Medical Information Processing Apparatus for Decentralized Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical information processing systems face challenges in decentralized machine learning due to the need for synchronized execution environments and parameters across multiple institutions, which is cumbersome and not suitable for medical data sharing, especially concerning data security and asset ownership.
Innovation Solution
A medical information processing apparatus and system that distributes a learning-purpose program to multiple medical institutions, adjusts and transmits parameters for machine learning processes, and displays and corrects data imbalances to ensure consistent model training without external data transfer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If medical data is aggregated from multiple medical institutions to create a trained model, then the model training quality is improved, but data security and asset ownership are compromised
Solution Approach 1:
The system segments the centralized data aggregation process into distributed local learning processes at each medical institution. Each institution keeps its data locally while contributing to the global model through parameter exchanges, thus maintaining data security while achieving model training quality improvement.
Solution Approach 2:
The system introduces parameter updates and gradients as intermediaries between medical institutions. Instead of directly sharing sensitive medical data, institutions exchange only the mathematical parameters needed for model training, acting as a secure mediator that enables collaboration without compromising data ownership.
2Reliability
If decentralized machine learning is performed by transmitting parameters among workers, then data security is maintained, but execution environment configuration becomes cumbersome
Solution Approach 1:
The system creates a universal execution environment where the same learning program can run across different medical institutions with varying hardware configurations. By standardizing the learning framework and allowing flexible parameter adjustments, the system achieves multi-functionality that works across diverse execution environments without requiring identical configurations.
Solution Approach 2:
The system dynamically adjusts learning parameters such as the number of iterations, batch sizes, and hyperparameters based on each institution's execution environment capabilities. This parameter flexibility allows the decentralized learning process to adapt to different hardware resources while maintaining data security.
3Stability of the object's composition
If execution environments are synchronized across multiple workers, then learning consistency is improved, but system complexity and preparation effort increase
Solution Approach 1:
The system performs preliminary actions by distributing a standardized learning program template to all participating institutions before the actual training begins. This pre-configured framework establishes consistent execution environments without requiring complex synchronization during the learning process, reducing preparation effort while maintaining learning consistency.
Data Source
AI summary
A medical information processing apparatus according to an embodiment includes a processing circuitry. The processing circuitry is configured: to distribute, to an information processing apparatus provided at each of a plurality of medical institutions, a program for causing a machine learning process to be executed by using medical data held at the medical institution having the information processing apparatus; to receive, from each of the information processing apparatuses, a change amount in a parameter related to the machine learning process, regarding a change caused in conjunction with the execution of the machine learning process; to adjust a value of the parameter on the basis of the received change amount; and to transmit the adjusted parameter to each of the information processing apparatuses to cause the machine learning process to be executed on the basis of the parameter.