USEFULNESS OF BRAIN SIGNALS FOR THE DETECTION OF

LOSS OF CONSCIOUSNESS IN ANESTHESIA

Overview of the Problem and Results from a Clinical Study

Carmen González

Instituto de Engenharia Mecânica, University of Porto, Rua Roberto Frias, Porto, Portugal

Marina Mendes, Pedro Amorim

Hospital Geral de Santo António, Porto, Portugal

Joaquim Mendes

Instituto de Engenharia Mecânica, University of Porto, Rua Roberto Frias, Porto, Portugal

Catarina S. Nunes

Division of Engineering, King’s College of London, Strand, London, United Kingdom

Keywords: General anesthesia, Loss of Consciousness detection, Bispectral index.

Abstract: Loss of consciousness (LOC) detection is essential for better anesthesia guidance. Clinical signs and brain

monitoring are currently used in operating rooms to assess the state of consciousness. However, a patient-

independent, accurate and reliable indicator of LOC is not currently available. We studied 69 patients

undergoing general anesthesia, investigating a possible relationship between loss of consciousness and BIS

and EMG signals registered during induction. Neither BIS and EMG values at LOC, nor their abrupt fall

proved to be good indicators of loss of consciousness. Further work needs to be done in order to reliably

detect loss of consciousness.

1 INTRODUCTION

Hypnosis, analgesia and muscular blockade are the

three components of general anesthesia. From all

three, hypnosis is the most important; it is a

pharmacologically induced sleep state, a reversible

state of unconsciousness (Bonhomme and Hans,

2004).

Loss of consciousness (LOC) detection has an

important role at induction of anesthesia. Identifying

the precise moment when it occurs will determine

the hypnotic dosage required for each individual

patient and provide important information to help

avoiding awareness episodes and overdoses.

Furthermore, detecting LOC is the first step in the

development of fully automated anesthesia delivery

systems based on the conscious state of the patients.

In this work, we investigated how the Bispectral

index (BIS), the most widely used anaesthesia

monitor of consciousness, reflects loss of

consciousness during induction of anesthesia. It has

been shown that BIS values at the moment of LOC

can vary widely (Tesniere et al., 2003). We

examined not only BIS values at LOC, but also the

abrupt fall that occurs in the BIS and

electromyography (EMG) signals during the

transition from the conscious state (awake) to

unconsciousness (“sleep”) at induction of anesthesia.

This paper is organized as follows: in section 2 a

short review of LOC detection methods is presented;

section 3 includes our clinical study and conclusions

are pointed out in section 4.

311

González C., Mendes M., Amorim P., Mendes J. and Nunes C. (2009).

USEFULNESS OF BRAIN SIGNALS FOR THE DETECTION OF LOSS OF CONSCIOUSNESS IN ANESTHESIA - Overview of the Problem and

Results from a Clinical Study.

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing, pages 311-315

DOI: 10.5220/0001546403110315

 SciTePress

2 SHORT REVIEW

Since the early days of anesthesia, several clinical

signs have been used for LOC assessment, such as

pupil diameter, sweat or tears production, heart rate

and muscular tone (Strickland and Drummond,

2001). Nowadays, in the clinical setting, LOC is

usually identified as the moment when the patient

looses the eyelash reflex or stops responding to

specific commands (proper name calling or tapping

in the forehead) or stimuli (pressing a button every

time they see a red light), which are very imprecise

and dependent on patient’s collaboration. These

methods have several disadvantages: assessment of

lack of response to name calling and mechanical

stimulus require the patient to be frequently

stimulated, so anesthesia induction cannot be smooth

and quiet; also any method based on a voluntary act

by the patient often fails because patients loose

critical judgement before losing consciousness.

In the last two decades, brain monitoring

techniques have been developed to assess the

hypnotic state of the patients. They were designed to

reflect real-time electrical activity of brain cells but

were not specifically created for LOC detection. In

clinical practice, they are used in combination with

clinical signs to provide as much information as

possible about the level of consciousness of the

patients.

Among all the developed indices, the Bispectral

Index (BIS) (Kelley, 2003) is the most well-known

and broadly used. It processes electrical activity

(electroencephalographic signal, EEG) detected in

the forehead by means of electrodes and shows a

value, from 0 to 100, which indicates the level of

hypnosis: 0 is associated to absence of electrical

activity and 100 corresponds to the awake state. The

BIS system uses a proprietary algorithm, based on

bispectral analysis, which performs amplification,

digitalisation and signal processing of EEG in order

to obtain the above mentioned index.

The BIS signal and the electromyopraphy of the

frontal muscle, produced spontaneously and also

acquired with the BIS unit, were the subject of a

clinical study presented in this paper, where their

relationship with the exact moment of LOC was also

analyzed.

3 CLINICAL STUDY

3.1 Patients and Methods

Adult patients scheduled for neurosurgical

procedures were studied during induction of general

anesthesia. Patients were monitored according to the

American Society of Anesthesiologists (ASA)

standards. BIS was also monitored in all patients. A

XP-Quattro sensor was applied to the forehead and

ASPECT A2000 or VISTA monitors were used. BIS

and EMG were recorded every 5 seconds using

Rugloop software. Anesthesia was induced with

Remifentanil (an opioid analgesic) and Propofol (an

hypnotic) administered intra-venously by Target

Controlled Infusion (TCI) using Rugloop software

and Asena pumps. During induction, LOC was

determined as the moment when patients failed to

open their eyes following name calling and a tap on

the forehead at every 15 seconds. The moment of

LOC was recorded for every patient.

BIS and EMG values collected during induction

were analyzed, as well as the moment of LOC. In

order to find another possible indicator of

consciousness on the BIS and EMG signals, the

moment when a significant fall in BIS and EMG

values occurred was assessed. Both signals were

processed using Matlab®. First, a moving average

filter of 11 samples was applied to the acquired

signals aiming to smooth them, and afterwards its

first derivative was analyzed. The minimum

derivative value corresponds to the abrupt fall

observed in the signals.

Once the falls were detected, we defined two

new variables: the difference of time between LOC

clinical detection and the moment when the falls in

BIS and EMG occurred (δ

BIS

and δ

EMG

respectively).

3.2 Results

We studied 69 patients of both genders; age was

51.5±14.7 years, height was 164.1±9.9 cm and

weight was 69.7±13.3 kg. The patients underwent

cranial or spinal surgeries.

BIS at LOC was 69,52±15,86 and EMG at LOC

was 42,04±7,60 dB (table 1). Typical tracings for

BIS and EMG trends during induction are shown in

figure 1.

BIOSIGNALS 2009 - International Conference on Bio-inspired Systems and Signal Processing

312

Figure 1: Example of BIS and EMG trends during

induction. The moment of LOC (dashed line) and of the

abrupt fall (straight line) are also indicated.

Table 1: Obtained values for EMG and BIS at LOC, and

for the difference of time between the moment of LOC

and the detected falls.

Mean Standard Deviation

BIS 69.52 15.86

EMG (dB) 42.04 7.60

BIS

(s)

-0.79 34.45

EMG

(s)

4.47 25.9

The difference of time between LOC clinical

detection and the moment when the falls occurred

was -0,79±34,45 seconds for BIS and 4,47±25,9

seconds for EMG (see table1). The fall detection

failed in 4 cases with BIS (5,7%) and 12 with EMG

(17,39%) because no abrupt fall clearly associated

with LOC existed in those cases (figure 2). The

statistical distribution of both variables is shown in

figure 3.

No significant correlation was found between

the new variables (δ

BIS

and δ

EMG

) and patient data,

such as age, sex, body mass index and the amount of

hypnotic administered until LOC. However, a

significant relationship was found between the

values of BIS at LOC and the difference of time

between LOC and BIS fall’s detection (figure 4).

Figure 2: BIS trend during induction and moment of LOC

in patient number 2: an abrupt fall in BIS cannot be clearly

identified.

Using a regression linear model, as shown in

figure 4, we confirmed the existence of a

relationship, with a P significance value lower than

-8

. No similar conclusion was obtained for EMG

signals and their abrupt fall.

3.3 Discussion

Some authors have previously reported BIS values at

the moment of LOC: BIS values in the transition

from consciousness to unconsciousness were: 78±14

(mean±standard deviation) (Tesniere, Billard, 2003);

82±10 (Kreuer et al., 2004) ; 73±18 (White et al.,

2006). Not only the BIS values at LOC were quite

different, but also a high standard deviation was

observed. Our results also show a high standard

deviation. This can be due to an existing inter-

subject variability or to a lack of ability of BIS to

reflect LOC. We investigated the possibility that

patients’ characteristics could influence BIS or EMG

performance to detect LOC, however, the fact none

of the demographic patient data showed significant

relationship with values at LOC suggests that BIS or

EMG do not perform well in this setting. Therefore,

it seems that there is not a BIS or EMG value

associated to loss of consciousness.

The values obtained for the difference of time

between the moment of LOC and the abrupt fall in

BIS and EMG signals also show a very high

standard deviation. Again, this can be due to inter-

subject variability. Furthermore, precise fall

detection in EMG signals failed in 17,39% of the

cases. Our finding that high values of BIS at LOC

correspond to negative δ values ( LOC detection

occurs before the abrupt fall) and low values are

associated to positive δ values (LOC detection

occurs after the abrupt fall), although interesting, has

no clinical application for the real time detection of

LOC using data derived from the BIS signal.

USEFULNESS OF BRAIN SIGNALS FOR THE DETECTION OF LOSS OF CONSCIOUSNESS IN ANESTHESIA -

Overview of the Problem and Results from a Clinical Study

313

Figure 3: Boxplot of the difference of time between LOC

and the BIS fall, and LOC and the EMG fall. For each

variable, values for maximum, third quartile, first quartile

and minimum are shown.

4 CONCLUSIONS

The BIS index provides useful information about

hypnotic state, but not about LOC detection;

transition from consciousness to unconsciousness

cannot be identified by the index value. We can

detect the moment when an abrupt fall in BIS (and

EMG) in most patients occurs, but it is usually

different from the moment of LOC. Our results

suggest that neither BIS nor EMG derivatives can be

used to determine the moment of LOC with

precision. Clinical signs are not ideal for LOC

detection. Clinical assessment of LOC is not

accurate, consistent or precise: different criteria turn

up into different LOC detections, and clinical signs

are not continuously checked.

We can conclude from this study that BIS index

does not reflect LOC: the index value does not have

any direct relation with the transition from

consciousness to unconsciousness, and the BIS fall

can only be identified using future values, so it is not

a useful tool in real-time scenarios and has no

clinical use to guide anesthesia induction.

No existing technique can detect the moment of

LOC with precision and without interfering with a

smooth induction or independently of patient’s

cooperation. We can approximately know when

LOC occurs, but small errors in its identification can

turn into over or under doses that can have

deleterious effects for patients undergoing general

anesthesia.

Figure 4: Linear regression model for the difference of the

time between LOC and BIS fall and BIS values at LOC

(significance P < 10

-8

More precision in LOC detection is also an

important step for automated anesthesia delivery

systems. The closed-loop systems developed until

now are controlled by the index value (e.g. BIS), and

the majority of them can only be used during

anesthesia maintenance (Liu et al., 2007). A proper

LOC detection would allow the development of

closed-loop systems able to deal with inter-subject

variability and to be used also during induction.

Further work should be done in this area in

order to find a reliable, objective and rigorous

method of LOC detection for both improving patient

safety and new automated anesthesia delivery

systems.

ACKNOWLEDGEMENTS

This work was supported by Fundação para a

Ciência e Tecnologia (FCT), as being part of the

project PTDC/EEA-ACR/75454/2006.

REFERENCES

Bonhomme, V. and Hans, P., Monitoring depth of

anaesthesia: is it worth the effort? European Journal

of Anaesthesiology, 2004. 21(6): p. 423-8.

Tesniere, A., Billard, V., Hartl, D., and Bourgain, J.-L.,

BIS™ and Spectral Entropy™ To Predict Loss of

Consciousness during Induction and Recovery from

Anesthesia with Propofol and Remifentanil. American

Society of Anesthesiologists, 2003. A329.

Strickland, T.L. and Drummond, G.B., Comparison of

pattern breathing with other measures of induction of

anaesthesia, using propofol, methohexital and

sevoflurane British Journal of Anaesthesia, 2001.

86(5): p. 639-44.

BIOSIGNALS 2009 - International Conference on Bio-inspired Systems and Signal Processing

314

Kelley, S.D., Monitoring level of consciousness during

anesthesia and sedation. A clinician's guide to the

Bispectral Index. Norwood (MA) 2003, Aspect

Medical Systems.

Kreuer, S., Bruhn, J., Larsen, R., Bialas, P., and Wilhelm,

W., Comparability of Narcotrend index and bispectral

index during propofol anaesthesia. British Journal of

Anaesthesia, 2004. 93(2): p. 235-40.

White, P.F., Tang, J., Romero, G.F., Wender, R.H.,

Naruse, R., Sloninsky, A., and Kariger, R., A

comparison of state and response entropy versus

bispectral index values during the perioperative

period. Anesthesia & Analgesia, 2006. 102(1): p. 160-

Liu, N., Chazot, T., Trillat, B., Dumont, G.A., and

Fischler, M., Titration automatisee du propofol guidee

par l'index bispectral. Annales Francaises

d'Anesthesie et de Reanimation, 2007. 26(10): p. 850-

USEFULNESS OF BRAIN SIGNALS FOR THE DETECTION OF LOSS OF CONSCIOUSNESS IN ANESTHESIA -

Overview of the Problem and Results from a Clinical Study

315