The main interactions of light with tissue are absorption and scattering. For the simple model system of a cuvette, the effects of scattering and absorption on the path of “individ- ual photons” is schematically illustrated in Fig. 2.
1The term “near infrared” indicates that this is the part of infrared light closest in wavelength (“near”) to visible light.
For a quantitative assessment of the concentration of absorbing chromophores, the Beer–Lambert Law states that light attenuation (A) is proportional to the concentration of the absorbing molecule (c). The proportionality factor is termed specific extinction coefficient ε :
A = ε ×c ×d.
However, this relationship assumes infinitesimal concen- trations and disregards scattering. This assumption clearly does not hold for spectroscopy in tissue. Scattering prolongs the path length of light as illustrated by photon 2 in Fig. 2 and thus the path length becomes longer than the distance between sender and receiver. In a typical transcranial study of the human brain, the mean path length of light is about six times as long as the distance between sender and receiver (Duncan et al., 1995). In order to account for the longer path length, in a modified Beer–Lambert Law a DPF (B) is intro- duced. A second modification of the Beer–Lambert equation is necessary since light may be lost due to scatter, not reach- ing the detector (photon 1). The detector cannot differentiate between the loss due to absorption and that due to scatter.
Therefore, in the modified Beer–Lambert Law a term, G, is introduced. This factor depends on the size of the detector and the geometry of the system. From these considerations, the modified Beer–Lambert Law is derived:
A = ε ×c ×d ×B + G.
Assuming constant B and G gives
∆ A = ε × ∆ c ×d ×B,
an equation frequently used for the assessment of con- centration changes.
The terms B and G are “correction” terms accounting for scatter, which is assumed constant to allow for the determi- nation of concentration changes. In principle, however, scatter by itself may also become a measurement parameter.
In fact, there is evidence that light scatter corresponds to physiological processes in nervous tissue as outlined below.
III. Other Optical Parameters Relevant for Near-Infrared Studies
Usually, the energy (of the photon) which is taken up by an absorbing molecule (e.g., hemoglobin or water) is translated into thermal energy. However, depending on the molecule, after a certain delay emission of light at a longer wavelength may occur. When the delay is shorter than 10–8s, the pheno- menon is termed fluorescence, when it lies between 10–8and 10–6s it is termed delayed fluorescence, when it is longer than 10–6s it is termed phosphorescence. Based on fluorescence and phosphorescence molecules can be detected at extremely low concentrations. Most fluorescent tracers are excited by visible light (see Chapter 1); however, tracers for near-
infrared studies are being designed, thus permitting noninva- sive fluorescence studies even in human subjects.
Another optical parameter represents a special case of light scattering. When light is scattered by moving particles the frequency slightly changes (Doppler shift). This fre- quency shift depends on the velocity of the moving particle, the direction of the movement, and the number of inter- actions with a moving particle. Thus the Doppler shift offers an opportunity to assess the movement of particles in tissue.
This method has been implemented as laser-Doppler flowmetry (LDF) for the measurement of cerebral blood flow (Haberl et al., 1989; Dirnagl et al., 1989) on the brain surface and through thinned skull preparations in the animal.
Near-infrared applications employing a correlation spec- troscopy approach allow for noninvasive examinations in the animal (Cheung et al., 2001).
IV. Technical Approaches for Near-Infrared Spectroscopy
and Imaging
The systems which are available for near-infrared studies in humans differ with respect to the principle approach of data acquisition (time-resolved, continuous wave), the tech- nical specifications (number of discrete wavelengths, con- tinuous spectrum), and the number of “channels” used for data acquisition.
The simplest approach uses a continuous wave (cw) light source with either discrete wavelengths (typically between two and seven) (Cope and Delpy, 1988) or a light source emitting across the entire NIR spectrum (Matcher and Cooper, 1994; Heekeren et al., 1999). The only optical parameter measured is attenuation. The light source may be a laser or a LED (for the discrete wavelength approach) or a simple halogen lamp (for the continuous spectrum approach). The advantage of the cw approach is its simplic- ity and flexibility; also a high signal-to-noise ratio is reached. A disadvantage is a strong contribution to the signal changes by superficial, extracerebral structures. The separa- tion of deep and superficial layers can be approximated by multiple source-detector separations.
Another approach uses a pulsed light source typically with a pulse duration on the order of picoseconds (Nomura and Tamura, 1991) [time resolved spectroscopy (TRS)]. In addition to the assessment of total light intensity, this approach assesses the distribution of photon arrival times.
Based on this additional parameter a multilayer depth reso- lution has been proposed (Steinbrink et al., 2001). Also by assessment of the individual DPF the quantification of the concentration changes can be much improved.
Since time-resolved measurements require rather demanding technology, frequency-domain monitors have been developed to more easily assess the mean time of flight
of photons. Instead of using a pulsed light source the inten- sity of the injected light is sinusoidally modulated at a high frequency (100–150 MHz). The reflected light will also show this modulation. The phase delay of the modulation is proportional to the mean time of flight [phase-modulated spectroscopy (PMS) or frequency domain NIRS system].
Measurements based on phase changes are more sensitive to deeper structures than intensity-based measurements (Hemelt and Kang, 1999); compared to a cw approach the depth resolution allows for a coarse differentiation between a superficial and a deeper layer, which may be sufficient to better detect truly cerebral changes. Furthermore, small changes in phase may be related to changes in light scatter- ing (Gratton et al., 1995).
All of the above-mentioned approaches may be per- formed over just a single site (one channel) or many sites.
When a larger number of emitter–detector pairs is used, it is possible to apply various image algorithms (Heekeren et al., 1999; Benaron et al., 2000; Igawa et al., 2001). An example for noninvasive functional optical imaging is given in Fig. 7 (Benaron et al., 2000). The instrument used in this experi- ment was a TRS system which also allowed for some depth resolution of the optical signal. Figure 7 illustrates the spatial match of the optical signal (right part) to the f MRI signal (left part) in the same subject performing a motor (finger tapping) task.
V. Physiological Parameters of NIRS Measurements
Light which is reflected from or transmitted through nervous tissue is influenced by changes in optical properties of the illuminated tissue. This phenomenon was known long
before modern imaging techniques had been developed (Hill and Keynes, 1949), and the generation of high-resolution functional maps of the brain is a cornerstone of invasive functional brain research by optical intrinsic signals (see Chapter 5) in animals (Grinvald, 1992) and human subjects (Haglund et al., 1993; Cannestra et al., 1998). To derive physiological parameters from such intrinsic optical signals, however, is a complicated task, since various physiological events which occur during brain activity are associated with different kinds of optical changes. Thus, in order to under- stand optical signals in terms of underlying physiological parameters, the different contributions to the intrinsic optical signals have to be disentangled (pioneered in the work of Malonek and Grinvald, 1996). In this section, we discuss which physiological parameters influence optical parameters in the tissue and how far the translation to non- invasive NIRS in the human is possible. An overview is given in Table 1.
Absorption changes in biological tissue are dominated by intrinsic chromophores. Based on their color (i.e., their absorption spectra) concentration changes of different com- pounds can be determined according to the above explained modified Beer–Lambert Law. For measurements assessing changes in optical properties only those endogenous chro- mophores which absorb light differentially depending on their functional state are relevant. For example, the absorp- tion spectrum of hemoglobin depends on its oxygenation state, a fact well known in the visible part of the spectrum (venous vs arterial blood) and extending into the NIR (see Fig. 1). Another relevant chromophore is cytochrome c oxidase (Cyt-Ox), the terminal enzyme of oxidative phos- phorylation whose absorption spectrum depends on its redox-state. Cyt-Ox measurements are possible with visible light and with light in the near-infrared wavelength range.
Table 1
Physical (optical) Physiological parameter measurable in Physiological parameter measurable parameter optical methods on exposed brain surface by transcranial near-infrared methods
Absorption [Oxy-Hb] [Oxy-Hb]
[Deoxy-Hb] [Deoxy-Hb]
[Total Hb] [Total Hb]
Cyt-Ox redox state Cyt-Ox redox state
Light scattering Fast scattering signal Fast scattering signal (?)
Membrane potential
Slow scattering signal Slow scattering signal (?)
Volume changes of cell and subcellular compartments
Fluorescence Autofluorescence: NADH Autofluorescence: No significant autofluorescence in NIR
Exogenous fluorescent tracers (contrast agents) NIR fluorescent tracers (contrast agents)
Phosphorescence Exogenous phosphorescent tracers (contrast agents) Feasible in principal with near- infrared phosphorescent dyes Doppler shift Cerebral blood flow (laser-Doppler flowmetry), Feasible transcranially in rodents (Cheung et al., 2001), not
speckle imaging clear whether feasible in humans
However, these measurements are not interchangeable, since measurements employing visible light mainly report on the Cyt-a/a3 redox state within Cyt-Ox, whereas meas- urements employing near-infrared light are dominated by another redox center of the enzyme, CuA. Many other absorbers are present in brain tissue, e.g., melanin or water;
however, these absorbers show no oxygenation-dependent change in color and will not change absolute concentration over the course of the experiment. These chromophores limit penetration depth of the light into tissue by back-
ground absorption, but will not influence the changes in absorption due to functional activation or other cerebral processes.
Independent of concentration changes of light-absorbing molecules, brain activity is accompanied by changes in scatter. Changes in the scattering properties of the tissue have been described on two time scales. Fast changes in scatter are temporally linked to changes in membrane poten- tial (Stepnoski et al., 1991), while slow scatter changes probably reflect cellular or subcellular volume changes (edema) (see below).
The optical phenomena of fluorescence and phosphores- cence can also be taken advantage of for optical studies.
However, the only relevant endogenous fluorescent compo- nent is NADH, sensitive to UV light, which has a very low penetration depth into tissue. In the near infrared, no rele- vant autofluorescent is present in biological tissue. The main relevance of fluorescence lies in detecting contrast agents (extrinsic signals; see also Chapter 4). For such studies, fluorescent dyes are used which change their fluorescence depending on certain physiological/biochemical parameters.
For example, dyes have been developed which are sensitive Figure 1 The extinction spectra of different biologically relevant chro-
mophores from 450 to 1000 nm. Note that the global absorption between 650 and 950 nm is quite low compared to the high absorption of the hemo- globins below 650 nm and the high absorption of water beyond 950 nm. In between light absorption is low enough to allow penetration to a depth of some centimeters, thus making spectroscopic analysis of optical changes in the cerebral cortex feasible. The magnification (×20) of the oxy-Hb, deoxy- Hb, and Cyt-Ox spectra within this “spectroscopic window” illustrates the option to differentiate between them by their differential spectral properties.
Figure 2 Cuvette model of the relation between scatter, absorption, and light attenuation. The different photons (1–4) demonstrate the different potential “fates” in a scattering and absorbing medium. Very few photons will directly reach the detector without scatter or absorption events (ballis- tic photon 3). Some photons will be absorbed by the chromophore (photon 2). Others will not reach the detector, since they are scattered out of the sampling volume (photon 4). Finally some photons will reach the detector but will have traveled a longer path than the geometrical distance (d) between light source and detector (photon 1). Assuming a constant scatter the ratio of photons reaching the detector is proportional to the number of type 2 photons. The fact of multiple scatter will introduce an enlargement of the sampling volume and the mean path length, accounted for by the dif- ferential path-length factor (DPF) in the modified Beer–Lambert Law.
Figure 3Sketch of NIRS setup in reflection mode. Light from the light source is guided to the head by a fiber-optic bundle, a so-called “optode.”
A second optode will collect the light which leaves the head at a distance of some centimeters. The probability that these photons have traveled a path in the banana-shaped sampling volume is much higher than the probability that the photons have traveled through deeper or more superficial layers of the head. Under the assumption of constant scatter the volume does not change over time and the changes in attenuation can be attributed to changes in chromophore concentration in this volume. The difficulty in exactly defining the shape and extent of the volume is reflected by the difficulty in quantifying the changes.
to ion concentrations (Ca, K, etc.), voltage across mem- branes, pH, etc. Linking a fluorescent tracer to antibodies allows for a mapping of receptors or other structures react- ing with the respective antibody. Whereas so far most fluorescent dyes have been designed to be excited with visible light, recently, more and more dyes for the near- infrared wavelength range have been developed, potentially useful for noninvasive near-infrared studies (Bremer et al., 2001). In the animal the phenomenon of phosphorescence has been employed for measuring oxygen concentration (Rumsey et al., 1988; Lindauer et al., 2001).