A Cloud-Edge Collaboration Framework for Cancer Survival Prediction to Develop Medical Consumer Electronic Devices

Consumer electronics, together with artificial intelligence technology, are being widely applied in the medical field, aiding medical data collecting, health monitoring, disease diagnosis, and survival prediction. Training an effective and robust medical artificial intelligence model for cancer survival prediction deployed in consumer electronic devices requires a large amount of high-quality annotated data. However, due to limitations in computing resources and energy, it is impractical to train artificial intelligence models on large-scale medical data directly in the cloud. To address the above problem, we propose a cloud-edge collaboration framework to develop medical electronic devices for cancer survival prediction. The framework trains local models on edge medical electronic devices and aggregates the model weights on the cloud server. Furthermore, when transmitting model weights between cloud and edge, we introduce differential privacy technology to ensure security concerns. We evaluate our proposed cloud-edge collaboration framework on two cancer datasets from The Cancer Genome Atlas (TCGA). Experimental results demonstrate that our cloud-edge collaboration framework is highly effective in both survival prediction performance and patient privacy preservation, with the potential to develop medical consumer electronics.