1 Brain Research Laboratories, New York University School of Medicine, New York, NY, USA. 2 Nathan S. Kline Institute for Psychiatric Research, Orangeburg, New York, NY, USA. 3 Department of Anesthesiology, Brigham and Womens Hospital, Boston, MA, USA. 4 Department of Anesthesiology and Intensive Care, Charité Hospital, Berlin, Germany
*Corresponding author. E-mail: leslie@brl4.med.nyu.edu
Accepted for publication: October 1, 2003
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Methods. The PSI was constructed from a systematic investigation of a database containing QEEG extracted from the analyses of continuous 19 channel EEG recordings obtained in 176 surgical patients. Induction was accomplished with etomidate, propofol, or thiopental. Anaesthesia was maintained by isoflurane, desflurane, or sevoflurane, total i.v. anaesthesia using propofol, or nitrous oxide/narcotics. It was hypothesized that a multivariate algorithm based on such measures of brain state, would vary significantly with changes in hypnotic state.
Results. Highly significant differences were found between mean PSI values obtained during the different anaesthetic states selected for study. The relationship between level of awareness and PSI value at different stages of anaesthetic delivery was also evaluated. Regression analysis for prediction of arousal level using PSI was found to be highly significant for the combination of all anaesthetics, and for the individual anaesthetics.
Conclusions. The PSI, based upon derived features of brain electrical activity in the anterior/posterior dimension, significantly co-varies with changes in state under general anaesthesia and can significantly predict the level of arousal in varying stages of anaesthetic delivery.
Br J Anaesth; 2004 92: 3939
Keywords: anaesthesia, general; monitoring; Patient State Index (PSI); sedation, hypnosis
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The Patient State Index (PSI) was constructed from a retrospective exploration of the multivariate changes in brain electrical activity observed from loss to return of consciousness. It was hypothesized that an algorithm accounting for the maximum electrophysiological variance of this process, minimizing redundancy and maximizing sensitivity to changes in state, could be used to construct an index sensitive to changes in hypnotic state. In an earlier publication,14 we described preliminary findings. This paper describes this process and demonstrates the relationship between different states of hypnosis and the PSI value.
![]() |
Methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Anaesthetic procedures
One of three anaesthetic regimens was administered at the discretion of the anaesthetist: (i) propofol total i.v. anaesthesia (TIVA); (ii) inhalation anaesthesia with isoflurane, sevoflurane, or desflurane (GAS); or (iii) nitrous oxide/narcotic (N/N), occasionally supplemented with propofol. All patients received either midazolam or fentanyl 30 min before induction of anaesthesia. At the start of induction, each patient was instructed to count backwards. The method of induction was at the discretion of the anaesthetist and included either etomidate, thiopental, gas or other agents. When patients stopped counting, the eyelash reflex was checked repeatedly until absent. The patient then received either a non-depolarizing neuromuscular blocking drug, or succinylcholine, in clinically determined amounts, in order to facilitate tracheal intubation. For the propofol and potent inhalation anaesthetic maintenance techniques, the anaesthetist chose whether to administer supplemental nitrous oxide and, for the propofol technique only, fentanyl.
EEG data acquisition
Nineteen electrodes were fixed to the scalp with paste, at positions corresponding to the International 10/20 Electrode Placement System.15 In addition, the following electrodes were used: mastoid electrodes, electro-oculogram electrodes diagonally above and below the orbit of the eye, for detection of eye movement artifact, a ground electrode placed upon the cheek, and an ECG lead on the chest. Recordings were monopolar, referenced to linked earlobes. Amplifiers had a band pass from 0.5 to 70 Hz (3 dB points), with a 60 Hz notch filter. All impedances were kept below 5000 ohms, checked regularly throughout the surgical procedure. The A/D converter sampled at 200 Hz per channel, with 12-bit resolution. The data were reduced to 100 Hz before analysis, using Fants resampling algorithm, which minimizes aliasing.16 All recordings were collected using Spectrum 32 EEG Acquisition Systems (Cadwell Laboratories, Kennewick, WA).
Experienced technicians, augmented by an automatic EEG artifact detection algorithm, edited EEG data visually. At each anaesthetic state (defined below), an artifact-free sample was selected for quantitative analysis, containing from 24 to 48 segments each 2.5 s in duration, with the exception of induction in which only the last 12 segments (30 s) before loss of consciousness were used because induction often occurred quickly and its duration varied widely.
For purposes of this study, the time intervals selected reflected stable periods with sufficient data for estimation of the brain state under the selected conditions. While this may result in less sensitivity of the PSI measure and longer response time to events, it was considered to represent a conservative approach. The implementation of the PSI in the Patient State Analyzer is based on continuous updating of the index every 1.25 s.
EEG data analysis
As duration of the surgical procedures varied from patient to patient (see Table 1), a standardized set of anaesthetic states was identified in each case. These states included data from: (1) pre-operative state, day before surgery, with no pre-operative medication; (2) baseline, recorded outside the operating room after delivery of pre-operative sedation (note: this state was taken as baseline in this study since the prior state was only obtained in a subset of patients); (3) induction, recorded from the pre-operatively sedated patient on the operating table during induction, while the patient counted backwards just before cessation of counting; (4) loss of consciousness, recorded immediately after cessation of counting, loss of eyelash reflex and loss of response to painful stimuli; (5) maintenance, averaged across an uneventful period of anaesthesia during surgery, at approximately the mid-point of anaesthetic delivery; (6) spontaneous somatic events, recorded during maintenance of anaesthesia just before a reported unexpected somatic event (e.g. eyes opened, arm or leg movement, head movement); (7) emergence, recorded approximately 10 min before eye opening (emergence 2) and approximately 5 min before eye opening after the maintenance anaesthetic had been discontinued (emergence 1); and (8) return of consciousness, recorded immediately after the patient opened their eyes, in response to a loud verbal command using the patients name, and showed aversion to a noxious stimulus.
|
Following neurometric QEEG procedures,17 18 all features were transformed to obtain normal distributions and standardized to yield Z-scores, relative to each subjects individual response to anaesthetic and the variance of the baseline (BL) state of the total anaesthetic database of 176 cases. The use of such self-norming takes into account individual variability of the EEG to anaesthetic agents.19 Using a proprietary multivariate discriminant algorithm based on these self-normed indicators of specific states, PSI values were computed for each of the selected states.
EEG data was collected utilizing circuitry optimized to exclude electrical contamination from the environment. Data from the subset of electrode sites used in calculating the PSI were selected from this full set of data. These sites include two anterior (FPI and FPZ'), a midline central (Cz) and a midline posterior (Pz) scalp locations, spanning anterior to posterior dimensions. All other data were maintained in a consciousness database for further study. The selected sites represent a minimal set of electrodes identified as necessary to reflect the significant changes that occur with loss and return of consciousness, with attention also to the gradient shifts between frontal and posterior regions. As, ultimately, the PSI was to reside in a clinical instrument for monitoring level of arousal, it was important that the required electrode sites could be easily incorporated into a clinical appliance for use in the operating room environment.
Computation of the PSI
Following the frequency analysis of the artefact-free EEG signals a subset of features found to account for most statistical variance related to hypnotic state are derived for input to a multivariate discriminant algorithm (proprietary). These features include: absolute power gradient between frontopolar and vertex regions in the gamma band; absolute power changes between midline frontal and central regions in the beta band and between midline frontal and parietal regions in the alpha band; total spectral power in the frontopolar region; mean frequency of the total spectrum in midline frontal region; absolute power in the delta band at the vertex; and posterior relative power in the slow delta range.
Every element in the set of selected features is transformed to a standard score (Z-score) relative to its distribution in a specific reference state and expressed as the probability of deviation from that state. The current values of these standardized scores are the inputs to the calculation of the PSI value. The PSI is the ratio of the probability that the observation belongs to the reference state vs the sum of the probabilities that the observation belongs to either the reference state or to a different level of arousal. Thus, the PSI value can range from 0 to 100. A recent publication by Drover and colleagues20 provides additional details of this computation.
Arousal scores
In order to estimate the statistical relationship between PSI and arousal level, arousal scores were assigned retrospectively to each selected stage using the Observers Assessment of Alertness/Sedation Scale (OAA/S, rated 05). (In the OAA/S scoring, 5=responds readily to name spoken in normal tone, 4=lethargic response to name spoken in normal tone, 3=lethargic response to name spoken loudly and repeatedly, 2=responds only to name spoken loudly after a mild painful stimulus (train of four), 1=responds only to name spoken loudly after a moderately painful stimulus (50 Hz electrical stimulation), 0=no response to verbal or painful stimulus. Same scale used in our previously published study by Gugino and colleagues.21) This was done using a conservative estimate of responsiveness agreed upon by the attending anaesthesiologists responsible for these cases as follows: awake sedated was assigned a 4.5, end of induction before intubation was assigned a 0, early surgical plane was assigned a 0, spontaneous somatic events during maintenance were assigned a 3, approximately 10 min before return of consciousness (eye opening) during emergence was assigned a 2.5 and return of consciousness (eye opening) was assigned a 4. It is noted that no significant site by state interactions were found (repeated measure ANOVA (P=0.15)), which served as assurance that the states were similar across the international sites.
Statistical analyses
The hypothesis that the PSI index would be significantly related to arousal state, that is the level of hypnosis, was tested using ANOVA for the significance of the difference between PSI values at baseline as compared with each of the other states, for all anaesthetics and for each class of anaesthetics separately. Additionally, the significance of the relationship between PSI and level of hypnosis was evaluated using regression analysis, with PSI as the independent variable and arousal score as the dependent variables.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Induction of anaesthesia was achieved by injection of a bolus of etomidate (in 36% of the cases), thiopental (in 27% of the cases), propofol (26% of the cases), inhalation via a facemask (4% of the cases), or injection of other agents (7% of the cases). For maintenance of anaesthesia, there were 49 TIVA cases, 68 GAS cases, and 59 N/N cases.
Nitrous oxide was used in 71% of the GAS cases. For patients whose anaesthetic was maintained with TIVA, propofol was administered at an infusion rate ranging from 100 to 200 µg kg1 min1. Eighty-four per cent (84%) of the TIVA cases received nitrous oxide, with an end-tidal concentration of 5060%. In the nitrous oxide narcotic technique, the end-tidal concentration of nitrous oxide was 6070%. The distribution, duration of analgesic/anaesthetic agents and length of surgery are shown in Table 1. Table 2 gives the average concentration of administered anaesthetic agents during the maintenance stage.
|
|
|
|
PSI vs OAA/S
The regression curves for the relationship between PSI value and OAA/S score are presented in Figure 2, for: (i) all cases combined, independent of anaesthetic (n=176, Fig. 2A); (ii) only volatile anaesthetics, GAS (n=63, Fig. 2C); (iii) only TIVA (n=46, Fig. 2B); and (iv) only N/N (n=57, Fig. 2D).
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The stable relationships, which have been retrospectively ascertained between clinical state and the value of the PSI in these numerous individual clinical cases, support the proposal that real-time computation of the PSI might serve as a reliable clinical monitor to assess the level of consciousness (sedation/hypnosis) throughout surgical procedures and with a wide variety of anaesthetic regimens. Further, while only a small number of spontaneous somatic events occurred in this population, the highly significant increase in PSI observed just before the event, suggests the clinical value of the index for predicting changes in state.
A further demonstration of the predictable relationship between anaesthetic state and PSI value was obtained in a normal volunteer study, in which more precise relationships between anaesthesia delivery and level of hypnosis could be evaluated. These results, presented elsewhere21 support the sensitivity of the index. Taken as a whole, such data suggest the clinical utility of monitoring QEEG, using PSI, throughout anaesthesia delivery, as an adjunct to standard clinical monitors. A prospective multi-site validation of the PSI is underway and results are presented elsewhere.20
![]() |
Acknowledgements |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 Schultz B, Schultz A, Grouven U, Zander I, Pichlmayr I. Changes with age in EEG during anesthesia. Anaesthesist 1995; 44: 46772[CrossRef][ISI][Medline]
3 Werry C, Neulinger A, Eckert O, Lehmkuhl P, Pichlmayr I. Age-related correlation between EEG parameters and depth of anesthesia under propofol. Effect of fentanyl. Anaesthesist 1996; 45: 72230[CrossRef][ISI][Medline]
4 John ER, Prichep LS, Kox W, et al. Invariant reversible QEEG effects of anesthetics. Conscious Cognit 2001; 10: 16583[CrossRef][ISI][Medline]
5 Kearse LA, Rosow C, Zaslavsky A, Connors P, Dershwitz M, Denman W. Bispectral analysis of the electroencephalogram predicts conscious processing of information during propofol sedation and hypnosis. Anesthesiology 1998; 88: 2534[ISI][Medline]
6 Rampil IJ. A primer for EEG signal processing in anesthesia. Anesthesiology 1998; 89: 9801002[ISI][Medline]
7 Glass PS, Bloom M, Kearse L, Rosow C, Sebel P, Manberg P. Bispectral analysis measures sedation and memory effects propofol, midazolam, isoflurane, and alfentanil in healthy volunteers. Anesthesiology 1997; 86: 83647[ISI][Medline]
8 Sebel PS, Lang E, Rampil IJ, et al. A multicenter study of bispectral electroencephalogram analysis monitoring anesthetic effect. Anesth Analg 1997; 84: 8919[Abstract]
9 Payne FB, Sebel PS, Glass PS. Bispectral Index (BIS) monitoring allows faster emergence from propofol, alfentanil/N2O anesthesia. Anesthesiology 1996; 85: A1056[ISI]
10 Song D, Joshi G, White PF. Titration of volatile anesthetics using bispectral index facilitated recovery after ambulatory anesthesia. Anesthesiology 1997; 87: 8428[ISI][Medline]
11 Viertio-Oja H, Sarkela M, Talja P, Tolvanen-Laakso H, Yli-Hankala A. Entropy of the EEG signal is a robust index for depth of hypnosis. In: 2000 ASA Meeting Abstracts. San Francisco, CA: Lippincott Williams & Wilkins, 2000
12 Roy RJ, Zhang XS. Clinical evaluation of EEG complexity measure for depth of anesthesia estimation. In: 2000 ASA Meeting Abstracts. San Francisco, CA: Lippincott Williams & Wilkins, 2000
13 Mantzaridis H, Kenny GN. Auditory evoked potential index: a quantitative measure of changes in auditory evoked potentials during general anaesthesia. Anaesthesia 1997; 52: 10306[ISI][Medline]
14 Prichep LS, Gugino LD, John ER, Chabot RJ, Rausch L, Kox W. In: Annual Meeting of Society for Technology in Anesthesia. San Diego, CA: Lippincott, Williams & Wilkins, 1999
15 Jasper HH. The 10/20 electrode system of the International Federation. EEG Clin Neurophysiol 1958; 10: 3715
16 Wolberg G. IEEE Computer Society Press. Los Alamitos, CA: Digital Image Warping, 1990
17 John ER, Prichep LS, Friedman J, Easton P. Neurometrics: computer assisted differential diagnosis of brain dysfunctions. Science 1988; 239: 1629[ISI][Medline]
18 John ER, Prichep LS. In: Niedermeyer E, Lopes Da Silva F, eds. EEG: Basic Principles, Clinical Applications and Related Fields. Amsterdam: Williams and Wilkins, 1993; 9891003
19 Gugino LD, Chabot RJ, Aglio LS, Maddi R, Gosnell J, Aranki S. QEEG and neuropsychological profiles of patients prior undergoing cardiopulmonary bypass surgical procedures. Clin EEG 1997; 28: 8797
20 Drover D, Lemmens H, Loyd G, et al. Patient state index (PSI): Sensitivity to changes in anesthetic state and optimization of delivery and recovery from anesthesia. Anesthesiology 2002; 97: 8289[CrossRef][ISI][Medline]
21 Gugino LD, Chabot RJ, Prichep LS, John ER, Formanek V, Aglio LS. Quantitative EEG changes associated with loss and return of consciousness in healthy adult volunteers anesthetized with propofol or sevoflurane. Br J Anaesth 2001; 87: 4218
22 John ER. A field theory of consciousness. Conscious Cognit 2001; 10: 184213[CrossRef][ISI][Medline]