Smart sock-based machine learning models development for phlebopathic patient screening

Internet of Things (IoT) systems are valid solutions for remote health monitoring. The combination of technology and network with medical knowledge leads to important telemedicine solutions. Telemedicine is the practice of medical care where patient and providers are separated by distance. Telemonitoring of chronic diseases is one of the most important aspects of telemedicine. Chronic diseases such as chronic venous disease and diabetic foot require continuous monitoring in order to prevent complications, which can become very burdensome, with variations in plantar pressures, deformations, ulcers, up to amputations. The quality of life of those suffering from these diseases is strongly affected and furthermore the continuous monitoring causes great expenses for the health system. We propose an IoT infrastructure for the screening and recognition of patients suffering from phlebopathic disease, based on the use of a smart sock. The smart sock integrates pressure sensors and inertial sensors, worn by patients during a dynamic test. The main contribution of this work is the development of machine learning models on the collected and processed data of a sample population of 24 participants, 13 healthy and 11 suffering from phlebopathic diseases. The best classifier obtained shows very good performance, with an accuracy of 95.83%, high precision, recall and a F1-Score equal to 0.95, making the proposed framework a good candidate as system for screening phlebopathic patients and for their early diagnosis.