An Event-Based Approach to Modeling Complex Data in Critical Care
Precision medicine is an emerging field in critical care research where treatments are individualized, ensuring the right patient receives the right medication at the right time. In this work, the focus is on the data analytics, however, interpreting and understanding biomedical data is challenging due to the heterogeneity, noise, potential interrelationships, and dynamic nature that makes the data complex. Computer science is needed to help organize and utilize this data. Exploratory data analysis is an approach to research that combines computing power and human involvement, to enable the interactive exploration of data through visualization with the goal of generating new hypotheses. Intensive care units (ICUs) are becoming a popular setting for precision medicine and exploratory data analysis due to the abundance and complexity of the data collected. One specific area of interest is in physiologic waveform monitoring and the responsiveness of these waveforms to treatment with vasoactive medications. In this thesis we first investigate statistical techniques for analyzing physiological waveform data. Problems with this approach are discussed, including the preponderance of significant p-values, and the obscuring of important relationships between data types. Due to these problems, time series analysis and machine learning methods including clustering are explored. We undertake a prospective observational study which generated a novel multi-modal dataset of waveform data with associated clinical events. Data from this study led to the development of EVENTIVe, an exploratory event-based interactive data visualization platform. EVENTIVe provides an efficient and user-directed means of investigating complex physiological and clinical time series. An event-based data structure is described, including semi-structured document-style storage formats. Time series changepoint detection, hierarchical clustering, and K-Means clustering are included as exploration options. EVENTIVe encourages users to review individual data points through visualization, in addition to measures of central tendency and dispersion. Through EVENTIVe insights into the effects of vasoactive interventions in the ICU on physiological parameters were explored with new hypotheses generated. This work sets the stage for many future studies into the short-term physiological disturbances occurring in critically ill patients, helping the progression towards the use of precision medicine in these settings.
URI for this recordhttp://hdl.handle.net/1974/24489
Request an alternative formatIf you require this document in an alternate, accessible format, please contact the Queen's Adaptive Technology Centre
The following license files are associated with this item: