JMIR Publications

ABSTRACT

Background:

Central collection of distributed patient data is problematic due to strict privacy regulations. Especially in clinical environments, such as clinical time-to-event studies, large sample sizes are critical but usually not available at a single institution. It has been shown recently that federated learning, combined with privacy-enhancing technologies, is an excellent and privacy-preserving alternative to data sharing.

Objective:

Many algorithms for federated time-to-event analysis, such as survival support vector machines, are still unavailable and not accessible to researchers and statisticians. Making these algorithms available for federated environments is critical for obtaining large sample sizes in cross-institutional time-to-event studies to comply with current privacy regulations.

Methods:

We extended the survival support vector machine algorithm to be applicable to federated environments. We further implemented it as a FeatureCloud app, enabling it to run in the federated infrastructure provided by the FeatureCloud platform. We evaluated our implemented algorithm on three benchmark datasets and a real-world microbiome dataset and compared it to the corresponding central method.

Results:

Our federated survival SVM produces highly similar results to the centralized model on all datasets. The maximal difference between the model weights of the central model and the federated model was only 0.001, and the mean difference over all datasets was 0.0002. We further show that by including more data in the analysis through federated learning, predictions are more accurate even in the presence of site-dependent batch effects.

Conclusions:

The federated survival support vector machine extends the palette of federated time-to-event analysis methods by a robust machine learning approach. To our knowledge, the implemented FeatureCloud app is the first publicly available implementation of a federated survival support vector machine, is freely accessible for all kinds of researchers, and can be directly used within the FeatureCloud platform.