Endoscope Image Processing with Distance-Based Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical image processing devices struggle to accurately estimate the state of an observation region imaged by an endoscope due to variations in imaging distance, as the appearance of the image changes significantly with distance, leading to inconsistent detection and classification of lesions.
Innovation Solution
A medical image processing device that acquires both the observation image and distance information from the endoscope, allowing it to select appropriate estimation modes (near view or distant view) based on the imaging distance and weight-average estimation results from multiple learning models to improve accuracy, using specialized light images and learning models for distance estimation and weighting coefficients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single learning model is used to estimate the state of the observation region, then the device complexity is reduced, but the estimation accuracy deteriorates when imaging distance varies
Solution Approach 1:
The patent divides the estimation task into multiple segments by creating separate learning models for different imaging distances (near view model and distant view model). Each model is specialized for a specific distance range, allowing accurate estimation without requiring a single complex universal model. This segmentation resolves the contradiction by improving accuracy through specialization while keeping individual models relatively simple.
Solution Approach 2:
The patent changes the parameter of imaging distance to select appropriate learning models. By acquiring distance information and using it to switch between near view and distant view models, the system adapts to different imaging conditions. This parameter-based selection resolves the contradiction by matching model characteristics to specific distance parameters, ensuring accuracy without universal complexity.
2Measurement precision
If multiple learning models are used for different imaging distances, then the estimation accuracy improves, but the device complexity increases
Solution Approach 1:
The patent implements a dynamic model selection mechanism that automatically switches between near view and distant view learning models based on the acquired distance information. This dynamic adaptation allows the system to use the most appropriate model for each imaging condition, improving accuracy while managing complexity through automated rather than manual model management.
Solution Approach 2:
The patent introduces distance information as an intermediary element that mediates between the imaging system and the learning models. This intermediary parameter enables automatic model selection without requiring direct complex interactions between multiple models, simplifying the system architecture while maintaining the benefits of multiple specialized models.
3Reliability
If distance information is acquired and used for model selection, then the estimation consistency across different distances improves, but the ease of operation decreases
Solution Approach 1:
The patent implements a self-service mechanism where the system automatically acquires distance information and selects the appropriate learning model without user intervention. The system serves itself by autonomously adapting to different imaging distances, improving consistency while maintaining ease of operation through automation rather than requiring user management of multiple models.
Solution Approach 2:
The patent uses distance information as feedback to automatically adjust model selection. By continuously acquiring distance data and using it to select the most appropriate learning model, the system maintains consistent estimation accuracy across varying imaging conditions without requiring manual adjustment, thus preserving ease of operation while improving reliability.
Data Source
AI summary
Provided are a medical image processing device, an operation method of the medical image processing device, a medical image processing program, and a recording medium that are capable of accurately estimating a state of the observation region of an observation image captured by an endoscope.In the medical image processing device including a processor 22, an image acquisition unit 110 of the processor 22 acquires an observation image 100 in which an observation region in a body is imaged by an endoscope. A distance information-acquisition unit 112 of the processor 22 acquires (estimates) distance information regarding a distance between the endoscope and the observation region from the observation image 100. A state estimation unit 114 of the processor 22 estimates a state of the observation region based on the observation image 100 and the distance information.