Secondary Logo

Journal Logo

A Pilot Study of Respiratory Inductance Plethysmography as a Safe, Noninvasive Detector of Jet Ventilation Under General Anesthesia

Atkins, Joshua H., MD, PhD*; Mandel, Jeff E., MD, MS*; Weinstein, Gregory S., MD; Mirza, Natasha, MD

doi: 10.1213/ANE.0b013e3181f10982
Technology, Computing, and Simulation: Research Reports
Free
SDC

BACKGROUND: High-frequency jet ventilation is an optimal mode of ventilation for many surgical procedures of the trachea and larynx but has limited monitoring modalities to assess adequacy of oxygenation and/or ventilation. Respiratory inductance plethysmography is a noninvasive monitor of chest and abdominal wall movement with well-established applications in the sleep laboratory. We performed an observational pilot study of respiratory inductance plethysmography as a detector of jet ventilation.

METHODS: Twenty-five patients underwent microdirect suspension laryngoscopy with high-frequency jet ventilation under general anesthesia with total IV anesthesia. Inductotrace® bands (Ambulatory Monitoring Inc., Ardsley, NY) were applied to the chest and abdomen in all patients and data collected from oxygen administration through emergence at 50-Hz sampling frequency in the DC mode using a 12-bit A-D converter and custom programmed LabVIEW interface. The raw data were filtered and a detector was developed based on a type I, IIR peak comb filter to differentiate apnea, cardiogenic oscillations, and jet ventilation– associated respiratory excursion. The primary end point was the ability of the detector to identify the presence of jet ventilation. Receiver operating characteristic curves were generated for the aggregate data of all patients.

RESULTS: Respiratory inductance plethysmography reliably detected jet ventilation. The data analysis program effectively extracted a relatively small amplitude jet ventilation signal from a baseline signal contaminated by cardiogenic noise. Sensitivity was in the range of 85%, with a filter bandwidth of 0.055 Hz. Increased sensitivity with increasing filter bandwidth was offset by a detection delay of 12.5 seconds.

CONCLUSIONS: Respiratory inductance plethysmography was successfully used to detect high-frequency jet ventilation in patients undergoing laryngotracheal surgery. This pilot study demonstrates the feasibility of respiratory inductance plethysmography as a monitor for use during jet ventilation.

Published ahead of print August 24, 2010

From the Departments of *Anesthesiology and Critical Care, and Otorhinolaryngology, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania.

Supported by the Department of Anesthesiology and Critical Care, University of Pennsylvania School of Medicine.

Disclosure: The authors report no conflicts of interest.

Reprints will not be available from the author.

Address correspondence to Joshua H. Atkins, MD, PhD, Department of Anesthesiology and Critical Care, University of Pennsylvania School of Medicine, 680 Dulles Building, 3400 Spruce St., Philadelphia, PA 19104. Address e-mail to Joshua.Atkins@uphs.upenn.edu.

Accepted June 19, 2010

Published ahead of print August 24, 2010

Surgical procedures of the larynx and trachea have evolved with development of techniques such as robotic surgery, vocal cord collagen injection, and laser ablation. Microdirect suspension laryngoscopy under total IV general anesthesia is a mainstay in the management of laryngotracheal disease. Endotracheal intubation complicates surgical procedures on the airway for reasons that include physical obstruction of the operative field, provision of a substrate for airway fire during laser surgery, potential contribution to airway edema, and traumatic injury resulting in vocal cord dysfunction, granulation, or stenosis.1

High-frequency jet ventilation (HFJV) is a valuable ventilation strategy for patients undergoing surgical procedures involving the airway. The subglottic technique produces limited interference with the operative field and minimal vocal cord movement, promotes expulsion of surgical debris, and improves safety by visual inspection of appropriate placement.2

A major disadvantage of HFJV is the inability to confirm ventilation with end-tidal carbon dioxide (CO2) or exhaled tidal volume measurement. Practitioners rely on pulse oximetry as the primary monitor of oxygenation and surrogate confirmation of ventilation during HFJV. This can be problematic. Desaturation is a delayed response when a high concentration of inspired oxygen is used. Moreover, adequate oxygenation does not ensure that there is adequate minute ventilation for CO2 removal. Hypercarbia, secondary to hypoventilation, may have untoward clinical sequelae in certain patient populations. Limited monitoring capability likely represents a barrier to the use of HFJV in clinical practices.3 The development of a safety monitor for use during HFJV holds potential to increase the use of jet ventilation during airway surgery.

Respiratory inductance plethysmography (RIP) is a noninvasive monitoring modality. Elastic bands embedded with insulated wires are placed around the thorax and abdomen. When charged with an electric current, the inductance of the bands varies as a function of thoracic and abdominal cross-sectional area.

RIP is widely used in sleep studies to detect apnea and airway obstruction.4 Limited applications for noninvasive assessment of ventilation and lung volumes in both ventilated intensive care unit patients and spontaneously breathing subjects have been described.5,6 RIP as a postoperative apnea monitor in preterm infants was recently reported.7 The use of RIP under operating room conditions and general anesthesia is not well characterized.

RIP is under active study in our group for use as a routine safety monitor during HFJV for interventional procedures under general anesthesia. We recently reported a case in which RIP detected lung hyperinflation and predicted pressure alarms on the jet ventilator during laser resection of severe tracheal stenosis.8 The goal of the present research was to apply RIP as a detection monitor for HFJV during airway procedures. Specifically, we aimed to determine the sensitivity of RIP to detect the presence of HFJV during surgery.

Back to Top | Article Outline

METHODS

The study was designed as an observational pilot study of 25 patients undergoing microdirect laryngoscopy. The study was approved by the IRB, and all enrolled patients consented to participation. Enrolled subjects were among those who presented to the operating room at the University of Pennsylvania for therapeutic or diagnostic microdirect suspension laryngoscopy with planned use of intraoperative jet ventilation. Exclusion criteria were baseline oxygen saturation <92% on room air, history of a difficult intubation, history of failed jet ventilation, body mass index >50 kg/m2, pregnancy, emergent surgery, and presence of an indwelling central venous catheter or pacemaker (to minimize risk of electric conduction by RIP band current to the heart). Recorded baseline patient characteristics included height, weight, oxygen saturation on room air, and chronic use of opioids.

After arrival in the operating room, ASA standard monitors and appropriately sized Inductotrace® RIP bands (Ambulatory Monitoring Inc., Ardsley, NY) were applied on the ribcage (just above the xiphoid cartilage) and the abdomen (at the level of the umbilicus). The bands were attached to the current oscillator, which was prewarmed for a minimum of 10 minutes to achieve stable temperature. Data collection was initiated with sampling at 50 Hz in the DC mode after the patient assumed the supine position used for the duration of the procedure. The voltage was zeroed on the oscillator. To maintain an output within the rails of the A-D converter, the gain of both bands was adjusted during patient peak inspiration and forced expiration. This maneuver was intended to ensure that the digital signal output would not be clipped during the acquisition period. The patient breathed 100% oxygen via a facemask. General anesthesia was induced with lidocaine (0.2–1 mg/kg), propofol (2 mg/kg), and remifentanil (2–4 μg/kg) over 4 minutes and then maintained by infusion to clinical indicators. During the procedure, additional boluses of remifentanil, propofol, or muscle relaxants were delivered at the discretion of the anesthesia team. Once achieved, apnea was maintained until a stable baseline RIP recording was obtained. The baseline was defined as the lack of detectable phasic signal other than cardiogenic oscillation. Mask ventilation was initiated using a pressure-controlled ventilator: 10 breaths/min, 20 cm H2O driving pressure, and 1:1 inspiratory/expiratory ratio. Mask ventilation was maintained for a minimum of 4 breaths.

Laryngoscopy was then performed by the surgeon. Once a stable laryngoscopic view was obtained, HFJV (Monsoon®; Acutronic Medical, Hirzel, Switzerland) was initiated via a metal ventilation cannula placed through the vocal cords. Jet settings (except in 1 case of excessive pressure alarms and during laser surgery) were as follows: 100% fraction of inspired oxygen, 120 breaths/min, 40% inspiratory time, 20 psi driving pressure, and 10% humidity. HFJV was started and stopped as directed by the attending surgeon. Initiation or cessation of jet ventilation was noted manually on the record as a timed event for purposes of data analysis. At the discretion of the attending surgeon, the trachea was intubated with a 6.0-mm endotracheal tube after completion of the portion of the procedure that used jet ventilation. In the subset of patients who were intubated, tidal volumes during volume-controlled ventilation after intubation were recorded as the digital output of the ventilator spirometer. Patients were tracheally extubated upon resumption of spontaneous ventilation and ability to follow commands. Otherwise, intermittent mask ventilation using pressure-controlled (as during induction) ventilation was initiated as necessary to maintain oxygen saturation >92% while awaiting resumption of spontaneous ventilation and emergence from anesthesia.

The RIP thoracic band signal was sampled at 50 Hz using a 12-bit A-D converter and imported into LabVIEW (Version 8.6.1; National Instruments, Austin, TX). The data were manually tagged during data collection to identify each 0.5-second period when HFJV was on (189 minutes) or off (321 minutes). A seventh-order elliptical high-pass filter with a 1-Hz stopband and 60-dB attenuation was implemented in LabVIEW. The sampled data were passed through the filter to remove DC offset. The data set was then subjected to a type I, IIR peak comb filter with a fundamental frequency of 2 Hz (Appendix A).9 The filter's frequency response is illustrated in Figure 1. The output of the comb filter was divided into 0.5-second epochs, and the peak-to-peak height of the signal in each 0.5-second epoch was calculated. Receiver operating characteristic (ROC) curves were constructed using the prec_rec routine in MATLAB (Version 2009b; The MathWorks, Natick, MA).10 The area under the ROC curve was determined by trapezoidal integration over a range of values for filter bandwidth, both with and without compensation for the rise time of the resulting filter.

Figure 1

Figure 1

Back to Top | Article Outline

RESULTS

Twenty-five patients were enrolled and all completed the study. There were no adverse events recorded. Jet ventilation was performed on all study patients. Cardiogenic oscillations were always present and often approached the magnitude of the jet ventilation–induced changes in thorax circumference, as illustrated in Figure 2. This considerable background noise can be progressively attenuated by decreasing the bandwidth of the filter from 0.1 to 0.01 Hz (Fig. 3). The effect of decreasing bandwidth is to introduce a detection delay due to the rise time of the filter as shown in Figure 4. A ROC curve for the detection of whether the jet ventilator is on or off in all 25 patients at 1 value of filter bandwidth is shown in Figure 5. The area under the ROC curve is equivalent to the Wilcoxon signed rank test; an area of 1 is a perfect test, and an area of 0.5 is a coin toss.11 The area under the plot for the sensitivity of the detector is shown in Figure 6. Analysis without correction for the rise-time delay is depicted by the green line. If we exclude data after the jet is turned on or off during this detection delay from the calculation of area under the curve, a more realistic estimate of performance is obtained (Fig. 6, red line). The detector exhibited the highest area under the curve at a bandwidth of 0.055 Hz, which is associated with a detector delay of 12.5 seconds (the rise time at that bandwidth). The rise time and associated detector improvement will vary as a function of fitter bandwidth as discussed below.

Figure 2

Figure 2

Figure 3

Figure 3

Figure 4

Figure 4

Figure 5

Figure 5

Figure 6

Figure 6

Back to Top | Article Outline

DISCUSSION

Confirmation and subsequent monitoring of ventilation during endotracheal anesthesia has been a significant factor in reducing anesthesia-associated morbidity and mortality. HFJV poses a vexing problem for the anesthesiologist, because many of the usual monitors are difficult to duplicate. There is no “gold standard” for monitoring of HFJV. One of the most important monitors of conventional ventilation is the disconnect alarm, which tells us if we are no longer ventilating the patient's lungs. Similarly, during endotracheal anesthesia, changes in phasic pressure, tidal volume, and exhaled CO2 provide robust confirmation of continued ventilation.

Although several authors have described approaches to intermittent assessment of end-tidal CO2 during tubeless surgery, less attention has been devoted to detection of HFJV failure that may occur by disconnect, catheter malposition, device failure, or high chest wall resistance/low pulmonary compliance.12,13 Intermittent measurement of CO2 in the airway may give an indication of long-term ventilation trends but is not a real-time monitor of HFJV.

Assessment of jet ventilation by chest excursion was described by Glenski et al.14 who used pneumobelts to assess manual oxygen injection during laryngoscopy. The respiratory rates were between 40 and 60 breaths/min. Inflation pressures were typically 40 psi and yielded estimated tidal volumes of 250 mL. No mention was made of the effect of cardiogenic oscillations on detection, but given the larger volumes and lower respiratory rates, these effects may not have been apparent. The principal use advocated for this assessment was detection of breath stacking, which might result in barotrauma. Bourgain et al.15 demonstrated a method for measuring end-expiratory pressure (EEP) in HFJV, and correlated this with volume above functional residual capacity using strain gauge measurements of chest and abdominal movement. The primary advantage of this technique is the ability to rapidly detect overdistension to avoid barotrauma. The ability to assess volume above functional residual capacity with EEP was discussed but the technique is limited because it becomes significantly less accurate in the presence of small airway disease. The potential impact of atelectasis on these measurements was not addressed. As such, the ability to rapidly determine efficacy of HFJV from end-tidal CO2, EEP, or chest wall tension may be limited with that technique.

The RIP monitor described herein detected when HFJV was turned off by measuring the absence of chest movement. The raw data were contaminated by cardiogenic oscillations, phasic variation during spontaneous respiration, and occasional artifact. This noise was filtered to reveal the jet ventilation signal. Spontaneous respiration and artifact were readily handled with a high-pass elliptical band filter as described above.

The HFJV signal is produced by a mechanical device with a defined frequency of 120 breaths/min. A comb filter was used because of its ability to produce a narrow passband. The comb filter is a recursive filter that combines a delayed version of the signal with the signal. The filter has a frequency response consisting of regularly spaced peaks that resemble a comb. The structure of the filter is simple; a single delay element and gain specify the filter, making it easy to implement. Because the filter requires few calculations to obtain the result, it is less sensitive to roundoff error. This makes the filter an attractive choice for implementation in inexpensive microcontrollers without double precision floating point capability. The principal design feature and controlled variable of the comb filter is the bandwidth (Df) at a default amplitude threshold of −3 dB. The effect of varying Df on the output of the detector from 0.01 to 0.1 Hz is illustrated in Figure 3. Because the fundamental frequency of the jet ventilator is known, it is a reasonable assumption that the energy imparted to the chest and abdomen will be at this fundamental frequency and its harmonics. In this regard, the comb filter is an efficient means of detecting the jet signal.

For the extraction of the HFJV signal, a type I peak comb filter was selected after consideration of other detection algorithms. The initial high-pass filter stage made a type II peak comb filter (which removes the DC peak present in the type I) redundant. This initial filter stage was chosen to permit us to combine multiple sequences derived from the 25 patients into a single data record to simplify subsequent analysis. A single stage composed of a type II filter might be more appropriate as a real-time, intraoperative detector during HFJV of a single patient and is currently under investigation by our group. Frequency domain techniques such as fast-Fourier transform were considered, but would have required considerably more computational power to achieve the spectral resolution of the comb filter. For detection of a single frequency in close proximity to the predominant noise frequency, the comb filter is a more parsimonious solution. It is relatively immune to fixed-precision arithmetic and thereby makes implementation in a small device more practical.

The detector was developed using comb filtering of the raw RIP signal from the thorax. The detector was able to differentiate between jet ventilation chest movement and chest movement induced by cardiogenic oscillations with a 12.5-second detection delay. Further noise reduction, however, is achieved at the expense of considerably longer rise times for the detector. This occurs because the filter incorporates more old information to permit filtering. This imposes a detection delay that is inversely proportional to the filter bandwidth. The filter has a finite rise time, but the ventilator is either on or off. For a period of time after the ventilator is turned on or off, the output of the filter is in transition. During the rise time, there is a greater probability of generating a false negative for any threshold we might choose to define jet on–jet off. If we use the output of the detector for this period in determining the ROC curve, it will decrease the estimate of the performance of the filter by inclusion of these false negatives. A filter that has a narrower bandwidth will exclude more noise, but will have a longer rise time. By excluding the very brief period of time after transitions in jet state, we get a more accurate estimate of the impact of filter bandwidth on detection accuracy. Depending on the clinical circumstances, a delay on the order of 12.5 seconds could be significant, in which case the anesthesiologist may elect to accept a lower sensitivity (i.e., a greater frequency of false alarms) in exchange for a minimal rise-time delay. The pulse oximeter exhibits a substantially longer delay in alerting the anesthesiologist to a change in the oxygenation status of the patient.

An ideal monitor would provide information regarding adequacy of ventilation during jet. This study was not designed to measure tidal volumes or minute ventilation using the RIP signal. The overall analysis, on which detector performance was measured, used uncalibrated raw data. However, many patients were intubated for a portion of the study. For these patients, calibration of the RIP signal, to obtain an estimate of volumes attained during HFJV, was obtained by placing the patient on volume-controlled ventilation. Volumes included in Figures 2 and 3 were obtained using this calibration methodology and were reported to provide estimations of the range of tidal volumes delivered in representative patients. Tidal volume estimates are not possible for the concatenated data set because calibration was not performed on patients who were not intubated at the end of the procedure.

The discriminating power of this system is high. Indeed, inaccuracy in manual tagging of the jet state may have factitiously decreased the estimated discrimination. This could be remedied by simultaneous, automated detection of ventilator activation that is likely to increase the overall sensitivity of the system. The ROC curve permits us to assess the performance of the detector across a range of thresholds, and informs us whether any threshold can be found that discriminates between the 2 conditions. This study was not intended to identify a single optimal threshold for detection but rather to demonstrate the feasibility of RIP for detection of HFJV within the dynamic range of signals generated. The high performance of the ensemble data suggests that a global threshold could be chosen for all patients, rather than requiring a period of training for each patient. It is conceivable that further research will identify conditions under which alternate thresholds might be desirable to optimize monitoring in certain patient populations.

This study was not powered or designed to predict circumstances under which jet ventilation will fail because of inadequate oxygenation. However, we observed in the first of the 25 patients enrolled that patient oxygen saturation could not be maintained >90% with jet ventilation using 100% fraction of inspired oxygen. Jet ventilation was abandoned and the patient was intubated for the procedure. Analysis of the approximately 1 minute of attempted jet ventilation with the comb filter reveals a jet ventilation signal with amplitude considerably lower than that seen in the other 24 cases. Indeed, in this case, the amplitude of the jet ventilation signal was lower than the cardiogenic oscillations. Applying our RIP system in real time to this patient should have detected no HFJV signal potentially allowing the team to proceed with intubation before the onset of desaturation after which resuscitative measures were the only option. This case also raises the question of what physiologic conditions would lead the HFJV signal to be lower than the cardiogenic oscillation signal. Possibilities include elevated small airway resistance and diminished alveolar compliance. These questions are not addressed by the current study and will require further investigation to resolve. The RIP monitoring system, as currently implemented, cannot quantify adequacy of ventilation or oxygenation. However, the study data demonstrate that RIP can be used to generate a high-fidelity signal under operating room conditions with general anesthesia. Moreover, the system, with the first-pass technical limitations described above, has excellent sensitivity to detect HFJV and thereby function as a monitor to alert the anesthesiologist to both ventilator disconnect or absence of detectable thoracic excursion under the extant conditions. This information might guide the clinician to (1) reconfirm jet ventilation catheter placement and system function, (2) alter jet ventilator settings, or (3) consider alternative modes of ventilation. A system capable of both detecting and quantifying minute ventilation during HFJV would be optimal and we continue to investigate ways to use RIP in this regard.

Back to Top | Article Outline

AUTHOR CONTRIBUTIONS

JHA helped with study design, conduct of study, data analysis, manuscript preparation, and is the author responsible for archiving. This author approved the final manuscript and reviewed the original study data and data analysis. JEM helped with study design, conduct of study, data analysis, and manuscript preparation. This author reviewed the original study data and data analysis. GEW and NM helped with conduct of study.

Back to Top | Article Outline

ACKNOWLEDGMENTS

The authors acknowledge the excellent research support for this study provided by Mary Hammond, BSN, of the Department of Anesthesiology and Critical Care.

Back to Top | Article Outline

REFERENCES

1. Davies JM, Hillel AD, Maronian NC, Posner KL. The Hunsaker Mon-Jet tube with jet ventilation is effective for microlaryngeal surgery. Can J Anaesth 2009;56:284–90
2. Evans KL, Keene MH, Bristow AS. High-frequency jet ventilation: a review of its role in laryngology. J Laryngol Otol 1994;108:23–5
3. Jaquet Y, Monnier P, Van Melle G, Ravussin P, Spahn DR, Chollet-Rivier M. Complications of different ventilation strategies in endoscopic laryngeal surgery: a 10-year review. Anesthesiology 2006;104:52–9
4. Luo YM, Tang J, Jolley C, Steier J, Zhong NS, Moxham J, Polkey MI. Distinguishing obstructive from central sleep apnea events: diaphragm electromyogram and esophageal pressure compared. Chest 2009;135:1133–41
5. Goodman NW, Stratford N. Effect of i.v. lignocaine on the breathing of patients anaesthetized with propofol. Br J Anaesth 1995;75:573–7
6. Brazelton TB III, Watson KF, Murphy M, Al-Khadra E, Thompson JE, Arnold JH. Identification of optimal lung volume during high-frequency oscillatory ventilation using respiratory inductive plethysmography. Crit Care Med 2001;29:2349–59
7. Brown KA, Aoude AA, Galiana HL, Kearney RE. Automated respiratory inductive plethysmography to evaluate breathing in infants at risk for postoperative apnea. Can J Anaesth 2008;55:739–47
8. Atkins JH, Mirza N, Mandel JE. Case report: respiratory inductance plethysmography as a monitor of ventilation during laser ablation and balloon dilatation of subglottic tracheal stenosis. ORL J Otorhinolaryngol Relat Spec 2009;71:289–91
9. National Instruments. Comb Filters (Digital Filter Design Toolkit). Available at: http://zone.ni.com/reference/en-XX/help/371988B-01/lvdfdtconcepts/comb_filters/. Accessed November 18, 2009
10. MATLAB Central. Schroedl S. Precision-Recall and ROC Curves. , 2008. Available at: http://www.mathworks.com/matlabcentral/fileexchange/21528-precision-recall-and-roc-curves. Accessed November 18, 2009
11. Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148:839–43
12. Gottschalk A, Mirza N, Weinstein GS, Edwards MW. Capnography during jet ventilation for laryngoscopy. Anesth Analg 1997;85:155
13. Klein U, Karzai W, Gottschall R, Gugel M, Bartel M. Respiratory gas monitoring during high-frequency jet ventilation for tracheal resection using a double-lumen jet catheter. Anesth Analg 1999;88:224–6
14. Glenski JA, MacKenzie RA, Maragos NE, Southorn PA. Assessing tidal volume and detecting hyperinflation during Venturi jet ventilation for microlaryngeal surgery. Anesthesiology 1985;63:554–7
15. Bourgain JL, Desruennes E, Cosset MF, Mamelle G, Belaiche S, Truffa-Bachi J. Measurement of end-expiratory pressure during transtracheal high frequency jet ventilation for laryngoscopy. Br J Anaesth 1990;65:737–43
Back to Top | Article Outline

APPENDIX A

A comb filter adds a delayed version of the signal to itself, creating a spectrum with regularly spaced spikes, resembling a comb. The comb filter can be implemented with feedforward and feedback; the current effort uses both. A block diagram is illustrated in Figure A1.

Figure A1

Figure A1

The z−N denotes a delay of N sampling intervals. For a sampling frequency (fs) of 50 Hz, a 25-element delay yields a 0.5-second delay, corresponding to the jet ventilator rate of 120 breaths/min. The transfer function is given by equation A1:

The filter is characterized by several parameters, as illustrated in Figure A2.

Figure A2

Figure A2

The parameter f0 is the center frequency for the fundamental frequency of the filter, in this case 2 Hz. The parameter Δf is the bandwidth at −3 dB. The choice of sampling frequency and the number of delay elements specifies f0, and Δf can be derived from the values of a and b. Although only 3 peaks are represented in Figure A2, peaks will occur at every integer multiple of f0 from 0 to fs/2.

The simplicity of the filter is evident; a delay line, 2 summing junctions, and 2 gains are all that are required to implement the filter. The filter is well suited to analysis of periodic signals that contain harmonics of a fundamental frequency because every harmonic is equally represented in the output at 0 phase shift.

The recursive nature of the filter causes the output to increase over time when presented with a signal, as illustrated in Figure A3. Here, a sine wave of amplitude 1.0 and frequency of 2 Hz is applied to the filter. The envelope of the signal asymptotically increases to 1.0 over approximately 20 seconds. This property of the filter explains why the decisions made during the first 12.5 seconds after a transition are less reliable; the filter output has not yet stabilized.

Figure A3

Figure A3

© 2010 International Anesthesia Research Society