Implemented global model for brain tumor detection using Federated Learning

Brain tumor (BT) is a severe medical issue that is challenging to detect and treat. For a few decades, there has been interest in applying machine learning and deep learning models to automate BT diagnosis. However, patients’ medical data is generally sensitive, and employing these models for BT diagnosis creates privacy problems. To overcome the issue mentioned above, federated learning (FL) architecture considers automating BT diagnosis. The proposed architecture develops and evaluates a global model using FL for BT identification. Using Backpropagation and DenseNet architecture, a model is designed for the MRI images and efficiently detects tumors for the patient. This approach uses a convolutional neural network to train a model on each device locally. The local models are combined to develop a global model to identify BT that allows clinics to transmit sensitive information while remaining anonymous. The results demonstrated that the framework accurately classified BT with good clinical accuracy and reliability.