This page summarises the motivation for using federated learning (FL) in neglected tropical disease (NTD) image classification.

Microscopy data in endemic settings is often siloed across facilities and cannot be freely centralised due to privacy and governance constraints. Federated learning enables a shared model to be trained while keeping raw data local.

The simulation mimics these constraints by using filename-mapped labels and synthetic model behaviour, focusing on interpretability of the pipeline rather than on real clinical performance.