Blockchain-Federated-Learning and Deep Learning Models for COVID-19\n detection using CT Imaging

Abstract

With the increase of COVID-19 cases worldwide, an effective way is required\nto diagnose COVID-19 patients. The primary problem in diagnosing COVID-19\npatients is the shortage and reliability of testing kits, due to the quick\nspread of the virus, medical practitioners are facing difficulty identifying\nthe positive cases. The second real-world problem is to share the data among\nthe hospitals globally while keeping in view the privacy concerns of the\norganizations. Building a collaborative model and preserving privacy are major\nconcerns for training a global deep learning model. This paper proposes a\nframework that collects a small amount of data from different sources (various\nhospitals) and trains a global deep learning model using blockchain based\nfederated learning. Blockchain technology authenticates the data and federated\nlearning trains the model globally while preserving the privacy of the\norganization. First, we propose a data normalization technique that deals with\nthe heterogeneity of data as the data is gathered from different hospitals\nhaving different kinds of CT scanners. Secondly, we use Capsule Network-based\nsegmentation and classification to detect COVID-19 patients. Thirdly, we design\na method that can collaboratively train a global model using blockchain\ntechnology with federated learning while preserving privacy. Additionally, we\ncollected real-life COVID-19 patients data, which is, open to the research\ncommunity. The proposed framework can utilize up-to-date data which improves\nthe recognition of computed tomography (CT) images. Finally, our results\ndemonstrate a better performance to detect COVID-19 patients.\n