Abstracts

RATIONALE: Patients admitted to the neuro-intensive care unit (NICU) are frequently post-operative neurosurgical or head trauma cases. Others commonly admitted to the NICU are patients in status epilepticus (SE). Patients may be in coma as a result of a complex combination of insults, or may have been rendered unconscious by the treatment (e.g. barbiturate coma to treat SE). An especially important parameter to follow in postop as well as SE patients is the level of anesthesia --- as manifest by a burst-suppression EEG pattern.
Development of a flexible research tool to study the feasibility and clinical usefulness of continuously recorded and processed neurophysiological data in acutely ill patients at risk for, or undergoing treatment for, epileptic seizures and stroke.
METHODS: We have developed a NICU monitoring system which continuously applies a battery of EEG and evoked potential tests, selected and programmed by the physician. Routine physiological parameters are also periodically obtained electronically from the bedside monitors. For each timed EEG epoch recording, the power spectral density is estimated from the averaged periodogram (squared Fourier spectrum). Power in the traditional EEG bands is calculated for each of the electrode derivations, typically 8 channels. For each electrode, a count of the number of bursts is performed.
The first step in the burst detection algorithm is removal of baseline drift. Next, a 700 ms window is moved across the data in steps overlapping by 100 ms. Power in each window, computed from the average sum of squares of the sample voltages, is compared with the power in the preceding window. If the energy has increased more than seven-fold, the event is counted as a burst. Successive bursts must be separated by a non-burst interval of at least 100 ms.
Trends are displayed continuously at the bedside and can also be accessed via the world wide web (with appropriate security). Remote access to the main menu and control protocol is via simple HTML pages, and trend results are generated at the moment of access by CGI scripts written in Tcl/Tk. Since the monitoring system is connected to the network, automatic email can be sent to the neurophysiologist attending to the patient, prompted by thresholds which have been exceeded or by error conditions. At our institution, the email facility can also be used to automatically send a message to the physician[ssquote]s pager.
The system has been initially tested by collecting data from one normal subject, and for a mean of 12 hours in 7 comatose NICU patients chosen because changes in their condition were anticipated.
RESULTS: The system was not hampered by the electrical noise in the NICU, however detachment of electrodes was a frequent problem. Detection, processing, archiving, and display was activated by the technologists, and pager notifications were generated when changes occurred.
CONCLUSIONS: Recording and processing of multi-modality clinical neurophysiological signals over extended periods of time has been successfully carried out. Identification of both normal and abnormal waveforms is satisfactory in our system, provided that the electrodes remain attached. The system can potentially detect changes in NICU patients as they occur.
[Supported by: CCF]