Federated Learning-Based Variational Auto-Encoder for Prediction of Breast Cancer in Cloud-Based Healthcare 5.0
Santosh Kumar, Tarunika Sharma, Balasubramanian Prabhu Kavin, Gan Hong Seng · 2024
Even in the modern era of cutting-edge intelligence and the intelligent healthcare sector 5.0, diagnosing individual illnesses remains a challenging task. For optimal human health in the smart healthcare sector 5.0, accurate illness forecasting is essential, especially for disorders related to breast cancer. Small wristwatches to huge airplanes have all evolved rapidly in recent times promoted worldwide by the Internet of Health Things. Due to the vast magnitude and widespread disposition of Internet of Health Things networks, safety and secrecy are among the most important features of the IoMT. These days, edge computing techniques like combined learning remain more significant for this use. These techniques can make use of the data without keeping it in a central location. This study assesses the efficacy of federated learning for breast cancer prediction. To cut the distinct texture characteristics of breast tissues in the suggested system, a GLCM (gray-level co-occurrence matrix) is combined with AlexNet architecture. Finally, a federated learning (FL) based variational auto-encoder approach is applied to achieve improved accuracy. The WI Indicative Chest Cancer (WBCD) dataset results show that the cloud-based FL-VAE technology performs better than competing methods. When comparing FL-VAE&s;s performance in standalone and cloud contexts, the latter produced the best results.