Machine Learning Technology for EEG-Forecast of the Blood-Brain Barrier Leakage and the Activation of the Brain's Drainage System during Isoflurane Anesthesia

Abstract

Anesthesia enables the painless performance of complex surgical procedures. However, the effects of anesthesia on the brain may not be limited only by its duration. Also, anesthetic agents may cause long-lasting changes in the brain. There is growing evidence that anesthesia can disrupt the integrity of the blood-brain barrier (BBB), leading to neuroinflammation and neurotoxicity. However, there are no widely used methods for real-time BBB monitoring during surgery. The development of technologies for an express diagnosis of the opening of the BBB (OBBB) is a challenge for reducing post-surgical/anesthesia consequences. In this study on male rats, we demonstrate a successful application of machine learning technology, such as artificial neural networks (ANNs), to recognize the OBBB induced by isoflurane, which is widely used in surgery. The ANNs were trained on our previously presented data obtained on the sound-induced OBBB with an 85% testing accuracy. Using an optical and nonlinear analysis of the OBBB, we found that 1% isoflurane does not induce any changes in the BBB, while 4% isoflurane caused significant BBB leakage in all tested rats. Both 1% and 4% isoflurane stimulate the brain's drainage system (BDS) in a dose-related manner. We show that ANNs can recognize the OBBB induced by 4% isoflurane in 57% of rats and BDS activation induced by 1% isoflurane in 81% of rats. These results open new perspectives for the development of clinically significant bedside technologies for EEG-monitoring of OBBB and BDS.

Keywords: anesthesia; blood–brain barrier; brain’s drainage system; machine learning technology; spectral power analysis.

Figure 1

The scheme of experimental data processing and ANN structure.

Figure 2

The scheme of ANN training with two dependencies of accuracy on training epoch for training and testing datasets.

Figure 3

The effects of different doses of isoflurane on BBB permeability to EBDC: (a) photo of a real-time confocal microscopy of BBB integrity under EEG control in anesthetized rat; (b,c) representative images of the cerebral vessels with intact (IBBB) in rats treated with 1% isoflurane and OBBB in rats treated with 4% isoflurane; (d,e) representative images of whole-brain slices with high resolution of the region of interest illustrating IBBB (d) and OBBB for EBAC (e); (f) quantitative analysis of the EBAC level in the brain of rats from the control group without anesthesia and from the groups treated with 1% and 4% isoflurane, the Mann–Whitney–Wilcoxon test, n = 7 in each group, ***—p < 0.001; (g) photos illustrating typical white color of the brain of rats treated with 1% isoflurane (IBBB) and blue color of the brain of rats treated with 4% isoflurane (significant OBBB for EBAC).

Figure 4

The effects of different doses of isoflurane on BDS activity: (a) schematic illustration of design of experiments; (b–j) representative images of FITCD distribution in dorsal (b–d) and ventral (e–g) parts of the brain as well as in dcLNs (h–j) in the control group (b,e,h) and in the group treated with 1% (c,f,i) and 4% (d,g,j) isoflurane; (k,l) quantitative analysis of intensity of fluorescent signal from FITCD in dorsal and ventral parts of the brain as well as in dcLNs in the tested groups, respectively; Mann–Whitney–Wilcoxon test, n = 7 in each group, ***—p < 0.001.

Figure 5

Comparison of temporal implementations of ANN responses (Figure 1) for anesthetized animals when ANN input is normalized according to Equation (2) (a,b) and when it is normalized according to Equation (3) (c,d). White, green and pink backgrounds indicate three different conditions, including before anesthesia and under general 1% and lethal 4% doses of isoflurane. Orange lines correspond to the direct response of the ANN, black lines are the average response to remove outliers, blue and red dashed lines represent the binary response with a threshold (it is equal to the average, indicated by a dash-dotted line). In blue curve 1, the BDS is activated; in 0, it is not. In the red curve, 1—OBBB, 0—IBBB or death.

Figure 6

The statistics of nonzero ANN responses to BDS activation (purple box plots) and the OBBB (orange box plots). In this figure, the background color indicates the conditions, including before anesthesia (white) and under general 1% isoflurane anesthesia (green) and lethal 4% concentration of isoflurane administration (pink). In order to calculate the statistics, we used two normalizations of input signals based on Equation (2) (left purple box plots) and on Equation (3) (right orange box plots) for all animals with isoflurane anesthesia.

Figure 7

Dynamics of the average EEG power (in animal No. 5 under anesthesia), estimated in time-domain sliding windows in frequency ranges (a)—δ, (b)—θ. White, green and pink backgrounds indicate the time before anesthesia, 1% (general) and 4% (lethal) isoflurane anesthesia, respectively.

Figure 8

The average values of the spectral power density in the δ (panel (a)) and θ (panel (b)) frequency ranges of the animal EEG recordings, which were measured under the effects of different doses of isoflurane anesthesia. The height of the bars shows the sample mean power.

Figure 9

Changes in the normalized power spectral density induced by the administration of general 1% isoflurane anesthesia in the δ and θ frequency ranges. The height of the bars shows the sample mean change in power.