|
|
October 16-20, 2010 San Diego, CA Home Abstract Archive Search Abstracts 2010 Session Grid 2010 Meeting Website Policy Statements ASA Website Feedback 
Visit Anesthesiology.org |
| Previous Abstract | Next Abstract |
| Printable Version |
|
A1107
October 15, 2007 9:00 AM - 11:00 AM Room Hall D, Area D, |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Clinical Evaluation of iAssist: A Novel Software Tool To Detect Change in Physiological Monitoring | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
J. Mark Ansermino, M.D., F.R.C.P.C., Joanne Lim, M.A.Sc., Ping Yang, M.A.Sc., Chris Brouse, B.S., Guy Dumont, Ph.D. Anesthesiology, Pharmacology & Therapeutics, University of British Columbia, Vancouver, British Columbia, Canada |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Introduction: Subtle changes in monitored physiological data can be used to guide clinical actions and give early warning of potential adverse events. Successful situation assessment includes detecting temporal changes to patterns in trends over time. We have developed a software tool (iAssist) for tracking the behavior of dynamic systems and automatically detecting key events in the processes over time. Following successful offline testing, the tool has been evaluated in real time in the operating room. Methods: Heart rate (HR), end tidal carbon dioxide (EtCO2), exhaled minute ventilation (MVexp), and respiratory rate (RR) were modeled using a dynamic linear growth model, whose noise covariances were estimated by an adaptive Kalman Filter based on a recursive Expectation- Maximization method. Events are detected by adaptive local Cumulative Sum (CUSUM) testing (1, 2). The events in mean noninvasive blood pressure (mean NIBP) and oxygen saturation (SpO2) are detected using CUSUM testing on a filtered residual from an Exponentially Weighted Moving Averaging filter (3). The algorithms were optimized to detect events in offline and real-time data and were implemented in a Java® software environment. Following ethical approval, the tool was evaluated in real time alongside current monitors. A median filter was optimized for each trend to remove artifacts and transients. Tuning parameters were fixed based on previous testing. Events from the stable period of anesthesia were marked by the clinician in real time. Each event detected by iAssist was graded (artifact (A), clinically insignificant (CI), clinically significant (CS), clinically significant with action taken (CSAT), clinically significant due to intervention (CSI)) using a graphical display of the trends and events marked on the trend display. Missed events were noted. The usefulness of each event was rated on a 10cm visual analogue scale (VAS). Results: Fifteen anesthesiologists (38 cases; mean duration of 103 (4.25) minutes) completed the evaluation in a variety of pediatric surgical cases of a least one hour duration. The type of events are shown in Table 1. The mean number of events per case was 22.8 (SD 23.4). The usefulness value ranged from 0 to 9.3 (for an EtC02 CSAT event). Only 6 significant events were reported as being missed by the algorithms. Discussion: The algorithms detected 13 events per hour of anesthesia, however many of these were related (e.g. MVexp, RR, CO2) to a single clinical event. Fifty percent of events were rated clinically significant (CS, CSAT or CSI). SpO2 had a low number of events but many artifacts. The CSI events could have been suppressed by information about clinical interventions. 1. Yang P, et al. IEEE Trans Biomed Eng 2006; 53(11): 2211-2219. 2. Yang P et al. 27th IEEE EMBS Conference; 2005 3. Yang P et al. 28th IEEE EMBS Conference; 2006.[table1] Anesthesiology 2007; 107: A1107 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|