A Deep Learning-Based IoT Implementation for Detection of Patients’ Falls in Hospitals

Abstract

Falls in hospitalized patients are a major problem for patient safety. Accidental falls are one of the most common incidents reported in hospitals. Thanks to the advances in technology, smart solutions can be developed for hospital environments as well as in all areas of life. Wearable devices, context-aware or computer vision-based systems can be designed to detect patients who fall in hospital. Internet of Things (IoT) can also be placed on wearable health products, and gathered sensors data is processed and analyzed with Machine Learning (ML) and Deep Learning (DL) algorithms. Furthermore, some DL algorithms such as LSTM are also applied to the analysis of time-series data. In this study, to minimize damage caused by falls, we’ve proposed a model that can achieve real-time fall detection by applying LSTM based deep learning technique on IoT sensor data. In result of the study, falling detection has been realized with 98% F1-score. Moreover, a mobile application has been successfully developed to inform caregivers about patients’ fall.