Federated Learning Data Allocation for Consistent Medical AI Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In federated learning for medical AI development, variations in learning data across different medical institutions lead to inconsistencies in inference accuracy due to random data sampling, lacking a systematic method for data collection beyond random sampling.
Innovation Solution
A machine learning system and method that includes a plurality of client terminals and an integration server, where the integration server designates search conditions, totals data possession, and distributes learning data types and quantities to each terminal, synchronizing local models with a master model to ensure consistent data usage and reduce variation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If random data sampling is used in federated learning, then data privacy is maintained, but inference accuracy variation occurs
Solution Approach 1:
The patent segments the data collection and management process into multiple functional modules: search condition management, data possession totalization, distribution information generation, and learning result integration. Each module operates independently to manage specific aspects of data handling, reducing overall system complexity while maintaining systematic control over data sampling across federated clients.
Solution Approach 2:
The patent implements feedback mechanisms where client terminals report data possession information back to the integration server, which then uses this feedback to generate distribution information for the next learning round. This closed-loop feedback system ensures consistent and controlled data sampling while maintaining privacy, thereby stabilizing inference accuracy across different clients.
2Reliability
If systematic data collection method is implemented, then inference accuracy consistency is improved, but system complexity increases
Solution Approach 1:
The integration server performs multiple functions through a unified system: it manages search conditions, receives data possession reports, generates distribution information, synchronizes models, and integrates learning results. This multi-functional approach consolidates what would otherwise be separate complex systems into a single coordinated platform, improving accuracy consistency without proportionally increasing complexity.
Solution Approach 2:
The patent systematically changes parameters such as search conditions, data selection criteria, and distribution ratios to optimize learning data composition. By adjusting these parameters in a controlled manner across different learning rounds, the system achieves consistent inference accuracy while maintaining manageable complexity through parameterized control rather than structural complexity.
3Reliability
If data is centrally extracted for training, then learning consistency is improved, but privacy risk increases
Solution Approach 1:
The patent extracts only the necessary control information (search conditions, distribution ratios, model parameters) from the central server to client terminals, while keeping the actual sensitive medical data localized at each institution. This selective extraction approach enables consistent learning control without extracting privacy-sensitive data, thereby maintaining learning consistency while eliminating privacy risks associated with central data collection.
Solution Approach 2:
The integration server acts as an intermediary that coordinates learning without directly accessing or storing sensitive patient data. It mediates between centralized control requirements and decentralized data storage by exchanging only aggregated statistics, search conditions, and model parameters, thus enabling learning consistency while preserving data privacy through the intermediary's buffering role.
Data Source
AI summary
Provided are a machine learning system and method, an integration server, an information processing apparatus, a program, and an inference model creation method capable of suppressing a variation in learning data in federated learning and suppressing a variation in an inference accuracy of a model. The integration server receives an input that designates a data search condition and transmits the designated search condition to a plurality of client terminals. Each client terminal performs searching within a medical institution system to which each terminal belongs and transmits a totalization result of the number of pieces of data that matches the search condition to the integration server. The integration server receives an input that designates the required number of pieces of learning data and distributes the number of pieces of learning data used for learning on each client terminal based on the designated required number and on the received totalization result. The client terminal executes machine learning of a local model to be trained using the data in the medical institution system according to the designated type and number of pieces of learning data and transmits the learning result to the integration server. The integration server integrates the received learning results to update a master model.