Publications by authors named "Marwin Saehn"
Article Synopsis
- Medical image analysis is increasingly reliant on deep learning, but creating effective models requires large datasets that are often inconsistently labeled by different institutions.
- Traditional federated learning isn’t suitable for training on these heterogeneous datasets, which is why flexible federated learning (FFL) is proposed as a solution.
- By using a large dataset of chest radiographs from multiple sources, FFL shows a significant improvement in training performance, suggesting it could enhance collaborative training in real-world healthcare applications.
