Comparative Analysis of Federated Learning and Centralized Approach for Detecting Different Lung Diseases
Kh. Fardin Zubair Nafis, Syeda Maisha Tarannum, Krity Haque Charu, Md Humaion Kabir Mehedi, Annajiat Alim Rasel · 2023
Access to a large dataset is necessary to improve disease detection with excellent accuracy. However, due to data confidentiality and privacy restrictions, collecting data from hospitals or other organizations is a significant challenge in the healthcare sector. Due to this, Federated Learning (FL), which adopts a decentralized approach, is developed to replace the conventional machine learning methodology in the development of improved screening methods. Since there is no requirement for data to be centralized in federated learning, patient data privacy is ensured. In this paper, we have compared the sequential model and the ensemble model for both federated learning and centralized approach, two different types of models. For each approach, these models were applied to separate X-ray images for the detection of two different lung diseases: lung cancer and tuberculosis. In this paper, we also have showed the analysis of their accuracy and demonstrated how FL can be the most effective strategy through comparison.