Ensemble Federated Learning Approach for Diagnostics of Multi-Order Lung Cancer.

Area of Science:

• Oncology
• Computer Science
• Artificial Intelligence

Background:

• Early lung cancer detection and classification are vital for patient outcomes.
• Traditional methods using single machine learning models are limited by centralized data availability and quality.
• Privacy and security concerns hinder the use of distributed datasets in traditional approaches.

Purpose of the Study:

• To propose an ensemble Federated Learning (FL) approach for multi-order lung cancer classification.
• To overcome the limitations of centralized data in traditional machine learning models.
• To enhance classification accuracy and generalization while ensuring data privacy and security.

Main Methods:

• Developed an ensemble Federated Learning framework combining multiple machine learning models.
• Trained models on distributed datasets without centralizing sensitive patient information.
• Evaluated the proposed approach on a Kaggle cancer dataset.

Main Results:

• Achieved an accuracy of 89.63% for lung cancer classification.
• Demonstrated improved accuracy and generalization compared to traditional single machine learning models.
• Validated the effectiveness of Federated Learning in handling distributed, private data.

Conclusions:

• The ensemble Federated Learning approach offers a robust solution for lung cancer classification.
• FL enables the utilization of distributed data, enhancing privacy and security in medical AI.
• This method shows significant potential for improving early lung cancer detection and patient outcomes.