FedOnco-Bench: A Reproducible Benchmark for Privacy-Aware Federated Tumor Segmentation with Synthetic CT Data
Viswa Chaitanya Marella, Suhasnadh Reddy Veluru, Sai Teja Erukude
TL;DR
This paper introduces FedOnco-Bench, a reproducible benchmark for privacy-aware federated learning in medical image segmentation using synthetic CT data. It systematically compares FedAvg, FedProx, FedBN, and FedAvg with DP-SGD, highlighting a clear trade-off between privacy leakage (measured by MIA AUC) and segmentation utility (Dice and CE). DP-SGD dramatically reduces privacy risk at the cost of some accuracy, while FedProx and FedBN offer balanced performance under data heterogeneity. The benchmark, implemented in PyTorch and made open source, provides a standardized platform to evaluate privacy-preserving FL methods and to study privacy–utility dynamics in clinical imaging scenarios.
Abstract
Federated Learning (FL) allows multiple institutions to cooperatively train machine learning models while retaining sensitive data at the source, which has great utility in privacy-sensitive environments. However, FL systems remain vulnerable to membership-inference attacks and data heterogeneity. This paper presents FedOnco-Bench, a reproducible benchmark for privacy-aware FL using synthetic oncologic CT scans with tumor annotations. It evaluates segmentation performance and privacy leakage across FL methods: FedAvg, FedProx, FedBN, and FedAvg with DP-SGD. Results show a distinct trade-off between privacy and utility: FedAvg is high performance (Dice around 0.85) with more privacy leakage (attack AUC about 0.72), while DP-SGD provides a higher level of privacy (AUC around 0.25) at the cost of accuracy (Dice about 0.79). FedProx and FedBN offer balanced performance under heterogeneous data, especially with non-identical distributed client data. FedOnco-Bench serves as a standardized, open-source platform for benchmarking and developing privacy-preserving FL methods for medical image segmentation.