New Sano publication on federated learning in medical imaging
How to train powerful AI models on hospital data without moving sensitive images
We are pleased to share that a new paper titled “Federated learning: A new frontier in the exploration of multi‑institutional medical imaging data”, co‑authored by Dominika Ciupek, Maciej Malawski and Tomasz Pieciak, is now available via ScienceDirect.
Modern deep learning methods in medical imaging need very large and diverse datasets to learn robust, clinically useful patterns. At the same time, sharing raw medical images between institutions is difficult due to strict privacy regulations, time‑consuming ethical procedures and the need for careful anonymization and data management. The paper focuses on federated learning (FL) – an emerging paradigm that allows hospitals to collaborate on AI models without sharing patient data directly.
The authors provide the theoretical foundations and a structured overview of federated learning, with a particular focus on medical imaging applications. They review general and specialized aggregation and learning algorithms that make it possible to train a globally generalized model across many institutions while keeping data on‑site. The article also highlights key challenges such as data and model heterogeneity, privacy and security risks, and limitations in computation and communication between participating sites, as well as an overview of regulatory frameworks relevant for deploying FL in healthcare.
Fig. 1. Comparison between centralized and federated learning approaches: A. In a centralized architecture, the institutions (here, 1, 2, 3) transfer their local datasets to the central server. Other centers (Institution 4) extract datasets from the global server or use its computing infrastructure to train the DL models. B. Each institution’s data remains locally preserved in a federated architecture while the parameters of locally trained models are transferred to the central server. The central server aggregates received parameters and sends back the parameters of a global model to each center.
The full text of “Federated learning: A new frontier in the exploration of multi‑institutional medical imaging data” is available here:
ScienceDirect – www.sciencedirect.com
