Privacy Focused Classification of Prostate Cancer Using Federated Learning

Prostate cancer detection is an extremely challenging problem due to the socially sensitive nature of the data. As such, it negatively impacts early detection, including new procedures becoming difficult to develop owing to a lack of enough datasets. The proposed approach, which is based on Federated Learning, is a decentralized strategy that uses weight averaging and can quickly find anomalies while retaining data confidentiality. Additionally, the results of the customized basic CNN model with three Conv2D layers and two state-of-the-art models, VGG19 and Xception, were compared to those obtained in centralized and decentralized configurations. To recapitulate, this research’s findings indicate that the decentralized technique yields results that are very consistent with those yielded by state-of-the-art centralized methods.