1 Introduction

Cerebrovascular autoregulation (CA) is a complex process and is considered one of the most important central nervous system auto-protective mechanisms against secondary brain injury after traumatic brain injury (TBI) [1]. CA is described as the ability of vessels to modulate their tone in response to changes in cerebral perfusion pressure and in so doing, maintain constant levels of cerebral blood flow (CBF) to match metabolic demand [2, 3]. It is often impaired after injury [4,5,6] and has been shown to be a significant predictor of outcome in patients with various acute neurological diseases, including severe TBI-related and ischemic stroke [7,8,9]. In addition, continuous monitoring of autoregulation is envisioned to be potentially beneficial as a means to personalize optimization of CPP on a patient-by-patient basis [10, 11]. The Brain Trauma Foundation currently does not recommend CA monitoring but it does recommend CPP-monitoring during the management of severe TBI (level IIb) [12].

Direct and indirect assessment of CA has proven to be difficult. In addition to standard physical examination and clinical judgment, approaches and methods such as transcranial-Doppler [13], brain tissue oxygenation [14], near-infrared spectroscopy measured hemoglobin saturation [15], and Laser-Doppler flowmetry of CBF [16] have proposed to evaluate CA non-invasively and continuously however, with mixed results [11]. Such methods hold promise but have several challenges due to the intermittent or invasive nature of the measure, inability to fully measure and track meaningful changes, or the need for a high level of operator experience. There is currently a lack of technology available in far-forward echelons of care for CA monitoring [11].

A newer approach to assess and monitor CA is the computation of the pressure reactivity index (PRx), which simultaneously tracks dynamic changes in mean arterial blood pressure (MAP) and intracranial pressure (ICP) as an indicator of intact or impaired CA after cerebral injury [11]. PRx is calculated as the moving Pearson correlation coefficient between MAP and ICP with positive values indicating impaired autoregulation (pressure-passive behavior of the arterial walls) and negative values indicating intact autoregulation (vascular bed with active vasomotor responses) [17,18,19]. PRx measurements provide an independent predictor of outcomes after TBI [11, 17,18,19] and appear reliable for the evaluation of CA. PRx assessment of CA however, is currently limited in its use. The invasive nature of the measurements, the complexity of signal processing, and inherent signal noise have potentially constrained its application to mainly academic institutions that are capable of adequate signal acquisition and have specialized processing software [17, 20]. For these reasons, its application at early stages of care is problematic.

Here we investigate a novel methodology to assess CA utilizing trans-ocular brain impedance (TOBI). Impedance (dz) is coupled with MAP to continuously calculate an index (DZx) capable of monitoring CA without the need for ICP measurement. TOBI uses alternating current delivered externally through closed eyelids that is capable of penetrating deep into the brain to measure changes in cerebral blood volume (CBV). We have shown previously that respiratory-induced brain bioimpedance changes (dz) track changes in cerebral blood volume (CBV) CBF, ICP, and CPP with high precision [21]. In this study, we use changes in dz to produce the novel DZx parameter and compare its performance with the more established PRx as a metric for the assessment of CA. The inherent relationship between ICP and CBV suggests DZx may be a potential technique to monitor and assess changes in CA similar to PRx. We hypothesized that DZx will track changes in PRx and assess CA with high precision.

2 Materials and methods

2.1 Anesthesia and surgical instrumentation

Twelve male Yorkshire mix swine with a mean (standard deviation) weight of 44(4.2) kg were initially sedated using ketamine/xylazine (20/2 mg kg−1). Anesthesia was induced using 2.5% isoflurane in a fraction of inspired oxygen (FiO2) of 100% and maintained at 1.5–2.5% isoflurane and an FiO2 of 30–40%. Animals were intubated using a standard cuffed 7.5 mm endotracheal tube. Core temperature was maintained at 37–38 °C using a warming blanket (Blanketrol®, Cincinnati Sub-Zero, Cincinnati, OH) for the duration of all procedures. Mechanical ventilation was set using volume control with tidal volume at 7–8 mL kg−1 (Draeger FabiusGS Premium. Draeger Inc. Telford, PA). Respiratory rate was adjusted to titrate end-tidal CO2 (PetCO2) to 35–45 mmHg (Biopac Data Acquisition System. Biopac Inc. Goleta, CA).

The right common femoral artery was cannulated for continuous arterial pressure measurements. The left and right external jugular veins were cannulated with 9Fr introducers (Arrow International Inc., Reading, PA) using an ultrasound guided Seldinger technique (Mindray M9. Mindray Corp. Mahwah, NJ). The right introducer was used for placement of a triple lumen central venous catheter (Arrow International Inc., Reading, PA) for central venous pressure measurements (CVP), drug delivery, and blood sampling, and the left for administration of fluids. At the conclusion of vessel cannulations, animals were turned to a prone position for the remainder of the experiment. Two, 2–3 mm burr holes were drilled into the left skull through a small circular skin window. A laser Doppler flow probe TSD144 (Biopac Systems Inc. Goleta, CA) and a 2F solid state pressure catheter SPR-320 Micro-Tip (Millar, Houston, TX) were placed directly into the brain parenchyma (cerebral cortex) through the burr holes for CBF and ICP monitoring respectively.

2.2 Trans-ocular bioimpedance setup and processing

Two disposable silver-silver chloride electrocardiographic (ECG) electrodes, (Positrace RTL, ConMed Corp. Utica, NY) were placed on the closed eyelids and secured by an adhesive tape. The electrodes were connected to a bioimpedance amplifier EBI100C (Biopac Systems Inc.) using shielded leads. A low electrical alternating current (0.1–1 mA at 50 kHz) was introduced through the electrodes, and the potential between the two electrodes was measured. The TOBI signal was passed through a series of digital filters to remove baseline drift and high-frequency noise, then movement artifacts were detected via adaptive thresholding and removed. The respiratory and cardiac components were then isolated via filtering. The amplitude of these components is quantified using a root-mean-square envelope in a frequency-independent manner, meaning that the resulting metrics are largely independent of respiratory and heart rate (Fig. 1). PRx is calculated as the moving Pearson correlation coefficient between MAP and ICP over a five-minute window with a step size of one second, while the bioimpedance index DZx was calculated as the moving Pearson correlation coefficient between MAP and dz over a five-minute window with a step size of one second. However, it is notable that DZx and PRx often move opposite one another due to the inverse relationship between dz and ICP caused by changes in cerebral blood volume. Being calculated as a Pearson correlation coefficient, both PRx and DZx may have values between − 1 and 1 [17].

Fig. 1
figure 1

Respiratory-induced changes in trans-ocular brain impedance. The figure shows four breaths. dz is calculated as respiratory amplitude in the impedance signal

Full size image

2.3 Hemodynamic perturbation (vasopressor challenge)

Following instrumentation and baseline sample collection, animals were subjected to maneuvers to manipulate CBF, MAP, ICP and disrupt CA. Norepinephrine solution (4 µg mL−1) was prepared by mixing 2 mg norepinephrine in 500 mL of 5% dextrose in 0.9% NaCl saline solution and was administered through an Alaris IVAC 7130B infusion pump (BD company, Franklin Lakes, NJ). The infusion rate was slowly titrated upward to increase blood pressure and reach a MAP of 150 mmHg or greater as well as to produce an increase in ICP which frequently was also associated with a loss of CA. MAP was maintained for 15 to 30 min followed by the titrated reduction or cessation of the infusion and allowing the animal’s MAP to return to near baseline. The vasopressor challenge was performed twice. Impedance data as well as all other hemodynamic data described above were recorded continuously using Biopac Data Acquisition System MP150 (Biopac Systems Inc.). MAP, CVP, ICP, CBF, and TOBI were all collected continuously throughout the duration of the experiment using a Biopac MP160 and Acknowledge data acquisition software (Biopac Inc. Goleta, CA).

2.4 Statistical analysis

Continuous data is presented either as mean and standard deviation or as mean and standard error (figures). Pearson correlation was used to test the strength and direction of the relationship between dz and hemodynamic metrics (ICP, CPP, and CBF). Percentage change from baseline for dz and hemodynamic metrics were calculated during the vasopressor challenge. Receiver operator curve (ROC) and area under the curve (AUC) were used to assess the predictive performance of DZx at different thresholds of PRx (> 0, > 0.10, > 0.20, and > 0.30). DeLong Methedology [22] was used for the calculation of the standard error of the AUCs. Binomial exact confidence intervals were calculated for the AUCs. Average value of sensitivity for all possible values of specificity, as well as the average value of specificity for all possible values of sensitivity were calculated. Optimal criterion values of DZx corresponding to the best fit of the coordinates of the ROC curves were calculated. Positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR), as well as negative likelihood ratio (NLR) for DZx were calculated. Statistical significance was set at α = 0.05, and statistical analyses were performed using MedCalc 19.1.3, GraphPad Prism 8, and MATLAB R2019a.

3 Results

During the vasopressor challenge, MAP, ICP, CPP, and CBF had a mean percent increase (SD) from baseline of 64(22.2)%, 29(23.2)%, 70(25)%, and 37(72.6)% respectively, all of which were statistically significant and considered clinically significant with changes above 15%. Of note is the considerable increase in MAP and CPP over ICP, which can be contributed to the direct systemic effect of norepinephrine on the vasculature whilst the moderate increase in ICP is more related to the increase in cerebral blood volume especially, when these animals lose autoregulation [23, 24]. There was a significant decrease in dz by 31(15.6) % (Fig. 2) and significant negative correlation between dz and ICP, CPP, and CBF, with across animals’ average (SD) correlation coefficients of − 0.64(0.15), − 0.57(0.36), and − 0.49(0.21) (p < 0.0001) respectively indicating that impedance decreased with increases in CBV, which is consistent with the conductive properties of blood when compared to other bodily tissues. In addition, there was a significant correlation between CPP and CBF (r = 0.74, p < 0.0001). Table 1 lists descriptive statistics (minimum, maximum, range as well as 95% confidence intervals) of the mean percent change. Figure 3 provides an example of the longitudinal changes in MAP, ICP, CBF, and dz during the vasopressor challenge. The figure demonstrates increases in MAP, ICP and CBF simultaneous with a decrease in impedance amplitude (dz), as a reflection of increased CBV. Pooled DZx from all animals was highly and inversely correlated to PRx (r = − 0.47, p < 0.0001).

Fig. 2
figure 2

Percentage change from baseline of MAP, ICP, CPP, dz, and CBF during vasopressors challenge. MAP mean arterial pressure, ICP intracranial pressure, CPP cerebral perfusion pressure, CBF cerebral blood flow, dz impedance respiratory amplitude. Data is presented as means and standard errors

Full size image
Table 1 Percent change during vasopressor challenge
Full size table
Fig. 3
figure 3

A sample longitudinal view of hemodynamics, and brain impedance during the vasopressor challenge. MAP mean arterial pressure, ICP intracranial pressure, CBF cerebral blood flow. The solid lines in the upper three tracing represent the mean change, while the line in the lower tracing represents changes in impedance respiratory amplitude

Full size image

ROC tests demonstrated the predictive performance of DZx when compared to PRx at different PRx thresholds between 0 and 0.3 yielding a high predictive power of DZx with high AUC, sensitivity and specificity. Table 2 lists DZx’s AUC, sensitivity, specificity, PPV, NPV, PLR, as well as NLR at their associated PRx thresholds. Figure 4a shows a representative ROC plot at a PRx threshold of 0.10. In addition, the ROC test yielded the highest DZx performance at a criterion between − 0.016 and − 0.08 (Table 2). Figure 4b shows the best DZx criterion and its associated sensitivity and specificity at the same PRx threshold of 0.10.

Table 2 Receiver operator curve and area under the curve of DZx
Full size table
Fig. 4
figure 4

a ROC for DZx at PRx threshold of 10. b Associated criterion, sensitivity and specificity of DZx at a PRx Threshold of 0.10

Full size image

4 Discussion

Our study has shown that utilizing a transocular pathway to measure changes in brain impedance might constitute an effective alternative to PRx to assess CA and CPP that does not require invasive ICP monitoring. Our novel impedance metrics of dz by itself, or as an index (DZx), were shown to be capable of tracking changes in PRx, CPP, and CBF with high fidelity and as a result provide assessment of CA.

In the vast majority of clinical settings, early changes in cerebral hemodynamics are difficult to assess, due to limited ability to visualize or detect changes in intracranial pressure or blood flow. Thus, detection and assessment of TBI severity and associated secondary injuries remains a challenge [25]. Monitoring modalities that enable assessment of ICP and CA might be a valuable tool to target optimized CPP and therefore maintain optimal CBF, as means to prevent and mitigate secondary brain injury [26]. PRx has been shown to be a promising methodology to assess CA, and predict severity and outcome post TBI [27]. However, because PRx requires ICP and intra-arterial blood pressure monitoring it seems only suitable in structured and controlled environments existing at higher echelons of care [28]. TOBI may allow early assessment of changes in CBV and cerebral autoregulation without the need for invasive ICP monitoring (but still with the need for invasive MAP monitoring). We are currently exploring the potential of using continuous noninvasive MAP monitoring to make the system completely noninvasive.

Bioimpedance has been used recently for monitoring and measurements related to cardiovascular and respiratory systems such as cardiac output, lung water content, and respiratory effort [29]. It is a measure of biological tissue’s passive electricity and resistance to an induced current or voltage [30, 31]. Blood is a good conductor of electrical current and blood volume has a distinct effect on impedance, which allows impedance measurements to detect changes in blood volume in an area of interest. In addition, bioimpedance is indirectly affected by respiration due to the change in blood volume produced during inspiration and expiration (Fig. 1). In this study, we measured brain bioimpedance through an ocular window by placing two electrodes on the animal’s closed eyelids using a bipolar arrangement as outlined by previous work [21]. As the low electrical current is applied through the ocular pathway, the impedance respiratory-induced differences primarily reflect the blood volume inside the skull. We have previously demonstrated that this ocular pathway and technique selectively targets the brain and overcomes issues with transcranial (scalp) impedance [21].

In this study, we investigated the loss of CA by the use a vasopressor challenge as a means to rapidly elevate CPP to a level above the compensatory range. We chose this model to create a reproducible and unequivocal state of loss of CA indicated by a rise in ICP during elevation of MAP and based on the knowledge that sudden elevation of CPP is known to be associated with aggravation of the secondary injury post TBI [32]. We also chose vasopressor as a reversible challenge to modulate changes in MAP, ICP, and CBF which are the metrics used for the calculation of PRx. Such modulation creates a wide range of these metrics that enables the determination of a DZx threshold where we can define autoregulation as intact or impaired and compare it to the more established PRx parameter. The establishment of such threshold will begin to enable further study of the use of such index during neurological patient management where perturbations are less extreme. In this study as with our previous study, dz once again correlated highly with ICP, CPP, and CBF changes. This correlation was expected as dz changes are sensitive to changes in blood volume during the vasopressor challenge, decreasing while blood volume increases and vice versa.

The receiver operator test revealed that DZx, as a non-invasive impedance metric, has a high predictive power with area under the curve in the upper 80s when compared to PRx. Although PPV might seem modest in this study, the combination of the high values of the sensitivity, specificity, and NPV are an indication of the robustness of the impedance metric. Furthermore, the combination of high PLR as well as low NLR indicate that DZx is capable of assessing the presence or absence of CA with high probability. TOBI and the presented indices of impedance represent a less invasive technique that may be performed in the early care of neurologic injury to assess CA and ICP. This technique may prove to have potential to enable early detection of severe TBI, guide triage and manage treatment, as well as improve outcomes by alleviating secondary brain injury associated with the loss of CA. It is envisioned to be applied as a rapid point of care diagnostic indicator of severity of TBI and other cerebral injuries allowing for earlier intervention in more far forward echelons of care.

4.1 Limitation

This study has several limitations. We have used male animals only in this investigation. Future studies will include both males and females to investigate the role of gender and its effect on impedance and autoregulation. This study tested the performance of bioimpedance and its indices in a large animal model of hemodynamic perturbations without TBI and while animals were under general anesthesia. More testing is needed in animals with TBI that also have perturbed physiology such as hypertension and hypotension to assess the ability of TOBI to track CA. We also measured CBF using LDF which provides only a relative measure of CBF in arbitrary units and is subject to noise. Furthermore, we used PRx as a gold standard metric for comparison. However, PRx is still not without controversy and with insufficient high-quality data to support a recommendation for use in clinical practice [33,34,35]. Further studies in animals and humans are currently underway to further validate this promising technology.

5 Conclusion

TOBI produces the indices dz and DZx which track changes in parameters that affect CA such as ICP, CPP, and CBF, and indicate the state of CA such as PRx. Further development of such indices might prove to be a more easily accessible alternative and suitable surrogate for the more invasive, time consuming, and expensive standard methods. The information that can be extracted from the impedance signal may prove to be useful as part of a portable, easily-applied high fidelity diagnostics device.