1.

Introduction

Out-of-hospital cardiac arrest occurs in over 350,000 people in North America and has a low survival rate of $<10\%$.^1,2 Approximately two-thirds of out-of-hospital cardiac arrest patients who are resuscitated subsequently die in hospital due to neurological injuries, which are a result of prolonged hypoperfusion and ischemia.^3,4 There is a need to monitor cerebral perfusion and oxygenation. The American Heart Association has acknowledged that brain injury after cardiac arrest should be a critical focus in clinical research.⁵ Cerebral monitoring is essential to provide optimal care to decrease ischemic brain injury after cardiac arrest.

There is currently no standard in measuring cerebral perfusion and oxygenation during cardiac arrest resuscitation. Near-infrared spectroscopy (NIRS) has recently been evaluated as a monitoring tool to measure cerebral oxygenation during and after cardiac arrest resuscitation.⁶ NIRS is based on the measurement of the intensity of near-infrared light in the range of 700 to 1100 nm that passes through the skin, skull, scalp, and brain, which depends on the absorption and scattering coefficients of each tissue. The major chromophores absorbing near-infrared light are oxygenated hemoglobin (${\mathrm{HbO}}_2$), deoxygenated hemoglobin (Hb), fat, water, and cytochrome c oxidase (CCO).^7,8 The measurement of ${\mathrm{HbO}}_2$ and Hb has been widely used to determine tissue oxygen saturation in the brain (${\mathrm{tSO}}_2$); however, CCO has been challenging to detect.⁸ CCO plays a key role in the mitochondrial oxygen metabolism. While the total tissue concentration of CCO does not rapidly change over time, the changes in the difference between the concentrations of the oxidized and reduced forms of CCO (redox) can be measured based on the property that oxidized and reduced forms of CCO have different shapes of their absorption spectra.⁸ The main challenge in measuring cerebral CCO redox changes ($\mathrm{\Delta }[\mathrm{CCO}]$) is that CCO concentration is significantly lower compared to hemoglobin.⁹ Commercially available NIRS devices use a few isolated wavelengths of near-infrared light [multispectral near-infrared spectroscopy (mNIRS)] instead of the full light spectrum, which may not be quantitatively accurate to measuring hemoglobin and CCO changes compared to using the whole broadband spectrum [hyperspectral near-infrared spectroscopy (hNIRS)].^9–13

We have previously shown that changes in $\mathrm{\Delta }[\mathrm{CCO}]$ occurred during functional activations and when oxygen delivery was compromised.^12,14–18 hNIRS studies on humans have established evidence that $\mathrm{\Delta }[\mathrm{CCO}]$ has clinical relevance representing the metabolic state of brain cells and a measure of changes in cerebral oxygen delivery.^15,19–21 Using NIRS simultaneously with transcranial Doppler ultrasonography, Tisdall et al.²⁰ found a linear relationship between $\mathrm{\Delta }[\mathrm{CCO}]$ and cerebral oxygen delivery during hypoxemia in healthy adult humans. $\mathrm{\Delta }[\mathrm{CCO}]$ has also been shown to increase compared to microdialysate lactate/pyruvate ratios, a marker of anaerobic metabolism.²¹ Using respiratory challenges, Holper and Mann²² showed that $\mathrm{\Delta }[\mathrm{CCO}]$ responded to hypocapnia and hypercapnia that were comparable to hemodynamic changes. Nosrati et al.¹² found that in a porcine model of cardiac arrest, cardiopulmonary resuscitation (CPR) resulted in a higher increase in $\mathrm{\Delta }[\mathrm{CCO}]$ than in cerebral tissue oxygen saturation (${\mathrm{tSO}}_2$). Furthermore, in small observational studies, patients who survived after cardiac arrest had significantly higher cerebral ${\mathrm{tSO}}_2$ compared to those who died.^23–25

The objective of this study was to examine the feasibility of using hNIRS to detect sudden changes in cerebral oxygenation and metabolism in patients undergoing transcatheter aortic valve insertion (TAVI).²⁶ During a TAVI procedure, rapid ventricular pacing (RVP) is used to decrease stroke volume during balloon valvuloplasty and valve implantation. RVP can be used to model cardiac arrest as it produces periods of low flow or no flow perfusion states. hNIRS represents a technology that can help measure cerebral hemoglobin content and metabolism in the cardiac arrest patients and to determine the real-time cerebral response to CPR.

2.

Materials and Methods

2.2.

Near-Infrared Spectroscopy Setup

Both hNIRS custom sensors (noninvasive stick pads) and mNIRS (Equanox 7600, Nonin, Michigan) were placed on each patient’s forehead for the entire TAVI procedure. The hNIRS sensor was placed over the left forehead and the mNIRS sensor was applied onto the right forehead. A band was applied across the forehead to ensure contact of both sensors. The Equanox 7600 utilizes four wavelengths (730, 760, 810, and 880 nm) with four channels and uses calculations based on the Beer–Lambert law to determine tissue oxygen saturation (${\mathrm{tSO}}_2$).²⁷ The sensor has four mNIRS channels: two at 2 cm and two at 4 cm source–detector separations to exclude the influence of the skull and scalp on the measurement of the cerebral ${\mathrm{tSO}}_2$.²⁷

The hNIRS experimental setup of the TAVI procedure is shown in Fig. 1. The spectra were collected at the sampling rate of 2 Hz by two fiber optic spectrometers: QE 65000 and USB 4000 (Ocean Optics, Dunedin, Florida) at 3 and 1 cm, respectively, to separate the extracerebral and cerebral measurements. Both of these spectrometers had their range from 650 to 1100 nm. QE65000 had a high signal-to-noise ratio (1000:1 single acquisition) sufficient to measure light at 3-cm distance from the source. Two custom-made 2-m-long optical fiber bundles (each made of seven 0.5 NA, $400\mu \mathrm{m}$ core $Ø$ multimode polymer-clad fibers with broad UV/VIS/NIR spectral range of 400 to 2200 nm Thorlabs, New Jersey) connected spectrometers with the patient’s head. The two other optical fiber bundles were used to connect the probe with a halogen lamp light source (Fiber-Lite Dc 950H Fiber Optic Illuminator, Dolan-Jenner, Massachusetts). The Spectra Suite (Ocean Optics, Florida) software was used to collect the broadband continuous-wave hNIRS data from both spectrometers with dark-signal correction.

Fig. 1

TAVI experimental setup.

2.3.

Hyperspectral Near-Infrared Spectroscopy Data Processing and Analysis

hNIRS measures the absolute concentrations of tissue hemoglobin [${\mathrm{HbO}}_2$] and [Hb], and changes $\mathrm{\Delta }[\mathrm{CCO}]$. hNIRS data acquired at 1- and 3-cm channels represented the extracerebral layer ($\text{scalp}+\text{skull}$) and a combined extracerebral and cerebral tissue volume, respectively. The data processing algorithm of hNIRS in our study was based on the analytical solution to the diffusion equation and implemented in MATLAB (MathWorks, Massachusetts, version R2016b).^12–13 The baseline concentrations of ${\mathrm{HbO}}_2$, Hb, and water were determined using nonlinear least-square fitting¹⁰ of the optical absorbance modeled as a function of the optical absorption coefficient ${\mu }_{\mathrm{a}}(\lambda )$

Eq. (1)

$${\mu }_{\mathrm{a}}(\lambda )=[\mathrm{Hb}]\epsilon {(\lambda )}_{\mathrm{Hb}}+[{\mathrm{HbO}}_2]\epsilon {(\lambda )}_{{\mathrm{HbO}}_2}+[{\mathrm{H}}_2\mathrm{O}]\epsilon {(\lambda )}_{{\mathrm{H}}_2\mathrm{O}},$$

and the reduced scattering coefficient ${\mu }_{\mathrm{s}}^{\prime }(\lambda )$ in the spectral band of 650 to 1000 nm. The reduced scattering coefficients (${\mu }_{\mathrm{s}}^{\prime }$) were assumed to be independent of chromophore concentrations and were calculated for each wavelength ($\lambda$) using the power law, as described by Yeganeh et al.¹⁰ Temporal changes in the hemoglobin concentrations ${\mathrm{HbO}}_2$, Hb, and CCO redox were resolved using the time-spectral domain-independent component analysis for the signal denoising²⁸ and a multistep data-fitting algorithm^12,15 based on the analytical solution to the diffusion equation by relating changes in ${\mathrm{HbO}}_2$, Hb, and CCO to the changes in the optical absorbance as

Eq. (2)

$$\mathrm{\Delta }{\mu }_{\mathrm{a}}(\lambda )=\mathrm{\Delta }[\mathrm{Hb}]\epsilon {(\lambda )}_{\mathrm{Hb}}+\mathrm{\Delta }[{\mathrm{HbO}}_2]\epsilon {(\lambda )}_{{\mathrm{HbO}}_2}+\mathrm{\Delta }[\mathrm{CCO}]\epsilon {(\lambda )}_{\mathrm{cco}},$$

where $\epsilon {(\lambda )}_x$ were the spectra of the extinction coefficients of ${\mathrm{HbO}}_2$, Hb, and CCO redox.^29,30 The data fitting was performed is two steps. First, $\mathrm{\Delta }[{\mathrm{HbO}}_2]$ and $\mathrm{\Delta }[\mathrm{Hb}]$ were calculated assuming $\mathrm{\Delta }[\mathrm{CCO}]=0$ and second, $\mathrm{\Delta }[\mathrm{CCO}]$ was calculated and retained only if the addition of $\mathrm{\Delta }[\mathrm{CCO}]$ resulted in the improvement of the fit quality. Additional details on recovering the absolute values and changes of chromophore concentrations from the hNIRS data have been previously described.^10,12,15

The cerebral tissue saturation of oxygen was calculated as the fraction of ${\mathrm{HbO}}_2$ relative to the total hemoglobin in the blood:

Eq. (3)

$${\mathrm{tSO}}_2=\frac{[{\mathrm{HbO}}_2]}{[{\mathrm{HbO}}_2]+[\mathrm{Hb}]}(\%).$$

The analysis of TAVI data using our homogenous model for the source–detector distances at 1 and 3 cm, respectively, was performed using MATLAB. All data were smoothed using the 0.5-Hz cutoff low-pass filter. Since ${\mathrm{tSO}}_2$ showed prominent dips in the end of the RVP periods Fig. 2(a), we used the time of the lowest ${\mathrm{tSO}}_2$ value as a central timestamp for the calculation of the signal change during RVPs. To calculate changes in NIRS parameters produced by RVPs, we averaged their values $X$ at five time points around the time of the lowest ${\mathrm{tSO}}_2$ value during RVP and subtracted the average baseline values $Y$ during 1 min prior to RVP. To assess the statistical significance of changes, we performed the left-tailed two-sample $t$-tests using the MATLAB function $[h,p]=t\text{test}2(X,Y,‘\text{Tail}’,‘\text{left}’)$, where $h$ was the test outcome and $p$ was the $p$-value.

Fig. 2

Example of the time course of the changes during a TAVI procedure (patient 10) for: (a) [${\mathrm{tSO}}_2$] for hNIRS at the 1- and 3-cm channels and mNIRS device, (b) [${\mathrm{HbO}}_2$] and [Hb] changes at the 1-cm channel, (c) at 3-cm channel, and (d) $\mathrm{\Delta }[\mathrm{CCO}]$ at the 1- and 3-cm channels. Dashed lines indicate RVP episodes.

3.

Results

Changes in [${\mathrm{HbO}}_2$], [Hb], ${\mathrm{tSO}}_2$, and $\mathrm{\Delta }[\mathrm{CCO}]$ measured at 1 cm and 3 cm by hNIRS were typical for all patients. We show these changes in one sample TAVI patient (Fig. 2). [${\mathrm{HbO}}_2$], ${\mathrm{tSO}}_2$, and $\mathrm{\Delta }[\mathrm{CCO}]$ showed rapid decrease while [Hb] showed an increase during three RVP episodes, which lasted for 20, 28, and 24 s, respectively. The longest (second) RVP caused significantly larger responses in all parameters compared to the first and third RVP.

Both [${\mathrm{HbO}}_2$] and [Hb] measured by the 1-cm channel were lower compared to the 3-cm channel [Figs. 2(b) and 2(c)]. The average changes during RVP in [${\mathrm{HbO}}_2$] and [Hb] were $\sim 5.2\pm 1.97\mu \mathrm{M}$ and $0.92\pm 0.26\mu \mathrm{M}$, respectively, at the 3-cm channel and $2.3\pm 2.3\mu \mathrm{M}$ and $0.28\pm 0.33\mu \mathrm{M}$, respectively, at the 1-cm channel. $\mathrm{\Delta }[\mathrm{CCO}]$ decreased significantly at the 3-cm channel during the second and third RVPs, whereas $\mathrm{\Delta }[\mathrm{CCO}]$ at the 1-cm channel did not show significant changes during RVP ($p>0.1$) [Fig. 2(d)].

Baseline values of ${\mathrm{tSO}}_2$ measured by hNIRS at 1- and 3-cm channels and by mNIRS channels were between 70% and 76%. The time course of all three ${\mathrm{tSO}}_2$ measurements showed similar trends during and after RVPs with significant decrease at the end of each RVP ($p<0.01$). The magnitude of ${\mathrm{tSO}}_2$ changes during RVPs measured by the mNIRS, and 1-cm hNIRS channels were similar and higher than at the 3-cm hNIRS channel. On average, ${\mathrm{tSO}}_2$ at 1-cm channel decreased by $6.95\%\pm 0.79\%$, compared to $2.96\%\pm 0.30\%$ at 3-cm channel.

The measurements of ${\mathrm{tSO}}_2$ and $\mathrm{\Delta }[\mathrm{CCO}]$ during RVPs for all 10 TAVI patients are summarized in Table 1. $\mathrm{\Delta }[\mathrm{CCO}]$ measured by the 3-cm hNIRS channel showed significant drops ($0.1\pm 0.07\mu \mathrm{M}$ on average, $p<0.05$) during 12 RVP episodes in 8 patients, whereas $\mathrm{\Delta }[\mathrm{CCO}]$ measured by the 1-cm channel was never significant ($p>0.05$). The correlation coefficients between $\mathrm{\Delta }[\mathrm{CCO}]$ and $\mathrm{\Delta }{\mathrm{tSO}}_2$ are shown in Table 2. The highest correlation $0.7\pm 0.2$ was between $\mathrm{\Delta }[\mathrm{CCO}]$ and $\mathrm{\Delta }{\mathrm{tSO}}_2$ measured by hNIRS at 3 cm.

Table 1

The hNIRS and mNIRS measured cerebral hemodynamics and metabolism for the 1- and 3-cm channels during RVP.

Table 2

The correlation coefficients between values from Table 1.

4.

Discussion

We measured cerebral oxygen saturation and metabolism using hNIRS in TAVI patients during transient periods of RVP, which can be used to model the low flow or no flow states of cardiac arrest. Overall, we observed significant changes in [${\mathrm{HbO}}_2$], [Hb], and ${\mathrm{tSO}}_2$ during RVP, which were consistent with previously published studies.^26,31–33 $\mathrm{\Delta }[\mathrm{CCO}]$ decreased during RVP, which may represent changes in the intracellular oxygen consumption by cerebral cells.^8,17–18,31

The 1-cm channel reflected the changes in the noncerebral tissue, such as scalp and skull, and the 3-cm channel measured both the noncerebral tissue and the cerebral cortex. Interestingly, ${\mathrm{tSO}}_2$ decreased during RVPs but not $\mathrm{\Delta }[\mathrm{CCO}]$ at the 1-cm channel. This may have been due to the brain being more sensitive to profound hypotension, even during short intervals of RVP, compared to the scalp. The metabolic rate is also higher in the brain compared to skin and muscle,³² which may help to explain these results. Decreases in brain $\mathrm{\Delta }[\mathrm{CCO}]$ may occur only after longer periods of hypotension if RVPs lasted longer in our TAVI patients. Alternatively, since the oxygen metabolic rate is higher in the brain^20,33,34 and hemodynamic noise at small separations was higher than at larger separations,³⁵ our hNIRS monitor may not have detected these small changes at the 1-cm distance.

In two patients, $\mathrm{\Delta }[\mathrm{CCO}]$ changes were not measured in both the 1- and 3-cm detector distances. This may have been due to smaller changes in $\mathrm{\Delta }[\mathrm{CCO}]$ that were not detected by our sensors compared to other patients in this study. Alternatively, as the brain atrophies with increased age, there is a need to increase detector separation ($>3\mathrm{cm}$) in order to penetrate brain parenchyma.^36,37 Okada and Delpy³⁸ suggested that the increased cerebrospinal fluid could substantially affect the penetration of near-infrared light into the brain. Furthermore, the average differential path-length factor also increases with age due to the higher scattering in extracerebral tissue.³⁹

We found that $\mathrm{\Delta }{\mathrm{tSO}}_2$ measured by hNIRS at 3 cm ($\sim 2.96\%$) was lower than at 1 cm ($\sim 6.95\%$) despite the reverse in ${\mathrm{HbO}}_2$. This may have been due to the differences in the proportion of vasculature (with higher [HbO2] and [Hb]) in the scalp and skull compared to the brain. This heterogeneity could limit the quantitative accuracy of cerebral ${\mathrm{tSO}}_2$ changes by hNIRS. The cerebral ${\mathrm{tSO}}_2$ as measured by mNIRS monitor may in fact be more reliable due to the calibration of Equanox 7600 for the influence of the skull and scalp.²⁷

Our results agree with previous NIRS studies comparing magnetic resonance imaging,⁴⁰ time-resolved spectroscopy measurement, and Monte Carlo simulations.¹⁶ These studies found that hemodynamic changes correlated with the cerebral hemoglobin signal in the extracerebral tissues. Monte Carlo simulations¹⁶ showed that the largest amount of detected photons propagating into the adult brain was from a 3-cm source–detector separation. However, $\mathrm{\Delta }[\mathrm{CCO}]$ was less prone to extracerebral contamination and more specific to mitochondrial concentration.³³ Our results also agree with Kolyva et al.¹⁷ who found that $\mathrm{\Delta }[\mathrm{CCO}]$ increased with longer source–detector distances in four different experimental settings: hypoxia, hyperoxia, hypocapnia, and hypercapnia.

This study had several limitations. The biological and optical properties and the thickness of the noncerebral tissue likely varied between patients, which was not determined in this study. Although this was beyond the scope of this study, we plan to apply a two-layer model to our data measured at 1- and 3-cm separations to improve the accuracy of measuring [HbO2], [Hb], ${\mathrm{tSO}}_2$, and $\mathrm{\Delta }[\mathrm{CCO}]$.¹¹ The application of a two-layer model may better evaluate the differences in the brain compared to the scalp and skull. Moreover, future studies will be performed to determine the optimal distances between optodes to accurately measure [HbO2], [Hb], ${\mathrm{tSO}}_2$, and $\mathrm{\Delta }[\mathrm{CCO}]$ in the brain.

5.

Conclusion

We have shown that hNIRS is sensitive to sudden changes in ${\mathrm{tSO}}_2$, [${\mathrm{HbO}}_2$], [Hb], and $\mathrm{\Delta }[\mathrm{CCO}]$ during profound hypoperfusion in TAVI patients. The changes in CCO redox measured by hNIRS were more specific marker of cerebral status than ${\mathrm{tSO}}_2$.