Internet of health things driven deep learning-based system for non-invasive patient discomfort detection using time frame rules and pairwise keypoints distance feature

Nowadays, digitization has opened up various possibilities in the healthcare sector. Smart healthcare is becoming a field of remarkable transformations and growth in the new era of smart cities and digitally interconnected societies. In smart healthcare systems, multiple sensors, devices, and sources are interconnected through networks and collect valuable data that gives insights about the patient's health status and ongoing health trends. Internet of Health Things (IoHT) is a concept that represents the identifiable sensors, devices, and sources connected to the internet and that can be utilized in various automated patient monitoring applications to provide remote healthcare services. This article aims to provide an IoHT driven deep learning-based system for non-invasive patient discomfort detection. The developed system comprises of an RGB camera device to identify discomfort in a patient's body instead of utilizing any wearable sensors or devices. In order to identify movement and discomfort, the system utilizes the YOLOv3 model to detect the patient's body and the Alphapose method to extract different keypoints information of the patient's body. Applying association rules, the detected keypoints are then converted into six central body organs. To detect a patient's discomfort, pairwise distance is measured between detected keypoints, and finally, time frame rules are applied to set a threshold in order to analyze movements as either associated with normal or discomfort conditions. The patient's body is monitored continuously for multiple frames, based on which comfort and discomfort levels are distinguished. Different video sequences are used for experimental evaluations; results reveal the effectiveness of the developed system by achieving a True Positive Rate of 95% with a 6% False Positive Rate.