File: mimic.txt G. Moody 28 April 1994 Last revised: 25 May 1997 MIMIC Database Copyright (C) Massachusetts Institute of Technology 1997. All rights reserved. Overview ======== As the sophistication of diagnostic and monitoring devices grows, care of acutely ill patients requires rapid assessment and accurate interpretation of a large and ever-growing volume of medical data. Automated decision support systems capable of assimilating these numerous data streams, tracking and anticipating clinically significant changes, and presenting pathophysiologic hypotheses, would be of great value to clinicians facing ``information overload'' in the intensive care unit (ICU). We have begun to explore the design of ``intelligent patient monitors'' to accomplish these goals. To support this effort, we are creating the MIMIC (Multi-parameter Intelligent Monitoring for Intensive Care) Database. The database, currently nearing completion, will include 100 patient records, each typically containing between 24 and 48 hours of continuous data recorded from patient monitors in the medical, surgical, and cardiac intensive care units of Boston's Beth Israel Hospital. Each record will be accompanied by detailed clinical data derived from the patient's medical record and from the hospital's on-line medical information systems. We select patients to record from those likely to be hemodynamically unstable during the planned recording period. Our goal is to record two or three ECG signals, blood pressure, and respiration in each case; any other available signals will also be recorded. The database will be preserved in archival form on recordable compact disks (CD-R media, readable on standard CD-ROM drives), each containing 1 to 3 patient records of 200 to 600 megabytes per record. This directory contains a 30-minute excerpt of the 42-hour record 237 of the MIMIC Database, readable as either a composite record named `237', or as three (ordinary) 10-minute records named `23700001', 23700002', and `23700003'. Note that software linked with versions of the DB library earlier than version 9.0 *cannot* read these records successfully, because they contain signals sampled at more than one frequency. The software provided elsewhere on this disk has been linked with a suitable version of the DB library, and may be used to read these records. Other software that uses the DB library can simply be relinked with the current version of that library, also provided on this disk (see software/lib for details; if you make no changes to existing software, the DB library transparently decimates all signals to a common sampling frequency). Typically, the data for each 24-hour period occupy 200 to 250 megabytes, so that MIMIC CD-ROMs generally contain between 2 and 3 patient days, which may include between one and three patient records. (Most of the recent records have had durations of 45 hours or more; these long records appear on individual CD-ROMs, occasionally paired with shorter records when available.) The waveforms were acquired and digitized by Hewlett-Packard Component Monitoring System (Merlin) monitors. The monitors were equipped with two dual-ported serial communications cards; one port on each card was used to transmit data acquired by the monitor at 38400 baud to a Gateway 2000 PC (model 4DX2-66V) equipped with a DigiBoard PC/4e serial interface card. The data files were created by custom software running in the PC. Each set of files contains 10 minutes' worth of data received from the monitor. (This scheme is employed so that, for example, loss of power to the PC will not result in loss of all recorded data.) Signal (.dat) files in the MIMIC database contain all samples acquired by the monitor (500 samples per second for each ECG signal, and 125 samples per second for all other signals). The `.txt' files contain a log of all numerical measurements, alarms, and monitor status messages obtained from the monitor. (The numerical measurements are reported at 1.024-second intervals.) The `.al' annotation files contain a sampling of the alarms obtained from the monitor (only the first alarm in each set of identical consecutive alarms is recorded in these files -- for the complete set of alarms, see the `.txt' files). The header (.hea) files for the ordinary records name the signal files and describe the signals; the header file for the composite record names the ordinary records that are its constituent parts. The format of the files on the MIT-BIH Arrhythmia and Polysomnographic Database CD-ROMs, the European ST-T Database CD-ROM, and the MGH/MF Waveform Database CD-ROMs is upwardly compatible with that used for the MIMIC Database (i.e., software linked with DB library version 9.0 or later can read any of these files). This directory also contains record 237n, which has been derived from the complete set of `.txt' numerical measurement files for record 237, and contains over 42 hours of heart rate, blood pressure (mean, systolic, and diastolic ABP and PAP), respiration rate, and SpO2 measurements formatted as a signal file. (There are occasional interruptions in the data as transducers were disconnected and reconnected. The MIMIC Database includes clinical data for each record. This information is derived from the subjects' medical records (including symptoms, diagnoses, progress notes, and medications administered) and from the hospital's clinical computing systems (primarily laboratory results). The clinical data are contained within the *.htm files in this directory; begin by opening `index.htm' using any web browser. If you have set up `wavescript' (under Linux, Solaris, or SunOS) or `wvscript' (under MS-Windows) as a viewer (helper) application for your web browser, the *.xws files in this directory provide links from the *.htm files to the signals in record 237 (and 237n, if you are using `wavescript'). Reference ========= Moody GB, Mark RG. A database to support development and evaluation of intelligent intensive care monitoring. Computers in Cardiology 1996, pp. 657-660. (Included on this disk in hypertext form; open ../html/mimic/mimic.htm using your web browser.) Further information =================== If you have questions about this sample record or about the MIMIC Database project, please write to: George B. Moody Harvard-MIT Division of Health Sciences and Technology Massachusetts Institute of Technology, Room 20A-113 77 Massachusetts Ave. Cambridge, MA 02139 USA e-mail: george@hstbme.mit.edu