Federated Learning With Deidentified Patient Data Aggregation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The training of machine learning models using patient data is complicated by data privacy laws, necessitating a method to utilize deidentified data for centralized training without sharing sensitive information across locations.
Innovation Solution
A federated learning approach is employed where machine learning models are trained locally using deidentified data, with a deidentification process that replaces personal information with surrogates, and the training data is then aggregated for a centralized model update, maintaining data security and privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If patient data is used for centralized model training, then model accuracy is improved, but data privacy compliance deteriorates
Solution Approach 1:
The system segments the centralized training process into distributed local training operations at multiple locations. Each location trains a local copy of the machine learning model using its own patient data, preventing centralization of sensitive data while still achieving collaborative model improvement through aggregation of training results.
Solution Approach 2:
The system introduces an intermediary deidentification process that transforms patient data into deidentified data before training. This intermediary step allows the model to learn from patient data patterns while removing personally identifiable information, thus maintaining privacy compliance while enabling model training.
2Object-affected harmful factors
If deidentified data is used for training, then data privacy compliance is improved, but model training quality deteriorates
Solution Approach 1:
The system performs preliminary local training at each location using deidentified data before centralized aggregation. This preliminary action ensures that training quality is maintained through local optimization while privacy is protected, and the locally trained models are then aggregated to achieve final model improvement without compromising either quality or privacy.
3Measurement precision
If data is shared across locations for training, then model accuracy is improved, but data security deteriorates
Solution Approach 1:
The system segments the training data and model copies across multiple locations, preventing centralization and sharing of raw patient data. Each location maintains its own data locally and trains a local model copy, achieving collaborative training without data sharing or centralization, thus maintaining both accuracy improvement and data security.
Data Source
AI summary
A method, computer program product, and computing system for processing a raw content portion to identify personal information, thus defining identified personal information; confirming the efficacy of the identified personal information; replacing the identified personal information within the raw content portion with one or more surrogates, thus defining a deidentified content portion; and performing a federated learning operation on the deidentified content portion to generate training data for a machine learning model.