Learning Model Training for Pancreas Recognition in Scope-Assisted Surgery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In laparoscopic surgeries, particularly those involving the pancreas, it is difficult to accurately recognize and identify the pancreas portion from operative field images, leading to challenges in reporting this information to the operator.
Innovation Solution
A learning model is trained to recognize the pancreas portion in operative field images using a neural network-based approach, such as SegNet, which segments and identifies the pancreas by excluding blood vessels, fat, and shadows, and superimposes the recognized pancreas portion on the image for clear visualization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image display methods are used in laparoscopic surgery, then the surgical procedure can be performed, but the pancreas portion cannot be accurately recognized or identified from the operative field image
Solution Approach 1:
A learning model serves as an intermediary between the operative field image and the operator. The learning model processes the image data and generates segmented images that highlight the pancreas portion, acting as a mediator that transforms raw image data into meaningful surgical information without requiring direct human interpretation of complex medical images
Solution Approach 2:
The patent replaces manual image interpretation (mechanical/human process) with an automated learning model (information processing system). The learning model automatically segments and identifies the pancreas portion through computational algorithms, substituting the need for human visual analysis and reducing information loss through automated detection
2Measurement precision
If a learning model is trained to recognize pancreas portions, then accurate recognition and visualization can be achieved, but additional processing time and computational resources are required
Solution Approach 1:
The learning model is trained in advance using pre-collected training data before actual surgical use. This preliminary training phase allows the model to learn pancreas portion characteristics offline, so that during surgery, the model can quickly process images without requiring real-time training, thus reducing operational time loss
Solution Approach 2:
The system dynamically adjusts between training mode and inference mode. During surgery, the pre-trained model performs rapid inference on new images. The system can also adaptively update the model between surgical procedures or during idle periods, optimizing the balance between recognition accuracy and processing time based on operational needs
Data Source
AI summary
A non-transitory computer readable recording medium storing a computer program causing a computer to execute processing of acquiring an operative field image obtained by shooting an operative field of a scope-assisted surgery, and recognizing a pancreas portion in the acquired operative field image by inputting the acquired operative field image to a learning model trained to output information relevant to the pancreas portion included in the operative field image in accordance with input of the operative field image.