TSEA: An Open Source Python-Based Annotation Tool for Time Series Data

We present the Time Series Event Annotator (TSEA), a graphical user interface annotation tool for time series data that enables rapid visualization, labeling, and annotation of signals, including individual points and ranges. Time series data are common to a variety of applications. Oftentimes there is a need to label segments and/or points of the signals, highlighting important elements that are later used for feature extraction or for signal analysis. A number of illustrative applications of the developed tool are discussed, particularly for the detection of "R" peaks from electrocardiogram signals. While algorithms for detection of "R" peaks can achieve good results when applied to an electrocardiogram signal with a high signal-to-noise ratio, they often lead to incorrect detections in the presence of noise or motion artifact commonly found in clinical setups. In such cases, the Time Series Event Annotator (TSEA) enables efficient imputing of missed or incorrect "R" peak detections, leading to increased data integrity for downstream analysis, at minimum cost. Considering that data cleaning often represents the majority of effort when developing a new machine learning pipeline, our annotation tool will accelerate the development of a wide range of new machine learning applications.