Enhancing cancer detection in medical imaging through federated learning and explainable artificial intelligence: A hybrid approach for optimized diagnostics

Abstract

The diagnosis of cancer are crucial medical responsibilities that assist medical practitioners correctly classify and treat them accordingly. Machine learning applications are widely used in medical field as they identify patterns from clinical data. Traditional machine learning approaches often struggle with accurately identifying malignancies due to the complexity and variability of medical data. This study aims to enhance the accuracy and interpretability of cancer detection models by integrating LightGBM with SHAP (SHapley Additive exPlanations) within a federated learning framework. The innovation of this research lies in the combination of LightGBM’s ability in handling high dimensional feature of large data size with SHAP’s detailed interpretability metrics. This integration not only facilitates accurate cancer detection but also provides insights into the contributing factors of the model’s predictions, making it easier for healthcare professionals to trust and utilize these models. The federated learning approach allows multiple institutions to collaborate in training the model without sharing raw patient data, ensuring data privacy while benefiting from diverse datasets. The integrated framework achieved a remarkable accuracy of 98.3% in cancer detection, with precision, recall, and F1 scores of 97.8%, 97.2%, and 95%, respectively. These results indicate that the proposed method effectively identifies cancer cases while maintaining high interpretability, allowing for better decision-making in clinical settings. The integration of LightGBM with SHAP within a federated learning framework provides a powerful and effective solution for cancer detection.