Visualizing Multivariate Time Series Data to Detect Specific Medical Conditions

Abstract

Efficient unsupervised algorithms for the detection of patterns in time series data, often called motifs, have been used in many applications, such as identifying words in different languages, detecting anomalies in ECG readings, and finding similarities between images. We present a process that creates a personalized multivariate time series representation—a Multivariate Time Series Amalgam (MTSA) — of physiological data and laboratory results that physicians can visually interpret. We then apply a technique that has demonstrated success with the interpretation of univariate data, named Symbolic Aggregate Approximation (SAX), to visualize patterns in the MTSAs that may differentiate between medical conditions such as renal and respiratory failure.