INTRODUCTION

TOP ABSTRACT INTRODUCTION THEORY METHODS RESULTS DISCUSSION REFERENCES

CEREBRAL BLOOD VOLUME (CBV) may be measured in human subjects by using positron emission tomography (PET) (12, 13, 19) or single-photon emission-computed tomography (SPECT) (29, 33). These methods require the use of radioisotopes and are not bedside tests. Near-infrared spectroscopy (NIRS), which was first described in 1977 (16), has been used to monitor changes in CBV in the head of a duck (3). In humans, changes in CBV have been studied for many years (10) on an arbitrary scale. NIRS has been used in neonates to show changes in CBV with blood pressure changes (1). More recently, quantified near-infrared (NIR) absorption methods for the measurement of CBV have been described in human infants (35) and human adults (8). These methods avoid the use of radioisotopes and X-rays and may be performed at the bedside. A measured change is induced in the concentrations of oxyhemoglobin (HbO₂) and deoxyhemoglobin (Hb) in the blood supplying the cerebral tissue, and changes in the cerebral concentrations of these chromophores are measured by using a cerebral NIR probe applied to the scalp. The volume of blood contained within the brain tissue may then be calculated, and, because a comparison is made between concentrations in the tracer and concentrations in the tissue, this is possible without knowledge of the volume of the brain tissue sampled. Changes in these chromophore concentrations are achieved by causing a small desaturation of ~5% of the hemoglobin in the arterial blood by reducing the inspired O₂ fraction.

NIRS has been used clinically for more than 10 yr for the monitoring of changes in cerebral oxygenation. In the NIR range (650-1,000 nm), there is relatively little absorption of light by tissue (2). Changes in NIR light absorption at different wavelengths are converted into changes in chromophore concentrations by using a modification of the Beer-Lambert law and solving for each chromophore by using an inverse matrix solution. The detailed theory is described elsewhere (4, 34). To calculate these chromophore changes, it is necessary to know the distance traveled by the light, which is referred to as the pathlength. The light takes a highly convoluted path through biological tissue with the consequent scattering and absorption dependent on both the wavelength and the tissue type (2). Because tissue is a high scatterer of light, the light travels a much longer distance than the geometric distance between the emitter and detector, and a differential pathlength factor (DPF) is used in the calculation to account for this. The product of the emitter-detector spacing and the DPF is the average propagation distance of the light through the tissue (7). In addition to changes in the concentrations of the two measured hemoglobin species, the use of multiple wavelengths of light enables changes in oxidized cytochrome oxidase to be measured; the changes in HbO₂ and Hb and the changes in the redox state of the Cu_A center of cytochrome oxidase are expressed in micromoles per liter of brain tissue. In the adult population, light must pass through a considerable thickness of extracerebral tissue as it enters and leaves the skull. This creates two distinct issues to be addressed separately: 1) the volume of tissue in which the hemodynamic and oxygenation changes mostly occur and 2) the total volume of tissue that is illuminated and the proportions that the cerebral and extracerebral tissues contribute to this total volume. Even though the extracerebral pathlength is long, the validity of NIRS measurements has been justified by the argument that, as the extracerebral path has a much lower specific blood volume than cerebral tissue, it merely acts as a dead space and is hemodynamically inert (17, 24, 25, 32). Recently, it has become increasingly recognized that a correction or scaling factor is required to allow for this extracerebral tissue volume, thus forming a separate issue from the fact that the changes in NIR light absorption are mostly due to hemodynamic or oxygenation changes occurring in the brain (25).

Indocyanine green (ICG) is used in this study as a highly absorbing intravascular chromophore. It is a tricarbocyanine dye that binds to albumin and therefore remains in the plasma. It shows strong absorption in the NIR range with maximal absorption at 805 nm, and recent work has accurately described the absorption spectrum of bound ICG. Changes in the attenuation of NIR light and, therefore, the optical density (OD) of biological tissue have been recorded after the administration of ICG (15, 22, 27, 28, 31). Because ICG is not an endogenous chromophore, a zero-concentration reference point is available before any dye is administered. This enables absolute tissue concentrations of the dye to be quantified by using NIRS (27, 28). We describe an integrated technique for absolute quantification of CBV in adults that uses ICG as the intravascular marker and NIRS to measure tissue ICG concentration. This avoids the requirement for hypoxia or ionizing radiation. It can be performed with portable equipment and has the potential to increase the accuracy of NIRS techniques for CBV measurement.

	THEORY

TOP ABSTRACT INTRODUCTION THEORY METHODS RESULTS DISCUSSION REFERENCES

CBV is the ratio of blood to tissue volume in the brain. When NIRS is used, it is not possible to know the volume of brain interrogated by the NIRS signal, and CBV is calculated by comparing concentrations of chromophores in the blood and the blood-containing cerebral tissue or induced changes in these concentrations. In the O₂ desaturation and resaturation method (8, 35), changes in brain tissue HbO₂ concentration measured by NIRS are compared with the HbO₂ changes in arterial blood measured by pulse oximetry to calculate CBV. All compartments of the vascular system can contribute to the concentration changes measured by NIRS; however, by far the largest contribution is provided by the venous and capillary compartment. It is assumed that a small arterial desaturation of ~5% is accompanied by a similar capillary and venous desaturation and that the cerebral O₂ consumption remains static. In our method, ICG is used as an exogenous chromophore. We assume that, because our measurement is made well after the bolus of dye has mixed with the plasma, concentrations in the cerebral blood (comprising arterial, venous, and capillary components) may be estimated by measurement of a peripheral venous sample, if allowance is made for the different hematocrit (Hct) of cerebral blood due to the Fahraeus effect (9, 26). ICG may allow greater accuracy because larger changes in tissue OD are observed (15). The disadvantage of large changes in OD is that complete extinction of the light may occur in some blood vessels (20), thereby altering the volume of tissue that is interrogated or the mean optical pathlength during the period of the measurement. We, therefore, used a dose of ICG that reliably gives a maximal attenuation of <0.4 OD (attenuation is calculated from raw data: log₁₀I₀/I, where I₀ is the incident light intensity and I is the detected light intensity) for the period of measurement. ICG concentration changes during the first 3 min after bolus administration are associated with higher attenuations and are partly due to redistribution of the bolus of dye. Accordingly, these measurements were not used. An integrated concentration time measurement within the first 20 min of elimination that does not use this redistribution period allows accuracy to be further improved, compared with single-point comparisons of blood and tissue chromophore concentrations, while still remaining in a timeframe in which single-compartment kinetics can be applied. The ratio between the blood and tissue concentration over the full measurement period is compared by using blood and tissue time integrals to derive a mean CBV, thus minimizing changes that may be seen in the shorter time span because of alterations in blood pressure and arterial CO₂ tension

(1)

where [ICG]_tissue is the concentration of ICG in the illuminated volume of tissue measured by NIRS; [ICG]_blood is the concentration of ICG in cerebral blood estimated by high-performance liquid chromatography (HPLC) analysis of the peripheral venous sample, compensating for the Fahraeus effect; and t₁ and t₂ are the start and finish times, respectively, of the measurement period.

ICG concentration decreases in a monoexponential manner during the measurement period as it is eliminated by the liver. The algorithm we have used to calculate tissue ICG concentrations uses an inverse matrix solution (34) and the absorption coefficients listed in Table 1. Coefficients for HbO₂ and Hb were described by Wray et al. (34). The absorption coefficients we used for albumin-bound ICG were previously used by Roberts et al. (28). The relationship between cranial optical pathlength and age has been described for particular wavelengths (7); however, for the wavelengths used by our apparatus, we did not have these data. We have assumed that the DPF remains constant during the measurement and have employed a value of 6.5 for all our subjects. There will be a small age-dependent error because of this assumption.

The integrated tissue and blood concentrations over a 17-min period (from minutes 3 to 20 after the ICG bolus) are determined by calculating the areas under the monoexponential elimination curves for NIRS tissue (AUC_tissue) measurements and plasma concentration (AUC_plasma) measurements. These areas are computed after fitting curves for the best solution of the following equation

(2)

where A is a constant and k is the disappearance rate constant, and the time integral (t) between minutes 3 and 20 is determined for plasma and tissue methods.

The molecular mass of ICG is 775 Da. This is used to transform the units of tissue concentration from micromoles per liter to milligrams per liter. The large-vessel Hct is assumed to be 0.4. The integrated large-vessel blood ICG concentration from the concentration-time curve is the product of AUC_plasma and (1  Hct).

Previous methods using changes in HbO₂ to derive CBV have made allowance for the Fahraeus effect (9, 26) in which the Hct of blood in the small cerebral vessels is less than that in the large vessels because of axial streaming and differing plasma and red cell flow rates. This is normally referred to as the cerebral-to-large-vessel Hct ratio and is known to vary among subjects, vary with Hct (30), and decrease with cerebral vasodilatation (29). ICG is plasma bound, and the variation in plasma volume between large vessels and the small cerebral vessels must be taken into consideration. We describe a new term, the cerebral-to-large-vessel plasma volume ratio (PVR). This variable, which we have calculated from the same data used to determine the previously quoted cerebral-to-large-vessel Hct ratio of 0.75 (29) is used in Eq. 3. The PVR will vary with the Hct and is calculated as 1.17 for a Hct of 0.4. The specific weight of brain (Swt_brain) is assumed to be 1,050 g/l (6). The factor 10⁵ takes account of the dimensions of the various measures and converts the CBV units from milligrams per liter to milliliters per 100 g of tissue, which is the usual unit of CBV (Table 2)

(3)

	METHODS

TOP ABSTRACT INTRODUCTION THEORY METHODS RESULTS DISCUSSION REFERENCES

Subjects

The Critikon 2001 NIRS monitor (Johnson & Johnson Medical) was used with the probe placed on the right side of the subjects' forehead in the mid-pupillary line in front of the hairline in a position to avoid the midline air sinuses and the temporalis muscle. The spacing between the emitter and the detector was 55 mm, and the supplied probe was applied to the scalp with a specially designed adhesive pad (Johnson & Johnson Medical) and further secured with elastoplast tape. The subjects were in the sitting position.

Each subject received 0.3 mg/kg of freshly reconstituted ICG (Becton Dickenson, Cockeysville, MD) as a bolus injection into an antecubital vein. Changes in absorption of NIR light were recorded with the instrument over the next 20 min while venous samples were taken from the other arm at 0, 1, 3, 5, 10, 15, 20, and 30 min after dye administration.

HPLC

NIRS

Analysis

	RESULTS

TOP ABSTRACT INTRODUCTION THEORY METHODS RESULTS DISCUSSION REFERENCES

A typical example of the inverse matrix solution for ICG concentration over the 20 min after bolus administration is shown in Fig. 1, demonstrating the initial peak and redistribution.

View larger version (12K): [in this window] [in a new window]   Fig. 1.   Typical concentration-time plot in 1 subject of indocyanine green concentration ([ICG]) measured by cerebral near-infrared spectroscopy (NIRS) after bolus of 0.3 mg/kg ICG. Data are time averaged for 3-s epochs. Initial peak and redistribution are seen.

NIRS signal strength after the bolus injection of ICG, as determined by the monitor receiving adequate levels of light to prevent low-signal alarms from being activated, was acceptable in 9 of the 10 subjects. In one subject, the signal was borderline before ICG administration and remained borderline after the bolus. A typical example of the concentration-time curves for plasma and tissue methods is shown in Fig. 2. OD changes obtained from the raw data at peak ICG concentration and at 3 min after administration are shown for each subject in Table 3.

View larger version (18K): [in this window] [in a new window]   Fig. 2.   Typical ICG elimination curves between minutes 3 and 20 after 0.3 mg/kg bolus administration for NIRS and high-performance liquid chromatography (HPLC) measurements in 1 subject. NIRS data points are time averaged for 15-s epochs. Units of tissue [ICG] have been transformed to milligrams per liter, and the 2 curves are almost superimposed, although concentrations measured differ between methods by a factor of ~100.

There was no statistically significant difference among the mean k values, which were 0.23 for each method. Values for k, A, and calculated CBV for individual subjects are shown in Table 4. The mean CBV was 1.1 ± 0.39 (SD) ml/100 g tissue. The correlation coefficients (r) of the monoexponential model to the data for tissue and plasma concentration measurements in each subject were 0.99 for all the HPLC-measured plasma curves; the r values for the NIRS-measured tissue curves are given in Table 3. In the subject with borderline signal strength, the NIRS tissue curve correlation was 0.89. 

	DISCUSSION

TOP ABSTRACT INTRODUCTION THEORY METHODS RESULTS DISCUSSION REFERENCES

The comparison of blood and tissue ICG concentration by an integration method is a new technique for the measurement of CBV that avoids hypoxia and ionizing radiation and has the potential to be used as a bedside test. The result for CBV is a mean for both the measurement period and the region of brain tissue that has been interrogated. The result for CBV of 1.1 ml/100 g is low in comparison with previous NIRS measurements in adults (8, 14, 25), which reported values of 2.3-5.38 ml/100 g, and lower than the reported positron-emission tomography (PET) (12, 13, 19) and single-photon emitted computer tomography (SPECT) (29, 33) measurements of 3.3-5.2 ml/100 g. This range of quoted values may be due to differing proportions of white and gray matter in the volume of tissue interrogated. There may be several reasons for our lower result, which are discussed below. If it is assumed that our volunteers had a normal CBV, then this low result also provides empirical evidence for the extracerebral contribution to the total optical pathlength when scalp recordings of NIR signals are made in adults. The main reason for our low result relates to the scaling factor, which needs to be applied to allow for this extracerebral pathlength, permit the estimation of real cerebral chromophore concentration changes, and correctly derive parameters such as CBV. At present, we cannot calculate this scaling factor, which some have referred to as a partial pathlength factor (23); we can only compare it, in the broadest sense, with methods such as SPECT. In the future, work with NIR imaging and NIR modeling may allow calculation of this scaling factor.

Physiological Assumptions of the Method

Optical properties. We assumed that the brain tissue interrogated for ICG concentration was an optically homogenous compartment. The optical properties of the head are complex: light tends to track preferentially through the cerebrospinal fluid (CSF) layer, and complex situations occur at the boundary between tissues of different vascularity and OD. The main cerebral arteries entering the parenchyma at the base of the brain are 1 mm in diameter; these branch to vessels 0.2 mm in diameter. Most vessels penetrating the cortex are ~0.04 mm in diameter. Firbank et al. (11) noted that, as the majority of cerebral vessels were of small diameter, the distribution of blood in the vessels would not have a large effect on the measurement of CBV. Time-resolved reflectance spectroscopy has shown that it may be difficult for photons to pass through larger vessels because they absorb the light to a greater extent than do the small vessels and the background tissue (20). This leads to the illuminated tissue, which is responsible for the changes in NIR absorption, being composed of a disproportionately large volume of capillaries. Monte Carlo modeling suggests that, at higher chromophore concentrations, as may occur soon after a bolus injection of ICG, the extinction of light may be a real phenomenon in larger blood vessels (20) and produce the effect that the volume of tissue examined is composed of proportionately more small vessels. There is then the possibility that the volume of tissue interrogated and the mean optical pathlength may change as the concentration of ICG decreases toward the end of the measurement period, increasing the mean size of the vessel sampled. This type of bias would change with changing ICG concentrations during the measurement and would be manifested as a distribution function rather than a constant linear change. To minimize any bias of this type, we have used small doses of ICG and a sensitive plasma assay. We compare our cerebral ICG concentrations with those of Roberts et al. (27, 28): the cerebral ICG concentration data we have used to calculate CBV are of lower values than that used in both of their studies. A methodological problem of this type would have a greater effect at higher cerebral tissue ICG concentrations at the beginning of the measurement than the lower ones at the end of the measurement, whereas the concentrations of the venous samples would be unaffected by any bias of this kind. This would prevent a good tissue curve monoexponential fit and good correlation of exponents between methods; our plasma disappearance rate constant was 0.23 for both methods.

Physiological instability. Cytochrome oxidase concentration was not measured and was assumed to remain constant in our population over the measurement period. We elected to maintain an overspecified system, i.e., three unknown species estimated by absorbance at four wavelengths to maintain accuracy in the algorithm. Cross talk in a four-species algorithm between ICG and cytochrome signals is inevitable because of the small magnitude of the cytochrome signal, the large magnitude of the ICG signal, and the similarity of the two respective absorption spectra. Changes in cytochrome oxidase are unlikely to affect our signal significantly, as changes in OD would be of a different order of magnitude to those caused by ICG administration. Arterial CO₂ tension, blood pressure, and actual CBV were assumed to remain constant throughout the measurement period. First, we would expect minimal change in arterial CO₂ tension in our volunteers, who were quietly breathing at rest. Second, we would also expect significant changes in arterial CO₂ to have an effect on the exponential curve and therefore on the goodness of fit. All but two correlation coefficients for curve fits were 0.99, making an error caused by CO₂ changes unlikely. Our method avoided the need for hypoxemia, which has been used in previous methods and which is known to be a potent vasodilator of the cerebral circulation, albeit techniques that use hypoxemia endeavor to avoid reducing blood O₂ tension to a level at which this occurs.

Cerebral-to-large-vessel Hct ratio. The problem of estimation of the Hct and plasma volume of the blood contained within the tissue illuminated by the NIR signal is complex, with a range of values quoted; a detailed discussion is provided elsewhere (18). We did not measure the individual Hct values of each subject. After review of the literature relating to Fahraeus' effect (9, 26, 30), we do not believe that measurement in each subject of the large-vessel Hct would improve the accuracy of the subsequent calculation of the cerebral vessel Hct and cerebral plasma volume in an individual subject. In animal studies, a lowering of large-vessel Hct is not associated with a significant lowering of the cerebral Hct (30). We, therefore, chose to apply a standardized cerebral Hct and plasma volume based on a large-vessel Hct of 0.4 and a cerebral-to-large-vessel ratio of 0.75 from the work of Sakai et al. (29). This is consistent with a PVR of 1.17. They quote a regional value for cerebral-to-large-vessel Hct ratio that is lower than that of previous authors who used two-dimensional and whole brain methods. They cited their lower value as due to their sample volume comprising proportionally more capillaries and arterioles, as the average Hct of smaller vessels is known to be less. This is as close an approximation to the type of sample that NIRS illuminates as is available in the literature. Radioisotope dual-tracer techniques may be used to calculate red cell volume and plasma volume individually, eliminating the need to know the PVR or cerebral-to-large-vessel Hct ratio (18, 29). A dual-indicator NIRS technique with the use of desaturation/resaturation to calculate red cell volume and ICG elimination to calculate plasma volume should now be possible, enabling the calculation of CBV without the need to estimate or assume values for cerebral-to-large-vessel Hct ratio or PVR.

Anatomic and Postural Considerations

Pathlength and Scaling Factors

Monte Carlo simulations and work with phantoms (23) suggest that the extracerebral pathlength may be up to 75% of the total optical pathlength. The CSF layer, which is known to cause light tracking, may make a major contribution to extracerebral pathlength. Observed chromophore concentration changes, although for the most part due to changes in the brain, should be seen as occurring in the total illuminated tissue volume, which includes the extracerebral tissue and CSF layer. At present, while some researchers are calculating the DPF, there are no commercially available instruments that allow this important measurement, which varies among individual subjects. Neither is it possible to determine the extracerebral pathlength component. Future work with NIR imaging techniques may allow correction for extracerebral pathlength by including a pathlength scaling factor in the algorithm.

If we assume that our volunteers had a normal CBV, then comparison with a value of 4.81 ml/100 g obtained by SPECT (29) suggests that the cerebral component of the total optical pathlength is ~23%. This is in broad agreement with Monte Carlo simulations that predict a 25% cerebral component of the pathlength with an optode spacing of 50 mm (23), and cerebral blood flow studies that used NIRS probes placed directly on the dura suggested a 30% cerebral component when recordings are made from the scalp (24).

Conclusions