Ovarian cancer, one of the most dangerous malignancies of the female reproductive system, demands accurate and early diagnosis. This study proposes a federated learning system for collaborative image-based ovarian cancer diagnosis, enabling multiple healthcare institutions to train a robust diagnostic model without compromising patient data privacy. By utilizing advanced techniques like differential privacy and secure aggregation, the system addresses challenges such as data heterogeneity and communication efficiency. Experimental results demonstrate that the proposed model achieves a test accuracy of 95.71% on a dataset of over 349 raw data points, performing comparably to centralized models while maintaining stringent privacy standards. This work highlights the potential of federated learning to revolutionize medical research and improve diagnostic outcomes for ovarian cancer.