Multi-Model Tissue Recognition for Surgical Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pathology diagnosis methods require specialized knowledge and often rely on single-class teaching data, limiting their ability to perform multi-class inference and integrating recognition results from multiple learning models for accurate tissue identification during surgical procedures.
Innovation Solution
An information processing device and method that utilize multiple learning models to derive an integrative recognition result for operative field images, combining computation results from first and second learning models to provide accurate identification of tissues like loose connective tissue and nerve tissue, and outputting this information to assist surgeons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If single-class teaching data is used to create a learning model, then the model can be trained with specialized knowledge from medical specialists, but the model can only perform single-class inference and cannot integrate multiple types of tissue recognition
Solution Approach 1:
The patent divides the learning model into multiple independent single-class learning models, each trained on specific tissue types (e.g., loose connective tissue, nerve tissue). Each computation unit executes a separate learning model for a specific tissue class, allowing specialized knowledge to be maintained while enabling multi-class inference through integration of multiple models.
Solution Approach 2:
The patent merges multiple single-class learning models into an integrative recognition system. The derivation unit combines computation results from multiple learning models to produce an integrative recognition result, enabling the system to handle multiple tissue types simultaneously while maintaining the specialized capabilities of each individual model.
2Measurement precision
If multiple learning models are used to recognize different tissue types, then the recognition accuracy and versatility improve, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the recognition task into multiple independent computation units, each handling a specific tissue type. This allows parallel processing of different tissue types simultaneously, maintaining high recognition accuracy for each class while distributing the computational load to manage overall complexity.
Solution Approach 2:
The patent creates a universal learning model framework that can recognize multiple tissue types through a common architectural structure. The derivation unit serves as a universal integrator that combines results from any number of single-class models, providing multi-functionality without requiring separate specialized systems for each tissue type.
3Measurement precision
If specialized single-class learning models are trained for each tissue type, then the models can capture detailed tissue characteristics, but the system cannot provide integrative recognition results that combine multiple tissue identifications
Solution Approach 1:
The patent merges the output information from multiple single-class learning models through the derivation unit. This integrative recognition result combines the precise identifications of different tissue types into a unified output, preserving the detailed characteristics captured by each specialized model while providing comprehensive multi-class recognition.
Solution Approach 2:
The patent implements a feedback mechanism where the derivation unit processes and integrates results from multiple computation units, then feeds back an integrative recognition result. This feedback loop ensures that detailed tissue characteristics from individual models are synthesized into a coherent overall recognition that maintains precision while adding integrative value.
Data Source
AI summary
An information processing device, includes one or more processors and a storage storing instructions causing any of the one or more processors to execute processing of executing computation by a first learning model in accordance with input of an operative field image, executing computation by a second learning model in accordance with the input of the operative field image, deriving an integrative recognition result for the operative field image based on a computation result based on the first learning model and the second learning model, and outputting information based on the derived recognition result.