Abstract

Background: In medicine, general anesthesia during surgery involves the administration of pharmacological (hypnotic) agents and clinical monitoring via the analysis of the patient's nervous systems (loss of consciousness and reactivity) during stimulation. Unfortunately, this clinical monitoring is complicated by factors such as curarization, shock, and drugs that block cardiovascular responsiveness. Additionally, inadequate anesthesia due to over- or underdosing increases morbidity rates, such as hypotension and respiratory depression in the case of overdose, and memorization, movement, hypertension, tachycardia, laryngospasm, and bronchospasm in the case of underdosing. Several anesthesia monitoring tools have been introduced to address this issue, such as bispectral analysis (BIS), auditory evoked potential (AEP), q-CON, and entropy monitors; however, these instruments are complicated by accuracy, noise, artifacts, and their correlation with hypnotics.

Methods: Noninvasive anesthesia monitoring methods include lower esophageal sphincter, AEP, entropy, and spontaneous electroencephalography (EEG), which is the most commonly used. This method involves BIS of the collected EEG signals and correlates well with consciousness and sedation scores regardless of the anesthetic agents used. In this paper, we present a noninvasive method for monitoring the depth of anesthesia during surgery using the AEP and BIS methods. This study aimed to reduce artifacts, optimize the hypnotics/analgesics dosage, limit the effects of pharmacological use, and ensure a better quality recovery.

Results: We applied two techniques, BIS and event-related potential (ERP), following multiple stimuli to determine the best anesthesia monitoring approach. A comparative study of the EEG signals showed that measuring cortical responses by ERP provided more precise data in space and time regarding the sedation state of the patient and better monitoring of the hypnotic dose. The BIS method was simpler and easier to implement; however, only average and static values regarding the sleep rate could be obtained.

Conclusion: BIS and ERP appear suitable for monitoring sedation and hypnotic dosage during anesthesia, with the best reliability rates and speed with a latency of < 4 ms and an accuracy of 92%.

Introduction

Electroencephalography (EEG) is a technique that records brainstem activity through the cortex and may involve cortical recordings in some cases. EEG measures spontaneous activity in the presence or absence of a stimulus. For example, the auditory evoked response (AER) results from an auditory stimulus. The event-related potential (ERP) is used to describe the neural responses to a specific motor, cognitive, or sensory stimulus. In the 1970s, Schmidt & Al1 introduced the clinical use of electrocochleography to diagnose Meniere’s disease using a set of electrodes placed on the exposed brain surface. Then, in 1971, Jewett and Williston conducted the first systematic study of human auditory brainstem responses2. Later, Salters and Brakman in 1976 introduced brainstem electrical response audiometry (BEA) for detecting tumor acoustics3. Cortical responses demonstrate several shapes and types, such as cortical auditory evoked potential (CAEP), long-latency or late-latency response (LLR), and auditory late response (ALR). The auditory steady state response (ASSR) was introduced in 2001, which allows the simultaneous presentation of multiple carrier frequencies in both ears and was adopted in clinical practice to estimate the hearing threshold4.

Our work focuses on two clinical applications of ERP responses: the monitoring and evaluation of cochlear prostheses by electroacoustic stimulation and the monitoring and control of anesthesia by cortical responses to stimuli. This study relates to the second application. The principle of general anesthesia is to temporarily block perception, consciousness, and motor responsiveness to stimuli, and the anesthesiologist maintains the stability of the unconscious patient’s vital functions. However, these goals (sleep and no reactivity) are distinct and can be achieved independently. Sleep is linked to hypnotic agents, while loss of reactivity is highly dependent on analgesic agents. Thus, pharmacological administration must be optimized, monitored, and adapted on a case-by-case basis throughout anesthesia. Clinical monitoring involves analyzing neural responses to stimulation, such as brain reactions to surgical incisions5. At the beginning of the last century, Guedel introduced anesthesia monitoring by describing four stages of sleep achieved with ether-chloroform (Guedel, 1920). In 1929, EEG was invented by Berger to study brain electrical activities by measuring electrical potentials6. Since 1990, the statistical and spectral analysis of signals from electroencephalograms (EEGs) has allowed the practical and clinical development of several anesthesia monitors.

In 1993, Kissin proposed a conceptual study and framework of anesthesia centered on the pharmacological effects of anesthetic drugs. Analysis of frontal cortical EEG has been shown to be beneficial in exploring the loss of consciousness component, while monitoring the electrical activity of subcortical structures allows an estimation of the patient's responsiveness to a noxious stimulus7. Among the most developed analyses involving EEG were performed by Ballard in 1997, who decomposed the input signal into a frequency spectrum by Fourier series. However, the bispectral analysis of EEG has captured the attention of the anesthetic community. The EEG bispectral index (BIS) is a statistical index derived from the EEG by an algorithm using a large patient database. This index is predictive of the depth of hypnosis induced by propofol8.

The use of depth of anesthesia monitoring in practice was introduced in 2006 following recommendations from the American Society of Anesthesiologists (ASA), followed by a Cochrane Library meta-analysis in 2007 and formal recommendations from French Society of Anesthesia & Intensive Care Medicine experts in 20099. Two major studies were conducted by Dwong, Liu, and Punjasawa. The first included 1,380 patients from 11 outpatient surgery studies. A reduction in hypnotic drug consumption of 19% was recorded. The second included 4,056 patients from 20 studies7. However, despite the development of BIS monitors, several factors limit their accuracy, resulting in an average accuracy of 70% to 80%. To solve these problems, we developed a noninvasive BIS monitor to monitor and control the anesthesia process. First, we replaced analog circuitry with programmable and embedded algorithms implemented in the Raspberry Pi electronic board. This approach prevents inaccuracies in the device's circuitry and reduces the influence of artifacts. Second, we estimated the optimal hypnotic ratio to achieve the desired sleep level (e.g., the BIS ratio was estimated to be 40% in the sleep state), unlike other BIS monitoring machines that adjust the anesthetic dose according to BIC value changes.

× Figure 1 . Latency auditory evoked potentials signals . ( A ) Time domain waveform, ( B ) Frequency domain Figure 1 . Latency auditory evoked potentials signals . ( A ) Time domain waveform, ( B ) Frequency domain × Figure 2 . AEP data acquisition signal . Figure 2 . AEP data acquisition signal . Methods Auditory evoked potentials

Auditory evoked potentials are a physiological measure of the response of subcortical and cortical nerve centers to an auditory stimulus. This response can be divided into three successive series of positive and negative waves: the first is the early response (EAEP), which reflects brainstem activity; the second is the mean latency response (MAEP), which reflects the early cortical response; and the final is the late cortical response (LAEP). Only the auditory evoked potentials at medium latency can be used to measure the anesthesia depth. The latency and amplitude of Nb waves and Pa (Figure 1) are the main parameters usually analyzed. The average latency of PEA predicts loss of consciousness under propofol, according to the findings of Iselin-Chaves in 200010.

× Figure 3 . The data acquisition unit of the BIS monitors [10]. Figure 3 . The data acquisition unit of the BIS monitors [10]. × Figure 4 . Algorithm of the BIS index . Figure 4 . Algorithm of the BIS index .

Stimuli used to produce AEP consist of spikes (brief 100-microsecond square wave) or tone bursts (brief sinusoidal waveforms). The choice of stimulus type has minimal impact on anesthesia applications. The stimuli must be delivered at intensities above the hearing level. The scalp location with the largest AEP amplitude is usually the vertex (Cz). The steady-state response evoked by stimuli delivered at rates near 40 seconds (the 40 Hz auditory steady-state response (ASSR); Figure 2) has been used extensively to assess anesthetic effects. Transient responses are classified according to their latency as fast (6–10 ms), middle (10–50 ms), slow (50–250 ms), and late (over 250 ms) responses11. The AEP acquisition protocol with standard brain response and latency values is illustrated in Figure 212.

In 2001, Danmeter13 introduced the first AEP monitor to the biomedical market. The monitor detects the AEP index using an autoregressive model (AAI), which can be displayed on two scales: 0–100 or 0–60. For optimal anesthesia, the index value varies ​​between 15 and 2514. Musizza and Ribaric (2010)15 presents the AEP algorithm and AAI index.

BIS monitoring

EEG BIS is a spectral and statistical analysis method based on an algorithm developed using a large patient database. It measures the coherence of the EEG components and their frequency synchronization. The deeper the level of anesthesia is, the greater the consistency and synchronization will be. The EEG signals are collected from disposable self-adhesive electrodes placed on the areas hairless of the scalp, as illustrated in Figure 3. These electrodes are connected to a converter amplifier and a signal processing unit for filtering, feature extraction, and processing. BIS is easy to use and predictive of the hypnosis depth induced by propofol5.

To date, only one large-scale study has examined the use of BIS while taking propofol. Zhang et al. performed a randomized controlled trial including 5,228 patients with propofol and found that the risk of regaining consciousness was significantly reduced in the BIS-guided cohort (0.14%) than that in the cohort without BIS (0.65%). However, BIS has several limitations, the most striking of which is the wide range of values obtained for the same endpoint, making it very difficult to establish a valid individual threshold for loss of consciousness16, 17. Many artifacts, such as pacemakers, drugs, intra-abdominal irrigation, and the electromyogram, interfere with the signal18.

Figure 4 represents the principle of the algorithm of the BIS monitor. After the acquisition and reading of the EEG signal, it is then digitized, preprocessed, and filtered to remove artifacts from eye movements and power grid interference. The preprocessed data are used to calculate the parameter of the ratio β. This parameter is calculated as the ratio between 30–47 Hz and 11–20 Hz frequency bands. The parameter synchrony-fast-slow is calculated by bispectral analysis. It is defined as the ratio between the sum of all spectral peaks between 0.5 and 47 Hz and the sum of all spectral peaks over the 40–47 Hz interval19. Finally, all parameters are computed to deduce the BIS index according to Equation 1. The BIS algorithm is based on the same principle as phase lag entropy20, 21.

The bispectrum measures the correlation between signal phases at different Fourier frequencies. It is defined as an FFT-2D of third-order cumulates of a random process and is characterized by the bicoherence index (BIC), which varies between 0 and 100%22. According to the power density P(f), the correlation index (BIC) expression can be expressed as:

With

× Figure 5 . Standard positions of EEG Electrodes on the scalp [10],[20] . Figure 5 . Standard positions of EEG Electrodes on the scalp [10],[20] . × Figure 6 . Illustration of brain waves . Figure 6 . Illustration of brain waves . × Figure 7 . Evolution of EEG wavesin function of anesthesia . Figure 7 . Evolution of EEG wavesin function of anesthesia . × Figure 8 . Real time acquisition: EEG with 3 from 16 channels . Delta wave : deep sleep, The theta wave : light sleep, alpha wave : brain sleeping, Beta wave : first activity with opening eyes, Gamma wave : mental activity. Figure 8 . Real time acquisition: EEG with 3 from 16 channels . Delta wave : deep sleep, The theta wave : light sleep, alpha wave : brain sleeping, Beta wave : first activity with opening eyes, Gamma wave : mental activity. × Figure 9 . Experimental electrodes position and names on the scalp . Figure 9 . Experimental electrodes position and names on the scalp . × Figure 10 . Real time acquisition with five sample epochs of EEG data. Figure 10 . Real time acquisition with five sample epochs of EEG data. × Figure 11 . ERP Channel activities for electrodes 1 and 14 (scalp, spectrogram, spectrum) (For a moderate anesthesia state) . Figure 11 . ERP Channel activities for electrodes 1 and 14 (scalp, spectrogram, spectrum) (For a moderate anesthesia state) . × Figure 12 . ERP Channel activities for electrodes 1 and 14 (scalp, spectrogram, spectrum) (For a sleep state). Figure 12 . ERP Channel activities for electrodes 1 and 14 (scalp, spectrogram, spectrum) (For a sleep state). × Figure 13 . EEG Real time acquisition: Power spectrum envelopes and ERP activities of five most significant independent components. Figure 13 . EEG Real time acquisition: Power spectrum envelopes and ERP activities of five most significant independent components. × Figure 14 . Real time acquisition: EEG with 4 from 16 channels. Figure 14 . Real time acquisition: EEG with 4 from 16 channels. × Figure 15 . Electrode potential vs Latency. Figure 15 . Electrode potential vs Latency.

Table 1.

Latency in function of hypnotic’s percentage

Pourcentage of hypnotics ERP index Latency in ms 1 % 92 128 2.6 % 20 430 4% 12 1285 × Figure 16 . Sedation state vs. BIS index and hypnotics. Figure 16 . Sedation state vs. BIS index and hypnotics. Results Electrode standard positions

The EEG represents the time-varying trace of the electrical potential collected on the skull for different points of the scalp. EEG acquisition facilitates the visualization of brain process activity and our understanding of neurophysiologic phenomena. EEG signal acquisition was performed using electrodes placed in contact with the scalp at positions determined according to the international standard 10/2023. These locations are illustrated in Figure 5 and are explained in detail in the following section. Since EEG measures the brain's electrical activity, it can indicate the brain's state of sleep or sedation or any other activity. Unfortunately, the obtained and measured signals are very weak (a few microvolts) and require processing and amplification. Their amplitudes vary according to the patient’s age, sex, and condition24.

Extraction and classification of EEG waves

The analysis of EEG signals allows the extraction of five waves: delta, theta, alpha, beta, and gamma. These feature extractions help identify patients’ awake or sleep states during the night or when under anesthesia, as confirmed by Alferd Loomis25. Loomis found that the brain demonstrates electrical activity in neuronal regions. Variations in the fluctuations of these generated waves allow us to determine which wave dominates at a given moment. Each of the five waves is characterized by its frequency band and potential amplitude, as shown in figure 6. Each wakefulness state indicates a specific action potential corresponding to three parameters (awake state, light sleep state, and deep sleep state).

EEG recording can provide indicators of a person’s physical and mental state. For example, an EEG that shows high-amplitude alpha waves over the occipital area of the brain indicates that the person is relaxed and has their eyes closed. The alpha waves will disappear if they open their eyes26. Additionally, sleep researchers use recordings from entire nights to study and classify the different stages of sleep. The EEG waveforms of epileptic patients can also help localize seizure activity in the brain27, 28.

Every brain wave can be affected by the brain state as follows (Figure 6):

DELTA Brain wave: THETA Brain wave: ALPHA Brain wave: Predominant brain wave when a person is in the phase between wakefulness and sleep or relaxation. It disappears when the eyes are open. The alpha activity increases immediately when the eyes are closed. Its frequency ranges from 8 to 13 Hz. This wave reflects the posterior part of the head in the occipital region and the cortex and its peripheries.BETA Brain wave: Predominant brain wave when a person is awake or concentrating. Its frequency ranges from 13 to 40 Hz. It is located in the temporal, occipital, and frontal lobes of the brain.GAMMA Brain wave: A rapidly oscillating brain wave that is predominant when a person is in the intellectual thinking phase. It shows substantial brain activity during creative processing and problem-solving. The Gamma wave is the only wave present in all parts of the brain. Its frequency ranges from 40 to 80 Hz with a low amplitude of a few microvolts.

Figure 7 shows the effect of anesthesia and hypnotics on a person’s state and variation in the EEG brain waves. Note that the deeper the anesthesia is, the more brain waves demonstrate low frequencies with high amplitudes, and vice versa. The higher the analgesic and hypnotic level, the faster the B–A–T–D transition (Beta to Alpha to Theta to Delta)

Experimental

EEG is measured by using small electrodes attached to the scalp's surface. The number of electrodes can vary from 1 to 32 in the 10/20 scalp system or from 1 to 256 in the 10/10 system. The electrodes are placed at predefined positions according to the international 10/20 system or its variants. The weak electrical activity detected by the electrodes varies from 5 to 100 µV, and the frequency range of interest is between 1 and 40 Hz.

We used the international 10/20 system with 29 electrodes cho