Anomaly Detection and Diagnosis Scheme for Mobile Health Applications

Abstract

Mobile healthcare applications highly depend on healthcare data, which is collected from wearable or implantable sensors. However, sensor readings may be inaccurate due to resource-constrained devices, sensor misplacement, patient with smearing, and other environmental related causes. Analyzing healthcare data is of paramount importance to provide high quality-care services and reduce false medical diagnosis. In this paper, we propose an online approach to detect inaccurate measurements and to raise alerts only when patients seem to be in emergency situations. The proposed approach is based on robust principal component analysis and adaptive threshold for multivariate anomaly detection, and on contribution plots for univariate anomaly diagnosis. We apply our proposed approach on real medical dataset. Our experimental results prove the effectiveness of our approach in detecting and diagnosing anomalous physiological measurements. The reduced time and space complexities of our approach make it useful and efficient for real time mobile health applications.