Related Experiment Video
Updated: Mar 22, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.7K
FedGA: Genetic Algorithm-Guided Federated Learning for Medical Image Segmentation with Non-IID Features
IEEE Journal of Biomedical and Health Informatics
|March 20, 2026
Summary
Federated learning (FL) improves medical image segmentation by using a novel genetic algorithm (FedGA) to handle diverse data domains. FedGA enhances precision and speeds up convergence in federated learning schemes.
Area of Science:
- Artificial Intelligence
- Medical Imaging
- Machine Learning
Background:
- Federated learning (FL) enables collaborative model training across decentralized data sources while preserving privacy, making it suitable for healthcare.
- Standard FL struggles with non-independent and identically distributed (non-IID) data, particularly in medical image segmentation where data domains vary significantly.
- Existing methods often focus on label distribution skew, leaving the challenge of multi-domain feature distribution in medical imaging less explored.
Purpose of the Study:
- To address the challenge of multi-domain federated learning in medical image segmentation.
- To propose a novel approach, FedGA, that optimizes global models using gradient-free genetic algorithms on the server side.
- To evaluate FedGA's effectiveness in improving segmentation precision and convergence efficiency.
Main Methods:
- Developed FedGA, a federated learning framework incorporating a genetic algorithm for server-side, gradient-free optimization post-aggregation.
- Applied FedGA to multi-domain medical image segmentation tasks, specifically breast lesion segmentation in ultrasound and prostate segmentation in MRI.
- Compared FedGA against existing approaches to assess improvements in segmentation accuracy and convergence metrics.
Main Results:
- FedGA demonstrated significant improvements in segmentation precision, particularly in critical boundary regions.
- The proposed method accelerated global model convergence and reduced the number of communication rounds needed for optimal performance.
- Empirical results confirmed FedGA's potential in enhancing federated learning efficiency for medical image segmentation.
Conclusions:
- FedGA effectively tackles the challenge of multi-domain data in federated medical image segmentation.
- The genetic algorithm-based optimization enhances segmentation accuracy and communication efficiency in FL.
- FedGA offers a promising solution for privacy-preserving, high-performance medical image analysis in diverse healthcare settings.
