Related Experiment Video
Updated: Jan 17, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Federated learning and differential privacy: Machine learning and deep learning for biomedical image data
Sobia Wassan1, Liudajun1, Han Ying1
1School of Equipment Engineering, Jiangsu Urban and Rural Construction Vocational College, Changzhou, China.
Digital Health
|September 15, 2025
Summary
Federated learning with differential privacy effectively classifies biomedical images, enhancing patient confidentiality without sacrificing model accuracy. This approach supports secure machine learning in medical diagnostics.
Area of Science:
- Biomedical image analysis
- Machine learning
- Data privacy
Background:
- Healthcare data privacy is crucial for patient confidentiality and accurate predictive modeling.
- Increasing privacy concerns necessitate methods that protect data without compromising model performance.
Purpose of the Study:
- Evaluate feedforward neural networks (FNNs), Gaussian processes (GPs), and multilayer perceptrons (MLPs) for biomedical image classification.
- Incorporate federated learning to enhance privacy preservation in these models.
Main Methods:
- Implemented FNN, GP, and MLP models using federated learning and differential privacy.
- Evaluated models based on accuracy, correlation coefficients, MAE, RMSE, RAE, and RRSE.
Main Results:
- The deep neural network (DNN) model demonstrated superior performance with a correlation coefficient of 0.9980, MAE of 36.80, and RMSE of 51.01.
- Federated learning successfully improved privacy while maintaining robust model performance across evaluated metrics.
Conclusions:
- Federated learning combined with differential privacy presents a viable solution for secure biomedical image classification.
- This privacy-preserving machine learning approach supports medical diagnostics without performance degradation.
