Classification-Based Screening of Phlebopathic Patients using Smart Socks

Abstract

Telemedicine consists in the delivery of health care services, where patients and providers are separated by distance. Telemonitoring facilities play an important role in remote assistance programs, particularly in assisting patients suffering from chronic afflictions, such as phlebopathic diseases (e.g. chronic venous disease and diabetic foot). When these pathologies worsen, complications can be serious. In fact, foot deformities lead to variations of plantar load, formation of ulcers and, in the worst case, to amputation. Consequently, these pathologies cause huge expenses for the health care system. We propose a framework for screening and early detection of phlebopathic diseases insurgence, based on dynamic tests for functional assessment where patients wear sensorized socks. Socks used in this study integrate force and inertial sensors to provide information on plantar pressures and person’s movement. We show results of a feasibility study including 42 patients, with a balance of 21 healthy patients and 21 with phlebopathic diseases. Data gathered from wearables were automatically elaborated through machine learning techniques in order to obtain a binary classifier identifying whether or not a patient shows pathological gait. Results show that our best classifier has high positive predictive value and high sensitivity, with F1-score equal to 92.1%.