Federated Learning in Healthcare Applications (FLHA)

FLHA is a one-day workshop co-located with FLICS 2026 in Valencia, Spain (9–12 June 2026). The exact workshop day will be announced with the FLICS workshop schedule.

Overview

Healthcare data is dispersed across hospitals, labs, registries, devices, and research networks, often stored in incompatible formats and governed by strict legal, ethical, and institutional constraints. These silos limit the scale and diversity of data available to any single organization, which in turn can hinder the development of robust, generalizable machine learning models. Centralizing sensitive clinical data is frequently impractical due to privacy risk, compliance requirements, security posture, and trust considerations.

Federated learning (FL) offers an alternative collaboration model: institutions keep data local while jointly training models by exchanging model updates rather than raw records. A coordinating service aggregates updates over multiple training rounds to form a global model that can benefit from multi-site data without requiring data pooling. In practice, FL is often combined with privacy-enhancing and security mechanisms such as secure aggregation, differential privacy, and encrypted computation, alongside governance controls that define participation, permissible uses, auditing, and accountability.

Deploying FL in healthcare raises challenges beyond distributed optimization. Data are typically heterogeneous across sites due to differences in populations, care pathways, coding practices, instruments, and acquisition protocols. This non-IID setting can lead to unstable training and uneven performance across institutions and subgroups. Effective solutions often require personalization strategies, robust aggregation, domain adaptation, careful evaluation and calibration, and ongoing monitoring for drift.

Real-world FL systems must also operate within healthcare IT constraints. Institutions vary widely in compute resources and MLOps maturity; networking and scheduling must accommodate operational limitations; and security reviews, procurement processes, and regulatory oversight can shape what is feasible. Achieving reproducible, auditable training and reliable deployment is therefore as much a systems and governance problem as an algorithmic one.

FLHA brings together researchers and practitioners to advance federated learning methods, platforms, and deployments in healthcare. The workshop emphasizes end-to-end realities, spanning algorithms and privacy, engineering and infrastructure, evaluation and benchmarking, and governance and operational lessons from deployed systems.

Submission Tracks

We invite:

Important Dates

All deadlines are Anywhere on Earth (AoE) unless stated otherwise.