Measuring and monitoring the depth of anesthesia (DOA) is important for patient safety during clinical surgery. Electroencephalogram (EEG) is commercially available for DOA monitoring [1]. However, it is easily affected by the presence of opioids and the patient’s condition, so it is not predictable enough to deliver anesthetics [2, 3]. Also, it has been reported to overestimate or underestimate DOA in many clinical situations [4].

To overcome or supplement these limits, we tried to use cerebral hemodynamic change during anesthesia. Since anesthetic drugs change neuronal activity in the brain and affect cerebral metabolism, it can be hypothesized that cerebral metabolic change correlates with DOA [5, 6]. Cerebral oxygenation levels can be monitored to evaluate cerebral metabolism using near-infrared spectroscopy (NIRS), which is an attractive, noninvasive method for regional monitoring of cerebral oxygenation [7, 8].

We have previously shown that there is a correlation between cerebral hemodynamic changes and variation of isoflurane concentrations in an animal model by simultaneously monitoring cerebral oxygenation and local field potential. It suggested that NIRS signals can also be used as a powerful parameter to measure DOA [9].

However, since it was a preliminary study to investigate the relationship between cerebral hemodynamics and anesthetic levels, we needed pure and noise-free information. So, we measured NIRS signals from the brain parenchyma in a very invasive manner with the insertion of optodes through it, which is almost impossible to apply to patients subject to anesthesia. We therefore should take a noninvasive approach for clinical usage. When we get NIRS signals in a noninvasive manner, we also must be aware that an NIRS system may give inaccurate or misleading information. Cerebral oxygenation values that are measured by NIRS can be affected by several factors such as skull thickness and the cerebrospinal fluid layer (CSFL) [10, 11].

In this regard, if we check whether the NIRS signals measured in a less invasive model including the skull and CSFL are similar to the previous study used as a gold standard, it would be helpful to confirm the clinical applicability of NIRS for DOA measurements.

Therefore, the goal of this study is to investigate whether the minimally invasive rat model where optodes are just attached to the intact skull without drilling holes could give similar information of tracking anesthetic changes as the previous invasive model where the NIRS signals are obtained directly from the brain.

2.1. Animal Care and Use

All the procedures of these experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of the Gwangju Institute of Science and Technology (Gwangju, South Korea) (Approval number of IACUC: GIST-2019-060). A total of eight Sprague–Dawley rats (male, 350-500 g body weight, 14-18 weeks from birth) were used in this study. The rats were maintained on a day-night cycle (lights on at 9 AM and off at 6 PM) at a constant room temperature (24 ± 1 ℃) with free access to food and water. All experiments were performed during the daytime.

2.2. Minimally Invasive Rat Model

In this follow-up study, we used a minimally invasive rat model where optodes made with ceramic ferrule and 400 μm multimode optical fiber (FT400EMT, Ø400 µm Core, 0.39 NA; Thorlabs, NJ, USA) were attached to the intact skull. Custom-made optodes were attached to the intact frontal skull for recording cerebral hemodynamic changes in the brain without drilling holes (Fig. 1).

Figure 1. Schematic comparison of invasive and minimally invasive models.

Before the surgical attachment of optodes, the rats were anesthetized with a mixture of xylazine (7 mg∙kg⁻¹) and ketamine hydrochloride (100 mg∙kg⁻¹). The subsequent surgical procedure was the same as the previous invasive model [9]. The rats were fixed in a stereotaxic frame and consistently anesthetized with 1.5% (v/v) isoflurane. We made an incision of about 2.0–2.5 cm in the scalp along with midline and the pericranium was removed to expose the skull. The optodes were placed 2.5 mm in the anterior of the bregma and ±3.0 mm lateral from the midline and fixed in the exact position by applying super glue (Fig. 1). The attachment was completed by securing the optodes on top of the skull using dental cement. The rats were then moved to a recovery box.

2.3. Experimental Protocol

The experiments were performed at least 3 days after optode attachment to allow the rats to fully recover from postoperative complications such as surgical bleeding, pain, shock, and infection. The left optode attached to the left side of the skull was connected to a light source and the right one was combined with a charge-coupled device (CCD) spectrometer through optical patch cables. Before inducing isoflurane anesthesia, the NIRS signals were recorded from freely moving rats for 30 minutes (awake state). We set these awake state signals as the individual baseline reference. After that, the NIRS signals were collected at various concentrations of isoflurane with the same protocol as the previous study [9]. In brief, the isoflurane concentration started at 2.5% and decreased by 0.5% every 5 min to 1.0%. After that, the anesthesia was stopped, and we waited for the first spontaneous movement to be detected. After this movement, we collected the NIRS data for 30 minutes (recovery state). All NIRS signals were measured continuously in real time during the entire experimental procedure. A gas monitoring system (B40; GE Healthcare, CA, USA) was used to check the end-tidal CO2 and O2, inhalation oxygen percentage and isoflurane concentration during the experiments. At the end of each experiment, the rats fully recovered from anesthesia and rested.

2.4. Near-infrared Spectroscopy System

We used a continuous-wave NIRS system to measure changes in oxy-(OHb), deoxy-(RHb) and total hemoglobin (THb) concentrations. The NIRS system consists of a tungsten halogen light source (HL-2000-HP; Ocean Optics Inc., FL, USA) and a 16-bit CCD spectrometer (USB4000, Ocean Optics Inc.) as a detector. The light was illuminated to the left frontal lobe through a custom-made optode and diffuse reflectance was measured by the other optode from the right frontal lobe [9]. The separation between the light source and detector optode was 6 mm.

2.5. NIRS Signals Process

The NIRS signals were recorded on a 3 Hz frequency. The intensity values of five wavelengths (730, 750, 800, 830, and 850 nm) were used to calculate the relative changes in OHb, RHb, and THb concentrations ([OHb], [RHb] and [THb]). Assuming that OHb and RHb are the dominant absorbers in the NIR range, and that the scattering is also constant, we can estimate the changes in [OHb], [RHb] and [THb] by following the modified Beer-Lambert law [12, 13]. Since anesthesia-induced respiratory change could significantly affect changes in cerebral blood volume (CBV) [14], we calculated the ratio of reflectance intensity (RRI) between 730 and 850 nm wavelengths to reduce the confounding effect [9].

2.6. Statistical Analysis of NIRS Signals

For statistical analysis of the NIRS parameters, the last 1 minute at each of the four different isoflurane concentrations was selected. We also selected 5 minutes of data from the awake state (20-25 minutes from the onset of the awakening state) and the recovery state (7-12 minutes from the first spontaneous movement of the subject such as eye blinking and whisker movement). The mean values and standard deviation for each section were then calculated. After confirming the normality of the NIRS data using Shapiro-Wilk, a two-tailed paired t-test was performed to compare each section. When the null hypothesis of this test was rejected, we performed the Wilcoxon-signed rank test. If the hemodynamic parameters for each anesthetic concentration showed a significant linear relationship, we estimated a coefficient of determination between them through a simple linear regression analysis [9].

2.7. Monte Carlo Simulation

In the previous invasive model study, the separation between the light source and detector was 5 mm [9], while in this minimally invasive model it was 6 mm. This was the maximum separation we could get from our rat model. The average skull thickness of the rats used in the experiments was about 1 mm, and we tried to find out the differences between the two models in obtaining NIRS signals through Monte Carlo simulation (MCS). Using this simulation, we compared the detection depth and light distribution in the brain tissue according to the difference in the separation and the skull passage.

The MCS was performed by modifying the code of OMLC’s MCXYZ [15]. The positions of the light source and detector were set the same as in our actual experiment, and when photons arrived at the detector during the simulation, the energy weight of the photon was recorded. The photons were excluded from the recording if they didn’t reach the detector until their energy was exhausted. The sum of the energy weights of the photons that reached the detector was calculated as diffuse reflectance [16].

The total number of photons we used as a light source for each simulation was 10 million. The absorption coefficient of the brain tissue was 0.056 cm⁻¹ and scattering coefficient was 9.553 cm⁻¹ (at 800 nm). The absorption coefficient of the skull was 4.4 × 10⁻⁴ cm⁻¹ and scattering coefficient was 18.75 cm⁻¹ (at 800 nm). The detection depth was defined as 1/e point of the maximum fluence rate (φ, [W∙cm⁻² per W delivered]) value [16, 17].

At the beginning of induction of anesthesia with isoflurane, [RHb] decreased sharply compared to the awake state, while [OHb] and [THb] increased. After that, overall [RHb] increased with the reduction of the isoflurane concentration, while [OHb] and [THb] gradually decreased (Fig. 2).

Figure 2. Representative graphs showing changes in [OHb], [RHb] and [THb] before, during and after anesthesia in a minimally invasive model. Depending on the existence of abrupt change in near-infrared spectroscopy (NIRS) signals during 1.0% isoflurane, the results are divided into the (a) not awakened during anesthesia (NAA) group (n = 5) and (b) awakened during anesthesia (AA) group (n = 3).

However, some rats (three out of eight) showed abrupt changes in hemodynamic parameters during 1.0% isoflurane [Fig. 2(b)]. These results tended to be similar to those of previous invasive models [9].

Depending on the absence or presence of these sudden hemodynamic changes during the 1% isoflurane, we distinguished the results into two groups with either not awakened during anesthesia (NAA) (n = 5), characterized by the absence of an abrupt increase in [OHb] and [THb] and decrease in [RHb]; or awakened during anesthesia (AA) (n = 3), characterized by the presence of a sudden increase in [OHb] and [THb] and decrease in [RHb] [9].

The changes in [OHb], [RHb], and [THb] according to the isoflurane concentration are compared statistically in Fig. 3. Compared with the awake and recovery state, [OHb] and [THb] significantly increased (P < 0.05) during anesthetized, whereas [RHb] decreased (P < 0.05). In the NAA group, all hemodynamic changes ([OHb], [RHb], and [THb]) corresponded linearly with decreasing isoflurane concentration. On the other hand, as in the previous study, it generally changed in the AA group according to the isoflurane concentration, and then showed a reversal at 1% isoflurane.

Figure 3. Changes in [OHb], [THb], and [RHb] with varying isoflurane concentrations in the not awakened during anesthesia (NAA) (a)–(c) and awakened during anesthesia (AA), (d)–(f) groups in a minimally invasive model. The single asterisk indicates a P-value less than 0.05 compared to awake state and the double asterisk indicates a P-value less than 0.05 compared to each other. The results of simple linear regression analysis are shown on the columns for the NAA group. □, 25%~75%; 工, Min~Max; —, Median line; *, Mean.

Changes in the RRI were also compared for each state according to the anesthetic concentration (Fig. 4). The trend of total RRI change was generally similar to [OHb] change (Figs. 3 and 4). For both groups, isoflurane anesthesia induced significantly increasing RRI values. The mean RRI value during the awake and recovery states was 1.44 ± 0.01 in the NAA group and 1.38 ± 0.02 in the AA group (Fig. 4). On the other hands, the mean value of RRI right after anesthetization was 1.67 ± 0.01 in the NAA group and 1.56 ± 0.03 in the AA group, and both groups maintained an RRI value above of 1.52 during all anesthetic concentrations.

Figure 4. Changes in ratio of reflectance intensity (RRI) value depending on the isoflurane concentration in the (a) not awakened during anesthesia (NAA) and (b) awakened during anesthesia (AA) groups in a minimally invasive model. The single asterisk indicates P-value less than 0.05 compared to awake state and the double asterisk indicates P-value less than 0.05 compared to each other.

The MCS showed that the average fluence is lower in the minimally invasive model, but the detection depth is slightly deeper than the invasive model (Fig. 5). The detection depth in the invasive model was 2.05 mm, while it was 2.67 mm (excluding the mean skull thickness of 1 mm) in the minimally invasive model. This shows that the detection depth of the minimally invasive model is 23% deeper than the invasive model.

Figure 5. Monte Carlo simulation results showing light distribution in the brain (a), (b), (d), (e) and detection depth (c), (f) in the invasive model (a)–(c) and the minimally invasive model (d)–(f).

In this study, we showed that the minimally invasive model could track changes in anesthetic level, which is meaningful since NIRS signals should be measured in a noninvasive way for clinical application. Our minimally invasive model gives a similar result in terms of NIRS signal changes with varying levels of anesthesia compared to our previous study using a completely invasive method that can be used as a gold standard [9].

In the minimally invasive model, it was demonstrated that there was a positive correlation between [OHb] increase and isoflurane concentration, similar to the previous study (Fig. 3). [OHb] and [THb] sharply increased right after anesthesia by isoflurane, while [RHb] decreased. Considering that anesthetics act as depressants of the central nervous system (CNS), it seems plausible that the inhibition of cerebral metabolism is related to anesthetic level [18^–20]. Isoflurane increases cerebral blood flow (CBF) and cerebral blood volume (CBV) by direct vasodilatation, but inhibits the global cerebral metabolic rate [21, 22]. Therefore, increased CBV and decreased metabolic rate by isoflurane may lead to increases in [OHb] and [THb] and a decrease in [RHb] during increasing anesthesia [9]. As anesthesia decreased, it was shown that [OHb] and [THb] gradually decreased, while [RHb] increased. Indeed, when we used the minimally invasive model, the results were consistent with our speculation and previous study. Overall, among the changes in NIRS parameters according to anesthetic concentration, [OHb] has the highest linear correlation coefficient (Fig. 3). This is strong evidence that [OHb] accurately represents anesthetic concentration as an indicator as in previous studies [9].

Since anesthesia-induced respiratory depression could induce hypercapnia, leading to CBV change [14], we calculated an RRI between 730 nm and 850 nm wavelengths to deduct the confounding effect. The RRI corresponded to the changes in isoflurane concentration (Fig. 4). Overall, the RRI showed a trend similar to [OHb], showing the greatest sensitivity to anesthetic concentrations in both the NAA and AA groups (Figs. 3 and 4). This suggests that the RRI may be used as a powerful new DOA indicator regardless of respiratory status in a less or noninvasive model.

Interestingly, three out of eight subjects showed a sudden increase in [OHb], [THb] and RRI during 1% isoflurane, which was also observed in a previous study. We classified them into the AA group and distinguished them from those that did not (NAA group) (Fig. 2). It could be explained that light anesthesia almost reaching the awake level speeded up recovery from anesthesia and increased the possibility of neurovascular connectivity interrupted by anesthetics. In fact, an isoflurane concentration of 1% is about 0.67 MAC (minimal alveolar concentration of anesthetic that induces immobility in response to a noxious stimulus in 50% of subjects) in normal rats, which corresponds to the anesthetic level with a higher probability of awaking from anesthesia stochastically [23^–25].

We changed the separation between the NIRS source and detector from 5 mm to 6 mm in this study. This was a way to make a fair comparison by compensating for the optical path length and detection depth difference between completely invasive and minimally invasive models. However, on the frontal lobe part of our experimental model, the maximum distance separation we could get for attaching the optodes was 6 mm, making it difficult to obtain larger separation. The average skull thickness of the rats used in this study was about 1 mm, and we tried to find out the differences between the two models in obtaining NIRS signals through MCS.

The MCS showed that the detection depth of the minimally invasive model is 23% deeper than the previous invasive model. The reason for this is that the skull acts as an optical diffuser and makes the light irradiation range wider when light is transmitted from the light source to the brain. The initial optical fiber core size in the minimally invasive model is 400 µm, but after passing through the skull layer, the simulation shows an effect where the core size becomes about 2 mm. This means that although the overall photon concentration in the minimally invasive model is lower than in the invasive model, it satisfies the conditions of a suitable concentration and detection depth to obtain a meaningful signal. This suggests that the presence of the skull does not cause a great deal of difficulty in making NIRS measurements based on signals from brain tissue.

We expected that the NIRS signals would be absorbed by the skull and CSFL, causing the signals and anesthetic information to be weakened and lost. However, the tendency and degree of hemodynamic change with varying anesthetic concentrations showed a similar pattern to the previous invasive model study. This indicates that the signal quality of the cerebral oxygenation measurement that passed through the skull and CSFL in the minimally invasive model is nearly compatible with that of the signal obtained directly from the brain. This suggests that the minimally invasive model can get sufficient NIRS signals to check the difference in DOA.

We note some limitations in this study. In fact, a completely noninvasive NIRS measurement should be done on the scalp or skin. So far, we have not taken them into account for anesthesia depth measurements. There are also many blood vessels in the scalp or skin. This means that when we get NIRS signals noninvasively, it does not just include signals from the brain. However, considering that the skull is the hardest and thickest tissue around the brain, rendering it difficult to penetrate and easy to scatter, our results may suggest that a completely noninvasive model will have results similar to our model. Nonetheless, to measure DOA using NIRS in a noninvasive way, it is necessary to consider how to treat the scalp and the hemodynamic information contained therein.

Our results showed a positive linear correlation between RRI and isoflurane concentration, similar to the previous study. This indicates that the signal quality of the cerebral oxygenation measurement that passed through the skull and CSFL in the minimally invasive model is as good as the signal obtained directly from the brain. Therefore, we believe that NIRS signals measured in a less invasive way can be used as a potential indicator to monitor DOA.

The authors declare no conflicts of interest.

Data underlying the results presented in this paper are not publicly available at the time of publication, which may be obtained from the authors upon reasonable request.

SGER grant through the National Research Foundation of Korea (NRF-2015R1D1A1A02062382); Healthcare AI Convergence Research & Development Program through the National IT Industry Promotion Agency of Korea (NIPA) funded by the Ministry of Science and ICT (No. S1601-20-1016); and GIST Research Institute (GRI) IIBR grant funded by the GIST in 2022.