Recent Trends of Federated Learning for Smart Healthcare Systems

1. Introduction

In the current era, technology has taken a sharper turn to change the world like it was never before. It is not only concerned with smartphones, laptops, or computers anymore. A multitude of devices are internet-connected, thus, creating more efficient “smart devices”. The Internet has grown explosively in all fields, probably because of its most popular and easy-to-use (“the Web”) which delivers information conveniently to desktops (Mansoor, 2002). People can easily access information throughout the world with just a computer and internet access. From online banking to education to social networking internet has entered its golden age (Mohamed, 2019). Even the household appliances can also be connected to network and can allow them to work independently (Mohamed, 2019). Siri, Alexa, and some other voice command devices have built intelligent conversational user interface using AI (Mohamed, 2019). The technologies have also boomed in real-world and healthcare sector (Mansoor, 2002). IoT in healthcare is often referred to as Internet of Medical Things (IoMT) (Yaqoob et al., 2017). With each passing day, the dependency of healthcare on IoMT and AI began to flourish. The main concept was to enhance the strength and quality of care along with reduction in the cost of care. By preserving each patient’s digital identity, IoT makes sure the personalization of healthcare services. Numerous health issues go undiagnosed in traditional healthcare systems because they are not readily available. However, sophisticated, ubiquitous, IoT-based devices have made it possible to easily monitor and analyse data of patient (Aminoshariae et al., 2021). This not only lessens the monopoly of private hospitals but also makes sure that government setups work fairly with patient community (Khanagar et al., 2023). For in-hospital patients, medical sensors are incorporated into different wearable devices. These devices regulate the pulse and oxygen saturation rate on regular basis. Even robots have been trained to do surgeries and other tasks like remote monitoring, patient care etc using AI technology (Yaqoob et al., 2017). In this way, advanced computational techniques were required for proper analysis, processing, and storage. Strict guidelines like Health Insurance Portability and Accountability Act (HIPAA) were framed to administer the process of interpretation and usage of data (Deshmukh, 2018). However, the large voluminous data led to many challenges to advanced techniques such as Machine learning (ML), Deep learning (DL) etc as they were not trained to access huge amount of data (Rani et al., 2023). To avoid the transmission of data, Federation learning (FL) came into highlights. An approach where issues related to security of sensitive data can be reduced significantly. The combination of FL, AI, and Explainable Artificial Intelligence (XAI) techniques can minimize several limitations and challenges in the healthcare system.

The healthcare industry is one of the most vulnerable to cybercrime and privacy violations because health data is extremely sensitive and is transferred to support studies, early diagnosis, and monitoring of diseases. Remote hospitals generally have imbalanced datasets with intermittent clients communicating with the FL global server (Elayan et al., 2022). In this context, FL benefits both the specialist as well as the patient. It reassures the patient that their sensitive data will not be shared with outside facilities, a breach in terms of security. Also, it works efficiently in predicting hospitalization for events related to the cardiac center by processing early signals for pathologies or knowledge tasks of patient data (Rani et al., 2023). FL-based models were also successful in the analysis and detection of COVID-19. They were able to analyse the imaging models like X-rays, and CT scans to detect various abnormalities (Loey et al., 2020; Zhang, Zhou, Lu et al, 2021).