Partea

Partea performs privacy-preserving federated time-to-event (survival) analyses using survival curves, cumulative hazard rates, log-rank tests, and Cox proportional hazards models to enable multi-institutional clinical research without centralizing sensitive medical data.

Key Features:

• Privacy-Preserving Data Analysis: Implements a hybrid privacy approach combining federated learning, additive secret sharing, and differential privacy to protect sensitive medical data.
• Federated Implementations of Time-to-Event Algorithms: Provides federated implementations of survival curves, cumulative hazard rates, log-rank tests, and Cox proportional hazards models that produce results comparable to centralized approaches.
• Benchmarking and Validation: Validates performance on several benchmark datasets and reproduces findings from prior clinical time-to-event studies in federated scenarios.

Scientific Applications:

• Multi-institutional clinical survival analysis: Enables pooled time-to-event analyses across institutions while preserving patient confidentiality.
• Federated clinical trial and observational study analysis: Supports reproduction and benchmarking of clinical time-to-event study results without centralizing patient-level datasets.

Methodology:

Combines federated learning, additive secret sharing, and differential privacy; implements federated survival curves, cumulative hazard rates, log-rank tests, and Cox proportional hazards models; and validates methods on benchmark datasets and prior clinical time-to-event studies.