectiveness and reliability of the information decreases due to inclusion of ineffective data
Solution Approach 1:
The patent extracts and removes ineffective medical diagnosis assistance information from the AI model output. The processor identifies and filters out ineffective information based on predetermined criteria, providing only effective information to the medical technician. This resolves the contradiction by reducing quantity of information while improving its reliability and effectiveness.
2Productivity
If endomicroscope systems provide tissue images without H&E staining, then the productivity and speed of biopsy is improved, but the ease of operation and familiarity for medical technicians deteriorates due to different image types
Solution Approach 1:
The patent introduces an AI-based image conversion system as an intermediary that transforms endomicroscope images into H&E-staining-like images. This mediator maintains the speed advantage of digital biopsy while providing familiar visual characteristics that medical technicians are accustomed to, thus resolving the contradiction between productivity and ease of operation.
Solution Approach 2:
The patent applies color transformation techniques to convert the visual characteristics of endomicroscope images into those resembling H&E-stained images. By changing the color palette and visual properties of the digital images, the system maintains rapid digital biopsy capabilities while providing familiar image types that improve ease of operation and diagnostic familiarity.
3Device complexity
If AI models are trained without considering image acquisition conditions, then the device complexity and training process is simplified, but the measurement precision and accuracy of diagnosis assistance information deteriorates
Solution Approach 1:
The patent implements preliminary action by having the AI model learn and store optimal imaging conditions (such as laser wavelength, power, and other acquisition parameters) before actual diagnosis. The system pre-trains the AI with various image acquisition conditions so that during actual operation, the AI can select and apply the most appropriate conditions, thereby improving measurement precision without significantly increasing operational complexity.
Data Source
AI summary
Proposed is a method, apparatus, and system for providing medical diagnosis assistance information showing whether a tissue related to a tissue image is normal or abnormal (e.g., cancer) using the AI technology. Proposed are a method, an apparatus, and a system for training an AI model for producing medical diagnosis assistance information in consideration of conditions for obtaining a tissue image in an endomicroscope system for digital biopsy.