Medical Image Diagnosis Assistance Using Style Learning Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current CAD systems for medical image analysis, despite using artificial neural networks, face challenges in ensuring accuracy and reliability due to the 'black box' nature of convolutional neural networks and the potential for unnecessary re-training based on user feedback without proper evaluation.
Innovation Solution
The proposed system separates a first artificial neural network for generating initial analysis results from a style learning model that receives and trains on user feedback. This allows for additional performance improvements based on user preferences without affecting the existing performance of the first neural network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional CAD systems use artificial neural networks for medical image analysis, then automated lesion detection capability is improved, but the black box nature of convolutional neural networks reduces reliability and accuracy assurance
Solution Approach 1:
The patent introduces an intermediary evaluation module that acts as a mediator between the CNN and the final diagnosis. This module evaluates the CNN's analysis results using multiple criteria (lesion detection accuracy, false positive rate, etc.) and only accepts results that meet predetermined thresholds, thereby ensuring reliability while maintaining automation.
Solution Approach 2:
The system implements feedback mechanisms where evaluation results are used to adjust and optimize the CNN model. The evaluation module provides continuous feedback on detection accuracy and false positive rates, enabling the system to improve reliability through iterative optimization while maintaining automated operation.
2Adaptability or versatility
If CAD systems re-train based on all user feedback, then adaptability to user preferences is improved, but unnecessary re-training occurs without proper evaluation reducing efficiency
Solution Approach 1:
The patent applies preliminary action by evaluating user feedback before initiating re-training. The evaluation module assesses whether feedback meets predetermined criteria (such as confidence thresholds or clinical significance) before the re-training process is triggered, preventing unnecessary re-training and improving efficiency.
Solution Approach 2:
The system implements selective feedback mechanisms where only evaluated and approved user feedback is used for re-training. The evaluation module filters feedback based on quality criteria, ensuring that only meaningful feedback triggers re-training, thus maintaining adaptability while improving productivity.
3Reliability
If user feedback is collected and processed for style learning, then user acceptance and diagnostic outcomes are improved, but system complexity increases due to separate style learning model
Solution Approach 1:
The patent applies segmentation by dividing the system into distinct functional modules: a CNN for lesion detection, an evaluation module for assessing results, and a style learning model for adapting to user preferences. This segmentation allows each module to specialize in its function, improving overall reliability and user acceptance while managing complexity through modular design.
Solution Approach 2:
The evaluation module serves as an intermediary that manages the interaction between user feedback and the style learning model. It filters, evaluates, and prepares feedback before passing it to the style learning model, thereby simplifying the overall system architecture while maintaining the benefits of user-adapted diagnostic outcomes.
Data Source
AI summary
Disclosed herein is an artificial neural network-based medical image diagnosis assistance apparatus for assisting in diagnosing a medical image based on a medical artificial neural network. A medical image diagnosis assistance apparatus according to an embodiment of the present invention includes a computing system, and the computing system includes at least one processor. The at least one processor is configured to acquire or receive a first analysis result obtained through the inference of a first artificial neural network about a first medical image, to detect user feedback on the first analysis result input by a user, to determine whether the first analysis result and the user feedback satisfy training conditions, and to transfer the first analysis result and the user feedback satisfying the training conditions to a style learning model so that the style learning model is trained on the first analysis result and the user feedback.