The Annotation Station: An open source technology for data visualization and annotation of large biomedical databases

LIBRARIES The author hereby grants to M.I.T. permission to reproduce and distribute publicly paper and electronic copies of this thesis and to grant others the right to0p so. Over the past 20 years, there has been a large ongoing effort in the Laboratory for Computational Physiology (LCP) to collect and annotate large databases of physiologic signals in order to facilitate an open initiative to develop algorithms to automate a variety of clinical tasks [1]. The success of this approach motivated the collection of a temporal database of signals typically collected in the intensive care unit (ICU) from the Beth Israel Deaconess Medical Center (BIDMC), known as the Multi-parameter Intelligent Monitoring for Intensive Care (MIMIC) database [11]. Although certain basic clinical information was recorded about the patients in this database, much of the information available from the ICU, such as nursing notes, were not recorded. Approximately 5 years ago, the LCP began an initiative to collect a massive temporal database consisting of almost all the information available in the BIDMC ICU [3] known as MIMIC II. This database not only consists of the beside monitor waveforms collected in the original MIMIC database, but also most of the sources of information available to the clinical staff, ranging from medication drip rates and fluid balances, to subjective evaluation scales and discharge and clinical progress notes detailing patient interventions, reactions, clinical visits and past medical history. The complexity of this database presents many challenges, but also many opportunities for the application of technology to improve patient care. In particular, previous review and annotation schemes are inappropriate for this type of data. A system for efficiently presenting this information in a coherent manner to an expert reviewer is required. Furthermore, an updated annotation structure for providing machine-readable descriptors of the clinical evaluation through time must be developed. This thesis details a system that attempts to provide a framework to facilitate the annotation process. The implemented solution is an Annotation Station used for clinical data review and annotation recording. 1.1 Challenges and opportunities in the ICU Current trends in ICU demographics include, aging populations [13], a reduction in the clinical staff to patient ratio [19], increased reliance on technology [14] and an increase in the volume of data recorded for each patient [15]. All of these factors lead to an increasing stress on modem ICUs. Within the ICU, a tremendous amount of data is collected, displayed …