Secretary bird earth radius optimization algorithm–Shepard pyramid forward harmonic net for prostate cancer detection in federated learning

Prostate cancer is the most aggressive and malignant cancer that impacts the prostate gland in men. The detection and staging of earlier prostate cancer employing magnetic resonance imaging (MRI) are difficult tasks that pose important challenges for radiologists. On the other hand, federated learning (FL) is useful to protect the data, but it also means that every device’s data might be different from others. This paper presents a FL-enabled secretary bird earth radius optimization algorithm–Shepard pyramid forward harmonic net (FL-enabled SBEROA-ShPyFHNet) for prostate cancer detection using MRI images, enhancing detection accuracy while preserving patient data privacy. At first, local training is performed. Thereafter, model aggregation is performed at the server. Then, the training model is updated with the downloaded global and local models. In the training model, preprocessing is executed by an adaptive nonlinear filter. Thereafter, area segmentation is conducted by Dense-Res-Inception Net (DRINET). Then, feature extraction is conducted to extract features. At last, the prostate cancer detection and severity-level classification is accomplished employing ShPyFHNet. Here, ShPyFHNet is trained employing the secretary bird earth radius optimization algorithm (SBEROA), which is the combination of SBOA and Al-Biruni earth radius (BER). Lastly, local updation and aggregation at the server are performed based on the average technique. The proposed FL-enabled SBEROA-ShPyFHNet achieved 96.75% accuracy, 0.033 loss, 0.043 mean squared error (MSE), 0.207 root mean squared error (RMSE) and 96.40% mean average precision (MAP). These results indicate that the model can reliably assist radiologists in detecting prostate cancer, potentially reducing unnecessary biopsies and enabling more precise patient management in clinical workflows.