Leveraging towards Privacy-preserving using Federated Machine Learning for Healthcare Systems

Abstract

With the advent of artificial intelligence, healthcare industry has been successful in improving medical diagnosis, medical prognosis, and other clinical tasks using machine learning. The success of improving healthcare services is largely due to the amount of patients data acquired usually in privacy-invasive ways for the process of machine learning. The issue of privacy-preserving and utilizing the large and diverse private datasets is solved using federated learning. It can perform predictive tasks alongside maintaining privacy of medical data. In this paper, we have proposed cancer text classification system using federated learning to maintain the integrity of sensitive medical data. The acquisition of data from various healthcare institutions is used for detection of cancer in preliminary stage in the patient's body. The experiment has been conducted for various types of healthcare environmental settings for the process of federated learning including change in the number of health institutions involved, change in data distribution among numerous healthcare institutions involved, and performance in case of applying various machine learning models. Here, we utilize several machine learning models namely, Recurrent Neural Network (RNN), Bi-RNN, Gated Recurrent Unit (GRU), and Long short-term Memory (LSTM) for text classification that attains accuracy of 57%, 59%, 60%, and 61% respectively. The paper proposes a novel federated learning based system capable of text classification in the healthcare industry with the use of federated machine learning alongside maintaining data privacy.