Federated Learning Approach for Early Detection of Chest Lesion Caused by COVID-19 Infection Using Particle Swarm Optimization

Abstract

The chest lesion caused by COVID-19 infection pandemic is threatening the lives and well-being of people all over the world. Artificial intelligence (AI)-based strategies are efficient methods for helping radiologists by assessing the vast number of chest X-ray images, which may play a significant role in simplifying and improving the diagnosis of chest lesion caused by COVID-19 infection. Machine learning (ML) and deep learning (DL) are such AI strategies that have helped researchers predict chest lesion caused by COVID-19 infection cases. But ML and DL strategies face challenges like transmission delays, a lack of computing power, communication delays, and privacy concerns. Federated Learning (FL) is a new development in ML that makes it easier to collect, process, and analyze large amounts of multidimensional data. This could help solve the challenges that have been identified in ML and DL. However, FL algorithms send and receive large amounts of weights from client-side trained models, resulting in significant communication overhead. To address this problem, we offer a unified framework combining FL and a particle swarm optimization algorithm (PSO) to speed up the government’s response time to chest lesion caused by COVID-19 infection outbreaks. The Federated Particle Swarm Optimization approach is tested on a multidimensional chest lesion caused by the COVID-19 infection image dataset and the chest X-ray (pneumonia) dataset from Kaggle’s repository. Our research shows that the proposed model works better when there is an uneven amount of data, has lower communication costs, and is therefore more efficient from a network’s point of view. The results of the proposed approach were validated; 96.15% prediction accuracy was achieved for chest lesions caused by the COVID-19 infection dataset, and 96.55% prediction accuracy was achieved for the chest X-ray (pneumonia) dataset. These results can be used to develop a progressive approach for the early detection of chest lesion caused by COVID-19 infection.