Learning Assistance Device for Automated Medical Data Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the medical field, collecting high-quality learning data with correct answer data for deep learning is challenging due to confidentiality concerns and the difficulty in associating correct answers with image data, making it inefficient and costly to prepare sufficient data for AI models.
Innovation Solution
A learning assistance device and system that acquire and aggregate discrimination results from multiple terminal devices to determine correct answer data, enabling the generation of new learning discriminators through iterative learning processes, thereby automatically creating labeled data for deep learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If correct answer data is manually collected and associated with image data in medical institutions, then data quality for deep learning is improved, but the cost and time required for data preparation increases significantly
Solution Approach 1:
The system enables self-service by allowing terminal devices to automatically generate correct answer data through their own discriminators. Each terminal device uses its discriminator to process image data and generate discrimination results, which are then aggregated to form correct answer data without requiring manual annotation, thus eliminating the time-consuming manual data preparation process while maintaining high data quality
Solution Approach 2:
The system implements feedback mechanisms where discrimination results from multiple terminal devices are aggregated and used to determine correct answer data, which is then fed back to improve the discriminators. This continuous feedback loop enables automatic refinement of both the correct answer data and the discriminators themselves, improving data quality without additional manual intervention time
2Quantity of substance
If a large amount of learning data is collected from multiple medical institutions, then the variety and quantity of learning data for deep learning is improved, but data confidentiality and security requirements increase
Solution Approach 1:
The system extracts only the necessary discrimination results from the terminal devices without requiring access to the original image data or patient information. By taking out only the processed discrimination outcomes rather than the raw medical data, the system enables data collection from multiple institutions while maintaining confidentiality and security requirements
Solution Approach 2:
The learning assistance device acts as an intermediary that aggregates discrimination results from multiple terminal devices without directly accessing the source medical data. This intermediary role allows the system to collect large amounts of learning data from multiple institutions while maintaining data confidentiality, as the intermediary only handles the processed results rather than the sensitive original data
3Measurement precision
If manual annotation of correct answer data is performed for each image, then data accuracy is improved, but the productivity and efficiency of data preparation decreases
Solution Approach 1:
The system merges discrimination results from multiple terminal devices to determine correct answer data. By combining multiple independent discrimination outcomes, the system achieves high accuracy through aggregation while maintaining high productivity, as the process is fully automated and parallelizable across multiple devices without requiring sequential manual annotation
Solution Approach 2:
The terminal devices perform self-service by automatically generating discrimination results that serve as correct answer data. This eliminates the need for manual annotation entirely, achieving both high accuracy through the discriminator's processing capability and high productivity through automation, with multiple devices working in parallel to prepare data efficiently
Data Source
AI summary
A learning assistance device acquires a plurality of learned discriminators obtained by causing learning discriminators provided in a plurality of respective terminal devices to perform learning using image correct answer data, acquires a plurality of discrimination results obtained by causing a plurality of learned discriminators to discriminate the same input image, determines the correct answer data of the input image on the basis of the plurality of discrimination results, causes the discriminator to perform learning the input image and the correct answer data, and outputs a result thereof as a new learning discriminator to each terminal device.