PCA-based multivariate anomaly detection in mobile healthcare applications

Abstract

Real time mobile Health applications highly depend on sensor readings to provide high-quality health services. However, real-time sensor readings may be inaccurate and cause abnormal physiological measurements due to internal and external factors. Thus, abnormal readings have a significant impact on the reliability of such applications and consequently affect the patient's life. This paper addresses the following issue by proposing a robust approach for online detection of abnormal medical measurements. The proposed approach is based on robust Principal Component Analysis (PCA) to analyze collected physiological measurements from sensors and detect the occurrence of multivariate anomalies based on squared prediction error at runtime. We apply our proposed approach on real medical dataset. Our simulation results prove the effectiveness of our approach in achieving good recall with a low false alarm rate. The reduced time and space complexity of our approach make it useful and efficient for real time settings.