Endoscopic Image Model Switching by Anatomical Region
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing endoscopic image processing systems face challenges in maintaining accuracy due to the use of a single machine learning model for diverse anatomical regions, leading to potential low accuracy in computer-aided detection and diagnosis, as the color tone and lesion shape vary across regions.
Innovation Solution
An endoscopic image processing apparatus that selects and switches machine learning models based on the current image pickup region, using a predetermined time interval to ensure accurate and efficient model switching, either reverting to the previous model if the interval is short or advancing to the next model if the interval is long, thereby optimizing model selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single machine learning model is trained using images of various regions, then the device complexity is reduced, but the manufacturing precision (accuracy) deteriorates
Solution Approach 1:
The patent segments the machine learning model into multiple region-specific models (pharynx model, esophagus model, stomach model, duodenum model). Each model is trained independently on images from its corresponding anatomical region, allowing specialized accuracy for each region while maintaining manageable complexity through automated selection based on image pickup position.
2Manufacturing precision
If multiple region-specific machine learning models are prepared, then the CAD accuracy is improved, but the device complexity increases
Solution Approach 1:
The system implements self-service through automated model selection. The image pickup position detection unit automatically identifies the current anatomical region, and the control unit automatically selects and switches between the appropriate machine learning model without requiring manual intervention from the endoscopist, thereby managing the complexity of multiple models transparently.
Solution Approach 2:
The machine learning apparatus is designed with multi-functionality to handle multiple anatomical regions (pharynx, esophagus, stomach, duodenum) using a single integrated system. The control unit can universally manage and switch between different region-specific models based on the detected image pickup position, making the system adaptable to various examination scenarios.
3Measurement precision
If manual switching of machine learning models is required, then the model selection precision is improved, but the loss of time increases
Solution Approach 1:
The patent replaces the mechanical/manual switching system with an automated control system. Instead of requiring the endoscopist to manually select models based on their knowledge of the current anatomical region, the system uses image pickup position detection (via sensors or image analysis) to automatically trigger model switching, eliminating manual intervention and reducing time loss.
4Productivity
If automatic model switching is implemented, then the productivity is improved, but the measurement precision (model selection accuracy) may deteriorate
Solution Approach 1:
The system implements feedback through the image pickup position detection unit that continuously monitors the endoscope's location in the anatomical tract. This real-time position information feeds back to the control unit, which automatically selects the appropriate machine learning model, ensuring accurate model selection that matches the current anatomical region being examined, thereby maintaining precision while improving productivity.
Data Source
AI summary
One or more processors select a model from among a plurality of types of machine learning models according to a region whose image is being picked up by an endoscope, generate notification information of a type of the model selected, receive an instruction signal for switching the model, measure a time interval from a selection of the model to a reception of the instruction signal, select a model selected immediately previously when the time interval is less than a first predetermined time, and select a model scheduled to be selected immediately subsequently when the time interval is equal to or greater than a second predetermined time.