Archimedes Optimization-Driven CNN in Federated Settings: A Case Study on Thyroid Disease Diagnosis

Artificial Intelligence (AI) is playing an increasingly vital role in the healthcare industry, driving the transformation toward smart systems that enhance diagnostic accuracy, enable personalized treatment, and support efficient clinical decision-making. In this context, predicting thyroid disease is critical for long-term patient management and improving survival outcomes. However, traditional machine learning approaches face limitations due to data privacy concerns and the distributed nature of healthcare data. In this study, we propose a federated learning (FL) framework for the classification of thyroid disease using convolutional neural networks (CNNs). To further improve model performance, we employ the Archimedes Optimization Algorithm (AOA) to optimize CNN parameters, ensuring efficient convergence and enhanced predictive accuracy. FL approach enables collaborative training across multiple healthcare institutions without exposing sensitive patient data, thus preserving privacy through decentralized learning. We evaluate and compare several federated learning algorithms and client trainings based on classification accuracy and f1-score. Experimental results demonstrate that the AOA-optimized CNN within the FL framework outperforms client local trainings. The accuracy rate of 90.68% obtained with the FedProx algorithm shows that it achieves performance comparable to that of the centralized learning approach.