14.1. INTRODUCTION

For several centuries, optical methods have provided powerful tools to measure an assortment of physiological variables. However, the use of diffuse optical light for physiological measurements and, in particular, for noninvasive monitoring, has a somewhat shorter history. Unlike traditional ray optics as used in light microscopes, the field of diffuse optics pertains to the propagation of light through highly scattering samples, such as biological tissue. In this diffuse regime, the photons can no longer be assumed to travel in a ballistic-like path through the sample, and instead the equations of light propagation must be replaced by stochastic models to account for multiple scattering events and the numerous direction changes experienced by these photons as they migrate through scattering mediums. On one hand, diffuse optical techniques allow measurements through much thicker samples than are possible with traditional optics, including the ability to record noninvasive measurements from the human brain, which is the subject of this chapter. However, on the other hand, there are penalties to be paid for this ability and the progress of these techniques has faced many technical and methodological challenges.

The slowed development of diffuse optical methods was a result of two primary obstacles. The first obstacle was sensitivity—a need to discover a low-absorbing wavelength range where light can penetrate into tissue and to develop sufficiently sensitive detectors to monitor light that exits after traveling relatively large distances through tissue. As it turns out, the near-infrared wavelength range between roughly 600 and 950 nm is relatively poorly absorbed by biological tissue. Within this wavelength region, often referred to as the “near-infrared window,” light can travel through up to several centimeters of tissue because of low intrinsic optical absorption of biological tissue. Moreover, although most components of biological tissue (e.g., proteins and water) have low absorption in these colors of light, oxy-hemoglobin and deoxy-hemoglobin (HbO₂ and Hb, respectively) are the dominant absorbers within this wavelength range. Thus, this spectral region is ideally suited for monitoring changes in these two hemoglobin chromophores, which can be achieved at relatively high signal-to-noise levels.

In 1977, F.F. Jöbsis [1] first demonstrated the use of near-infrared light to noninvasively measure changes in oxy- and deoxy-hemoglobin levels within the brain. This work employed what is now called near-infrared spectroscopy (NIRS), a point-source measurement of absorption changes within tissue. In the early years following Jöbsis’s initial work, many groups followed to examine a variety of biologically relevant parameters in tissue, including the brain [2,3]. Much of this early research, during the 1980s and early 1990s, focused on the use of diffuse near-infrared light for measuring cerebral hemoglobin oxygen saturation within neonate and adult humans utilizing a small number of point-source measurements [4–8]. It was to be several years before actual spatially resolved imaging with diffuse light (diffuse optical tomography, or DOT) would appear. For this, diffuse optical researchers needed to overcome a second obstacle—the need for a better understanding of how light propagates through highly scattering (diffusive) tissue. Today, although still an active area of research, considerable work has appeared on the theory of light propagation through scattering media in recent years (for a review see Reference [9]), and this work, along with phantom studies, has made imaging with diffuse light a reality.

Over the past 30 years since Jöbsis’s initial work, the number and diversity of optical studies has grown tremendously. Much of this growth can be attributed to increased availability of optical instruments and the recognition of several advantages to NIRS imaging, which complement other existing neuroimaging modalities, including functional magnetic resonance imaging (fMRI) with additional temporal and spectroscopic resolutions. In addition, optical experiments are typically less costly to perform than conventional functional MRI. In general, optical instruments are also smaller, less expensive to purchase, and require little to no maintenance costs. A further attraction of NIRS is that many instruments are portable, which allows experimental procedures to be conducted in a more controllable environment than the noisy and confining scanners used in MRI and several other neuroimaging modalities. This has allowed the NIRS technique to study subject populations that have been traditionally difficult to study with fMRI including infants [10–12], children [13–15], and the elderly [16–18]. This feature of optical systems also has allowed optical measurements in a wider range of study types such as those requiring movement [19], long-term monitoring [20], or studies that are sensitive to experimental conditions (e.g., noise or the potentially claustrophobia-inducing nature of the MR scanner).

The advantages of optical imaging must, of course, be weighed against the limitations of these techniques and, in particular, the current lack of anatomical imaging for localization purposes and the relatively shallow penetration depth (particularly during brain monitoring). In the remainder of this chapter, we review the basic principles of NIRS and DOT and discuss these limitations and work being done to improve these methods. We will discuss example instrumentation implementations, and introduce techniques for the analysis of optical brain studies. Finally, we summarize with a discussion of the future directions for functional optical brain imaging.

14.2. THEORY OF OPTICAL IMAGING

In comparison to traditional ray optics, the theory of diffuse optics requires several important modifications in order to account for the effects of multiple scattering events and the propagation of light through thick biological samples. Briefly, diffuse optical techniques are based on the measurement of the absorption and scattering of near-infrared light. Low levels of light are sent into tissue while detectors measure the amount of light exiting at positions away from the point of entry. The absorption and scattering properties of the underlying tissue determines the amount and distribution of the light which exits the tissue. The total absorption through such a sample can be quantified using a modification of the Beer–Lambert law adjusted for the increased distance light travels due to multiple scattering and geometry effects. The modified Beer–Lambert law (MBLL) is an empirical description of optical attenuation in a highly scattering medium [4,21], and is given by the expression

where OD is the optical density (also referred to as absorption), I_o is the incident light intensity, I is the detected light intensity, ɛ is the extinction coefficient of the chromophore, C is the concentration of a particular chromophore (e.g., an absorbing compound such as hemoglobin). L is the linear distance between where the light enters the tissue and where the detected light exits the tissue, and B is a pathlength factor that multiplies L to account for the increased effective distance traveled by light due to scattering. Finally, G is a factor that accounts for the measurement geometry and light lost at the boundaries of the sample. By convention, optical density is expressed in log base e. The calculation of the concentration of chromophores in the sample region (e.g., hemoglobin changes in the brain) requires a value for the factor B, which must include both a differential pathlength factor (DPF) and partial volume factor for each wavelength [8,22,23]. The DPF is a scaling factor transforming the source-detector separation into a measure of the effective pathlength through the tissue, which increases due to multiple scattering events. While the DPF can be measured by using frequency domain and time domain optical systems [24–26], the complexity and cost of such systems often prohibits DPF measurement for each subject prior to functional recording. Alternatively, one can calibrate the DPF against the cardiac-induced arterial pulsation [27], but this requires an accurate measurement of the cardiac pulsation within the optical signal, which is not always achievable. When using a continuous wave optical system, one most often has to rely on previously tabulated DPF values [24,28]. However, this approach suffers from uncertainties in the DPF, which may vary between individuals and tissue composition, and may change with age or anatomical changes [24,29]. Since the path traveled by light through cortical tissue varies with the wavelength-dependent scattering properties of the head and head layers, errors in the DPF correction used in MBLL can give rise to cross-talk errors in the separation of oxy- and deoxy-hemoglobin estimates. However, this cross-talk may be minimized by an optimal choice of wavelength pairs as discussed in several recent studies [30–32].

The MBLL assumes that changes in chromophore concentrations are spatially uniform over the measurement sampling volume. Thus, application of the MBLL must also account for potential partial volume factors in the value of B. For example, during functional brain imaging, changes in optical properties result from localized changes in the blood oxygenation and blood volume within the brain. This is most clearly understood by considering the geometry involved in noninvasive measurements. In an adult human, the first layer of tissue is 0.5–1 cm of scalp, followed by 0.5–1 cm of skull, followed by a thin layer (0–2 mm) of cerebral spinal fluid, and finally the gray and white matter of the brain. Functional activation changes primarily occur in the cortex, the outermost layer of the brain, which are therefore up to 2 cm below the source and detector. Thus, light traveling from the source to the detector will at best encounter the functional change over only a very small portion of the optical sampling volume, even for a spatially large activation. Such local changes violate the MBLL assumption of a global change and, thereby, introduce an error reminiscent of partial volume averaging. Typical partial volume corrections are often assumed to be approximately a factor of 40 to 60 [30], but vary between different brain regions, individuals, and functional conditions

In most functional brain studies, one is generally interested in dynamic measurements of optical absorption changes rather than on the absolute quantification of total absorption. According to the MBLL (Equation 14.1), a change in the chromophore concentration results in changes to the intensity of light passing through the sample. For such differential measurements, the extinction coefficient ɛ, and distance L remain constant. In addition, it is often be assumed that B and G are also constant. This is generally true for reasonably small changes in absorption compared to total absorption from background tissue. For measurements of changes in light intensity, equation 1 can be rewritten as

where ΔOD = OD_Final – OD_Initial is the change in optical density, I_Initial and I_final are the measured intensities before and after the concentration change, and ΔC is the change in concentration. While in principle, the measurement of absolute light attenuation would allow for the quantification of baseline levels of hemoglobin, in practice this would require precise calibration of the probe and instrument components. In particular with regard to measurements on the brain, where the coupling of light sources to the scalp, hair, and various other experimental factors make this calibration extremely difficult, absolute quantification of absorption based on the intensity of light is not currently possible. Because ΔOD is calculated from normalized light intensities, most of the calibration factors determined by the power of the instrument lasers, efficiency of fiber optics and photon detectors, and similar instrumentation issues can be ignored, and thus the voltage of the instrument photon detectors can be used in place of the light fluence (assuming that the detector voltage is linear with the level of incident light).

When multiple species of absorbing chromophores are present—as is the case of brain activation studies where both oxy- and deoxy-hemoglobin concentrations are expected to change—the MBLL is expanded to consider the contributions of each chromophore species. The total changes in optical density are expressed as a linear combination of the contributions from each change in species concentration weighted by the respective molar extinction coefficient.

where λ indicates a particular wavelength. Equation 14.3 explicitly accounts for independent concentration changes in oxy-hemoglobin (Δ[HbO₂]) and deoxy-hemoglobin (Δ[Hb]). By measuring ΔOD at two or more wavelengths and using the known extinction coefficients of oxy-hemoglobin (ɛ_HbO2) and deoxy-hemoglobin (ɛ_Hb) at those wavelengths, we can then determine the concentration changes of oxy-hemoglobin and deoxy-hemoglobin by solving the system of linear equations defined in Equation 14.3. For the common example of optical measurements at two distinct wavelengths (λ₁ and λ₂), the concentration change in oxy- and deoxy-hemoglobin can be given by

The generalization of this inverse formula for more than two wavelengths can be found in Arridge et al. [21] Figure 14.1 plots the extinction coefficients for oxy- and deoxy-hemoglobin versus wavelength as measured by Cope et al. [5,21,33] The other chromophores of significance in tissue in this wavelength range are water, lipids, and cytochrome oxidase. We do not consider these chromophores in this chapter, as their contribution in general is an order of magnitude less significant than hemoglobin, and are not easily measured without use of six or more wavelengths.

FIGURE 14.1

Absorption spectrum of oxy-hemoglobin (HbO) and deoxy-hemoglobin (Hb). Below 600 nm the absorption increases by more than a factor of 10 while above 950 nm water absorption increases significantly.

14.2.1. Photon Diffusion in Tissue

While the MBLL has been shown to work well in highly scattering media with uniform properties, it has known deficiencies in media with more complex structures (e.g., the layered structure of the scalp, skull, and brain) [23], and it does not provide a framework for reconstructing images. In order to form images, one must model where the light has gone within the tissue. For light propagation through samples that are much thicker than several times the scattering length of light (around 0.1 mm for tissue), which is the case of diffuse imaging of the human brain, a diffuse approximation is well justified. Analogous to the expression from Fick’s second law to describe a random-walk of macroscopic particles, the photon diffusion approximation is given by the equation [34–37],

$$-D{\nabla }^2\mathrm{\Phi }(r,t)+v{\mu }_a\mathrm{\Phi }(r,t)+\frac{\partial \mathrm{\Phi }(r,t)}{\partial t}=vS(r,t)$$

14.5

Ψ (r, t) is the photon fluence at position r and time t. The photon fluence is defined as the number of photons intersecting a differential area and is proportional to the intensity of light. S(r ,t) is the source distribution of photons. D = v / (3μ_s′) is the photon diffusion coefficient [38,39], μ_s′is the reduced scattering coefficient, μ_a is the absorption coefficient, and v is the speed of light in the medium. For a combination of the hemoglobin chromophores, absorption coefficient is defined by the spectrally weighted contribution of each absorbing species, i.e.,

$${\mu }_a={\varepsilon }_{Hb{O}_2}\left[Hb{O}_2\right]+{\varepsilon }_{Hb}\left[Hb\right].$$

14.6

Equation 14.5 accurately models the migration of light through highly scattering media, provided that the probability of scattering is much greater than the absorption probability. Solutions of the photon diffusion equation can be used to predict the photon fluence detected for typical diffuse measurements [37].

Assuming only a small perturbation in absorption, the analytic solution to Equation 14.5 for measurements of changes in optical fluence is only known for a limited number of simple geometries. One of the simplest of these geometries is the semi-infinite homogenous geometry, which is defined by a medium that has only one surface. In practice, this geometry approximates many objects whose boundaries are located sufficiently far enough away from the imaging volume. Assuming that hemoglobin concentration changes are both global and small, the solution of the photon diffusion equation for a semi-infinite medium is

$$\mathrm{\Delta }OD=\frac{1}{2}\sqrt{\left(\frac{3{\mu }_s^{\prime }}{{\mu }_A^{Initial}}\right)}\left(1-\frac{1}{1+L\sqrt{\left(3{\mu }_S^{\prime }{}^{Initial}{\mu }_A^{Initial}\right)}}\right)\mathrm{\Delta }{\mu }_{A}L.$$

14.7

The solution of the photon diffusion equation for the representative semi-infinite tissue geometry (Equation 14.7) tells us that the modified Beer–Lambert law is reasonable for tissues with spatially uniform optical properties when the chromophore concentration does not change significantly (i.e., Δ[X]/[X]<< 1). Equation 14.7 also explicitly defines the pathlength factor B for a semi-infinite medium, which was introducedin Equation 14.3, as

$$B=\frac{1}{2}\sqrt{\left(\frac{3{\mu }_s^{\prime }}{{\mu }_A^{Initial}}\right)}\left(1-\frac{1}{1+L\sqrt{\left(3{\mu }_S^{\prime }{}^{Initial}{\mu }_A^{Initial}\right)}}\right)$$

14.8

This shows that B depends on tissue scattering, initial chromophore concentration, extinction coefficient (and thus B is wavelength dependent), and optode separation. In practice, the validity of the assumption that B is independent of μ_a and L has often been ignored since B is in general empirically determined and/or the changes in μ_a are typically small.

In Figure 14.2, we show the sensitivity profile for a simple reflectance geometry with a continuous-wave (CW) light source using typical tissue optical properties of μ_s ′ = 10 cm⁻¹ and μ_a= 0.1 cm⁻¹. The most important point to notice is that the spatial sensitivity profile is not localized but covers a large volume and therefore intrinsically has relatively poor spatial resolution. Second, notice that as the separation between the source and detector increases, the penetration depth also increases. This fact can be used to probe deeper into the brain and possibly even to distinguish superficial scalp from deeper brain signals.

FIGURE 14.2

Photon migration sensitivity profile shown for a semi-infinite medium for a 2 cm and 4 cm source-detector separation. Contour lines are shown every half order of magnitude.

14.2.2. Theoretical Optical Sensitivity to the Brain

The diffusion approximation to the optical forward problem as described in the previous section may be used to model the propagation of light through tissue. By using numerical solvers to Equation 14.5 for finite element meshes or by using Monte Carlo techniques [28,35,36,40–42], we can simulate the propagation of photons through tissue and obtain a spatial sensitivity map for a given grid of source and detector measurement combinations. For example, in our Monte Carlo simulations, individual photon trajectories are traced by sampling appropriate probability distributions for scattering events [41]. After 10⁶–10⁸ photons are traced, quantities such as the fraction of photons reaching a particular detector or the spatial sampling of the tissue can be determined. In our Monte Carlo model, the tissue is a 3D volume with optical properties assigned to each voxel, and source and detectors can be placed at any voxel. Photons are propagated until they exit the tissue, and the total path length in each tissue type is recorded for all photons reaching a detector. Only recently has the computational power become widely available for the practical use of Monte Carlo simulations (one simulation typically takes 3–4 hours on current high-end desktop systems).

Figure 14.3 shows an anatomical MRI of a human head, segmented into five tissue types (air, scalp, skull, cerebral spinal fluid, and gray/white matter; see Reference [43]), with a contour overlay indicating the photon migration spatial sensitivity profile for (a) continuous-wave, (b) 200 MHz modulation, and (c,d) pulsed measurements. One contour line is shown for each half order of magnitude (10 dB) signal loss, and the contours end after 3 orders of magnitude in loss (60 dB). For 3D Monte Carlo simulation, we assumed that μ_s ′ = 10 cm⁻¹ and μ_a = 0.4 cm⁻¹ for the scalp and skull, μ_s′ = 0.1 cm⁻¹ and μ_a = 0.01 cm⁻¹ for the CSF, and μ_s ′ = 12.5 cm⁻¹ and μ_a = 0.25 cm⁻¹ for the gray/white matter. Note how the contours extend several millimeters into the brain tissue, indicating sensitivity to changes in cortical optical properties. The depth penetration difference between the continuous-wave and 200 MHz measurements is difficult discern. A ratio of the two sensitivity profiles (not shown) shows that the 200 MHz profile is shifted slightly toward the surface. The time-domain sensitivity profiles suggest the possibility of obtaining greater penetration depths in the head from measurements made at longer delay times.

FIGURE 14.3

Contour plots of the photon migration sensitivity profile in a 3D human head as determined from Monte Carlo simulations. Details provided in the text.

14.2.3. Diffuse Optical Imaging and Tomography

It is a relatively simple step, conceptually speaking at least, to move from the simple NIRS point-measurement to a spatially resolved imaging of these same variables. In a typical diffuse optical imaging study of brain function, a grid of light sources and light detectors is positioned on the head of a subject as pictured in Figure 14.4. Often this is done using fiber optics where one end of these fibers, which we refer to as an optode, is secured into a head gear device, and the other end of the fiber is connected to the imaging instrument. The arrangement of the optodes on the head determines the depth-sensitivity of the measurements to underlying cerebral changes, which is a function of the distance between the source and detector pair as shown in Figure 14.2. In addition, the arrangement of optodes will also dictate the lateral spatial resolution. In the case of true tomographic imaging (DOT), a dense array of optodes and overlapping measurement combinations of the sample volume are used to provide uniform spatial coverage and achieve the best spatial resolution [31,44,45]. However, optical tomography is generally more experimentally difficult than less dense array probes because of the number of optodes involved—requiring additional setup time for adjustment of the probe and quality assurance of the signals—and the range of light intensities that need to be recorded at the detector optodes. In a tomographic probe as shown in Figure 14.4b, light from sources located various distances away must be recorded by a single detector. Since the intensity of light reaching the surface decays exponentially as the detector moves away from the source, several orders of magnitude influence will be reduced between a near and far source-detector spacing. For example, there is approximately a 50-fold intensity difference between light from a 1.5 cm and 3 cm spacing [46]. The collection of overlapping measurements requires either detectors with a high dynamic range (e.g., Reference [47]) or source-switching schemes where only a fraction of the sources are turned on at any one time (e.g., References [45,48]).

FIGURE 14.4

Probe arrangement for NIRS and DOT experiments. Panels (a) and (c) show a nearest-neighbor NIRS probe. Panels (b) and (d) show a tomographic-style (DOT) probe. The sensitivity profiles for these two probes are shown in panels (c) and (d) with contour (more...)

In general, optical imaging—as opposed to tomography—is more common in brain activation studies. Unlike optical tomography, diffuse optical imaging does not require overlapping sets of measurements and, thus, the design and implementation of these experiments is somewhat easier with a compromise of less uniform spatial resolution and lower resolution. Optical imaging typically uses a grid of optodes arranged to collect nearest-neighbor combinations of sources and detectors. However, a nearest-neighbor probe geometry (Figure 14.4a) may often contain of areas of low measurement sensitivity. In particular, optical measurements are most sensitive in an area immediately between the source and detector positions and are less sensitive to changes occurring immediately beneath these optode positions as shown in Figure 14.4. The choice between a point-source measurement, imaging, or true tomography probe design depends on the application and specific hypothesis of the experiment. There are several practical trade-offs to consider when designing this probe, not the least of which is the additional setup time required to position and adjust the probe on the head of a subject.

Whether optical measurements are collected from a true tomography or imaging geometry, the image reconstruction theory and procedure is very similar (reviewed in References [9,49]). In order to generate an image of spatial variations in Hb and HbO₂, in principle, all we need to do is to measure the photon fluence at multiple source and detector positions. Each measurement samples a different volume of underlying tissue and can be modeled using photon migration theory discussed in the previous sections of this chapter. This set of measurements can then be back-projected to generate the image of underlying concentration changes. The general solution of the photon diffusion equation at the detector position r_d for a medium with spatially varying absorption is

$$\mathrm{\Phi }\left(r_d\right)={\mathrm{\Phi }}_{incident}\left(r_d\right)-\int \mathrm{\Phi }\left(r\right)\frac{v}{D}\left({\varepsilon }_{Hb{O}_2}\mathrm{\Delta }\left[{\text{HbO}}_2\right]+{\varepsilon }_{Hb}\mathrm{\Delta }\left[\text{Hb}\right]\right)G\left(r,r_d\right)\text{d}r.$$

14.9

For reflectance in a semi-infinite medium, the solutions for Φ_incident and G(r, r_d) can be found in Farrell [50]. Equation 14.9 is an implicit equation for the measured fluence as Ψ (r) appears in the integral on the right-hand side. One way to solve this equation uses the first Born approximation, which assumes that Ψ (r) = Φ_incident + Φ_sc [51]. In this approximation, the measured fluence is given by

$${\mathrm{\Phi }}_{sc}\left(r_d\right)=-\int {\mathrm{\Phi }}_{incident}\left(r\right)\frac{v}{D}\left({\varepsilon }_{Hb{O}_2}\mathrm{\Delta }\left[{\text{HbO}}_2\right]+{\varepsilon }_{Hb}\mathrm{\Delta }\left[\text{Hb}\right]\right)G\left(r,r_d\right)\text{d}r.$$

14.10

By measuring Φ_sc(r_d ) using multiple source and detector positions and multiple wavelengths, we can invert the integral equation to obtain images of Δ[HbO₂] and Δ[Hb]. Equation 10 defines the contribution of each position in the underlying volume to the measurement made between a particular source and detector pairing and defines the measurement sensitivity for each measurement on the optode grid. The process of image reconstruction from such measurements involves the inversion of Equation 14.10 in order to solve for the changes in oxy- and deoxy-hemoglobin at each position in the sample volume.

For small changes in absorption, the optical trajectories may be assumed to remain constant and thus the hemoglobin terms may be moved outside of the integral in Equation 14.10. Many imaging procedures relay on reducing Equation 14.10 to a matrix formulation by rewriting this spatial integral as a discrete sum over the volume elements (voxels) in the desired image, i.e., Y = A · x where Y is the vector of measurements (i.e., Φ_sc(r_d )), x is the vector of image voxels, and A is the transformation matrix obtained from the integrand of Equation 14.10. In the case of fewer measurements than unknowns, the linear inverse problem is underdetermined and, thus, cannot be uniquely solved [49]. In addition, this inverse problem is ill-posed, which means that multiple equivalent solutions to the inverse problem may exist. The optical inverse problem requires regularization. The usual form of the regularized optical inverse problem is given by the (regularized) Moore–Penrose generalized inverse [9]

$$x=-A^T{\left(A{A}^T+\lambda I\right)}^{-1}·Y,$$

14.11

where I is the identity matrix, and λ = α max(AA^T) is termed the Tikhonov regularization parameter. Y is the optical density measurements and x is the estimated absorption changes in the image. Note that a weighted back-projection (x = −s ·A^T ·Y) is an approximation to Equation 14.10 in that the matrix product A · A^T approaches an identity matrix

Figure 14.5 illustrates the method of reconstructing an image with diffuse light. The process is conceptually similar to that of x-ray computed tomography, which is included for comparison. In the case of x-ray tomography, a collimated beam transilluminates the tissue, maintaining high spatial resolution, and the attenuation of the beam is measured. The measured attenuation is then back-projected into the tissue along the axis of the x-ray. For a simple object positioned in the center of the plane, we see that a single x-ray projection sharply defines the lateral boundaries of the object, but does not localize the object along the x-ray propagation axis. Through the combination of additional measurements at different angles with respect to the object, it is possible to obtain a high-resolution image of the x-ray-attenuating object within the tissue. Figure 14.5B shows the result from two orthogonal projections. The artifacts near the boundaries are reduced with additional projections. Figures 14.5C,D show the results obtained with diffuse light by following the same back-projection procedure. The striking difference obtained with a single projection is that the boundaries of the object are not sharply defined, and the object appears to be near the sources and detectors where the sensitivity of the measurement is greatest. The first observation is an intrinsic limitation of imaging with diffuse light, while the second observation is an artifact of the back-projection procedure. Utilizing an additional orthogonal projection (Figure 14.5D), we see better localization of the absorbing object. We note that while the back-projection method provides quantitative images with x-ray CT, it is only an approximation for diffuse optical tomography. This difference results from the fact that the measured diffuse light has sampled a volume of tissue as opposed to the measured x-ray that samples only along a line through the sample. The back-projection method, therefore, does not produce an image that optimally fits the experimental data. More sophisticated optimization is needed to obtain quantitative image reconstruction with diffuse light [49].

FIGURE 14.5

Comparison of x-ray computed tomography (A,B) and diffuse optical tomography (C,D) of a square absorbing object in the center of the outer square. X-ray tomography has greater resolution as indicated by the sharper edges of the reconstructed image. Diffuse (more...)

14.3. INSTRUMENTATION FOR BRAIN STUDIES

For brain imaging experiments, several technical solutions exist for implementation of NIRS and DOT, including time domain (TD), frequency domain (FD), and continuous-wave (CW) systems. The principle difference between these methods is the characteristics of the incident light used and subsequently the type of detectors. TD systems [2,9,52–54] introduce into tissue extremely short (picosecond) incident pulses of light. These narrow pulses are broadened and attenuated by the various tissue layers such as skin, skull, cerebrospinal fluid (CSF), and brain. A TD system detects the temporal distribution of photons as they leave the tissue, and the shape of this distribution provides information about tissue scattering and absorption values. The collection of time-of-fiight information from TD systems allows for the discrimination absorption changes from different depths. For example, deeper absorption changes affect the tail of the temporal distribution function representing the photons that have traveled for the longest duration in the tissue. By modeling this temporal point-spread function, TD methods can be used to increase sensitivity to deeper changes [55]. In principle, TD systems can theoretically obtain the highest spatial resolution and can accurately determine absorption and scattering. However, drawbacks of TD measurements include long acquisition times to achieve reasonable signal-to-noise ratios, a need to mechanically stabilize the instrument, and the large dimensions and high cost of the necessary ultrafast lasers [9]. In addition, TD methods require acquisition of data from multiple temporal windows in order to reconstruct the temporal distribution of light exiting the tissue; thus, the overall acquisition rate is inversely related to the number of time-gates measured and consequently the depth resolution. Although TD instruments have advanced in recent years [56], their cost and complexity has still generally limited their widespread use.

Another implementation of NIRS uses frequency domain (FD) signals [57–60]. In FD systems the light source is on continuously, but amplitude-modulated at frequencies on the order of tens to hundreds of megahertz. Information about the absorption and scattering properties of tissue are obtained by recording amplitude decay and phase shift (delay) of the detected signal with respect to the incident light [61]. State-of-the-art FD systems cover a wide range of applications for clinical use, and can achieve considerably higher temporal resolution than TD systems. FD instruments are typically less expensive to build than TD systems, but are still more so than CW systems, and they require more careful engineering and optimization to eliminate noise, ground loops, and RF-transmitter effects. The detection of FD light requires sensors with high dynamic range, and such systems typically require photo multiplier tubes (PMT) detectors. In recent years, FD systems have been developed commercially (e.g., the Imagent system from ISS Inc.), and this has allowed these methods to be more readily accessible.

Finally, in CW systems [62–64], light sources emit light continuously as FD systems do, but at constant amplitude or modulated at frequencies not higher than a few tens of kilohertz. In CW setups, only the amplitude of the light is detected. Thus, the primary shortcoming of a CW system compared to TD and FD systems is the inability to uniquely quantify the effects of light scattering and absorption [65]. While CW imaging techniques have the potential to provide quantitative images of hemodynamics changes during brain activation, quantitative imaging of baseline brain states is likely to only be possible with FD and TD methods, as they provide both time-delay (e.g., optical pathlength) and amplitude information. However, CW technology can be engineered with relatively inexpensive and widely available components; a hospital pulse-oximeter being an example for the small dimensions achievable for a CW NIRS instrument. There are currently a number of commercially available such systems (Hamamatsu: the NIRO 500, and the INVOS 3100 systems; NIRx Medical technologies: the Dynot system; Hitachi Medical Corporation: the ETG Optical Topography System; and Techen: the NIRS2 and CW5 systems). The commercialization of these instruments has made this technology more available to a variety of researchers. It is likely that CW technology will quickly find widespread use among brain researchers because of its relatively low cost, portability, and ease of implementation and use compared to FD and TD systems, while still displaying sensitivity to cerebral hemodynamic features. Arguably, a CW imaging system will also achieve the best signal-to-noise ratio at image frame rates faster than 1 Hz and up to several hundred Hz. In the following two sections, we will describe two examples of basic instrument design of CW NIRS and DOT systems from our lab.

14.3.1. NIRS Instrument

A basic NIRS system delivers light of preferably two or more wavelengths to a single position on the tissue, and then collects and detects the diffusely reemitted light at one or more locations. Because the light must travel several centimeters through the scalp and skull to sample the adult human brain, it is necessary that light sources be chosen with sufficient power, and in the wavelength range between 600 and 950 nm where light absorption is minimized. This typically requires the use of laser diodes although filtered white light sources have also been successfully used in adult humans [27]. For safe, long-term tissue exposure to laser light, the optical power incident on the tissue must be no more than 4 mW per 1 mm [2,66]. To gain sensitivity to the brain in the adult human, detectors must be placed at least 2.5 cm from the source (less for babies, with their thinner scalps and skulls), as the depth of sensitivity is roughly proportional to the source-detector separation. With such separations, the amount of light reaching the detector is on the order of 10 pico-Watts (an attenuation of 9 decades). To obtain a decent signal-to-noise ratio at desired bandwidths of 1 Hz or greater, therefore, it is necessary to use high-sensitivity detectors: photo-multiplier tubes (PMT’s) or avalanche photodiodes (APDs), or CCD cameras.

Figure 14.6 shows a photograph of our CW NIRS system. The sources in this system are two low-power laser diodes emitting light at discrete wavelengths, typically 690 nm (Sanyo, DL7140–201) and 830 nm (Hitachi, HL8325G). These lasers are powered by a stabilized current, intensity-modulated by an approximately 5 kHz square wave at a 50% duty-cycle. Both diodes are driven at the same frequency, but phase shifted by 90º with respect to one another. This phase encoding—known as an in-phase/quadrature-phase (IQ) demodulation circuit—allows simultaneous laser operation as well as separation of the contributions of each source to a given detector’s signal. The system has four separate detector modules (Hamamatsu C5460-01), which are optically and electrically isolated from each other. Each module consists of a silicon avalanche photodiode (APD) with a built-in high-speed current-to-voltage amplifier, and temperature-compensation. They achieve a gain of typically 10⁸ V/W with a noise equivalent power of 0.04 pW/Hz resulting from the high detector sensitivity. The incoming signal at a given detector is composed of both source colors, which are separated by synchronous (lock-in) detection using the two source signals as a reference. The output of the decoding portion of the IQ circuit includes two signal components, corresponding to the two laser wavelengths. These two components are low-pass filtered at 20 Hz and digitized by a computer at 200 samples per second.

FIGURE 14.6

Photograph of the NIRS system and cap used to hold the fiber optics on the head.

14.3.2. Imaging Instrumentation

Conceptually, an imaging instrument is a simple extension of a NIRS system to include more sources and detectors. Adding more detectors is straightforward as each detector operates independently without interaction with the other detectors. The sources, on the other hand, all introduce light into the tissue, and detectors can only detect the total amount of light present at a given time. Thus, in order to properly assign light received at a detector to each individual source, the source light must be encoded. Several encoding strategies exist: time-sharing, time-encoding, and frequency-encoding. Time-sharing is the easiest to implement. In this scheme, each source is turned on, one at a time, long enough for the detectors to acquire a decent signal (typically 10–100 ms). Large numbers of sources gives rise to slow image frame rates and temporal skew between sources. Time-encoding overcomes the temporal skew problem by switching between sources at a much faster rate (100 μs or faster) but integrating the signal from each source over several switch cycles (e.g., the same 10–100 ms), resulting in roughly the same image frame rate as with time-sharing. Finally, frequency encoding turns all sources on at the same time but modulates the intensity of each one at a slightly different frequency. The individual source signals are then discriminated at the detector by employing simple analog or digital band-pass filters at the frequencies of source modulation. While the duty cycle in this scheme is maximized, the effective dynamic range of the system is limited by the fact that all sources are on at the same time. That is, it can be difficult to detect a distant source that is several orders of magnitude weaker than a source close-by given that each detector has a fixed dynamic range (typically 3–4 orders of magnitude).

Figure 14.7 shows a block diagram and photograph of our DOT imaging system with 32 lasers and 32 detectors. The detectors are avalanche photodiodes (APDs, Hamamatsu C5460-01). A master clock generates the 32 distinct carrier frequencies between 4.0 kHz and 7.4 kHz in approximately 200 Hz steps. These frequencies are then used to drive the individual lasers with current stabilized square-wave modulation. Following each APD module is a band-pass filter, cut-on frequency of ~500 Hz to reduce 1/f noise and the 60 Hz room light signal, and a cut-off frequency of ~10 kHz to reduce the third harmonics of the square-wave signals. After the band-pass filter is a programmable gain stage to match the signal levels with the acquisition level on the analog-to-digital converter within the computer. Each detector is digitized at ~40 kHz and the individual source signals separated using an IQ demodulation scheme against the carrier frequency of each laser.

FIGURE 14.7

Photograph of the optical tomography system. Details provided in the text.

14.4. EXPERIMENTAL DESIGN OF OPTICAL STUDIES

In the reminder of this chapter, we will discuss applications of NIRS and DOT instruments. Over the past several years near-IR optical methods have been applied to study a variety of cerebral regions including the visual [67–69], auditory [70–72], and somatosensory cortices [73,74]. Other areas of investigation have included the motor [75–78], prefrontal/cognitive [79–82], and language systems [83]. In addition to its use in brain mapping experiments, optical imaging is beginning to be applied to study several clinical problems, in part motivated by the portability and bedside applicability of optical instruments. Clinically oriented applications have included the prevention and treatment of seizures [84–89] and Alzheimer’s disease [18,90,91], and provided a potential tool to monitor stroke rehabilitation [92–99]. NIRS has also been applied to psychiatric health concerns such as depression [79,100–104] and schizophrenia [104–109].

Several groups have also investigated the possibility of using optical methods to measure the cytochrome oxidase (cytox) redox state [1,110]. Cytox oxidation is a marker of intracellular energy metabolism, and it has been suggested as a possible parameter to assess the functional state of the brain [27,111–114]. These optical measurements are usually conducted over a wide spectral range (extending up to 500 nm), because the absorption contributions of cytox are an order of magnitude (or more) smaller than those of hemoglobin.

Lastly, a few groups have also worked on using diffuse light to directly measure neuronal activity, rather than indirectly via hemoglobin. For example, it has been shown in vitro that neuronal activity is associated with an increase in light scattering, induced by a change in the index of refraction of the neuronal membranes [115–117]. Following up on this work, near-infrared optical methods have been successfully used to detect such light-scattering changes in vivo in adult humans [118–120]. This fast optical signal appears to show the same time course as the electrophysiological response measured with EEG/electrophysiological techniques.

Conceptually, the design of an optical imaging brain experiment study is similar to many other imaging modalities (particularly functional magnetic resonance imaging; fMRI). There are two primary things to consider in designing an optical study: the expected characteristics of the activation signal and the specific hypothesis that will be addressed by the study. More so than in fMRI experiments, optical studies generally begin with some foreknowledge of the location and extent of the expected activation signals. It is particularly important when designing an optical study to consider the expected depth of the activation signal as this determines the type of optical method used and the arrangement of the head probe. Although, in general optical measurements are limited to the outer layers (approximately 5–8 mm) of the brain, this depth sensitivity may be adjusted based on the source-to-detector separation distance and on the characteristics of the underlying tissue (see Figure 14.2). For example, in the visual cortex, where both bone and venous blood vessel densities are high, optical penetration is lower and the signal strength at a given distance may be lower requiring closer spacing to achieve a given ratio of signal-to-noise. In contrast, cognitive studies of brain activity in the frontal lobe may need to consider the proximity to sinus cavities and underlying cerebral spinal fluid layers, which results in a lower attenuation of optical signals, and allow for larger optode separation distances at similar signal-to-noise ratios. These factors should be noted when designing an optical probe for such a study.

A second consideration in the design of optical experiments is the specific hypothesis that will be addressed. In comparison to the majority of fMRI based studies, which are based on the amplitude of the evoked response, optical measurements have several unique characteristics that are factors in designing an experimental hypothesis. First, optical signals generally have high temporal resolutions (often far exceeding the typically slow dynamics of the hemodynamic activation response). This characteristic allows experimenters to develop hypotheses based on the temporal dynamics of the evoked signal. Examples of this approach include examination of differences in the early or late phases of the evoked response between subject populations or tasks (e.g., see Reference [121]). A second unique characteristic of optical measurements is the added spectroscopic information. This allows hypotheses based on relative changes, for example, the relationship of oxy- and deoxyhemoglobin. This information may be used to (at least qualitatively) infer the influence of flow verses oxygen metabolism effects (e.g., see References [122,123]). In contrast, for the majority of fMRI studies, hypotheses are formulated to test for differences in the amplitude or spatial characteristics of the response rather then these timing differences. Although this can also be done in optical imaging, due to the generally low spatial resolution and depth sensitivity of optical measurements, the amplitude of optically measured evoked response may be distorted by partial volume errors [31]. Thus, particular care must be taken to spatial maps between conditions or in drawing conclusions about the amplitude of evoked responses, since conditional differences in the depth or location of the activation may offer alternative explanations of such results. As discussed in the prior section, the design of the optical experiment—for example, the choice between a NIRS versus a DOT (tomography) study—will determine the sensitivity, reproducibility, and accuracy of the results.

A final unique consideration to the design of optical studies is the potential contamination from background physiological signals. Since optical measurements are made through the superficial layers of skin on the head, these measurements are sensitive to systemic fluctuations. Further discussion of the sources of this noise and analysis methods for reducing their impact will be discussed later in this chapter. However, in designing an optical study, it is important to also consider the possibility for systemic effects of the stimulation task. For example, does heart rate change when the subject performs the task? In addition, we have observed that systemic fluctuations may occasionally be influenced by the design of the optical experiment. For example, blocked-design experiments, which present stimuli at regular time intervals, may exhibit coupling between physiological oscillations (Mayer waves) and stimulus presentation. The presence of these systemic oscillations may depend on subject posture, breathing rate, or simply reflect individual variations in physiology [124]. Additionally, the performance of difficult tasks, such as motor activity, may change this physiology and produce systemic evoked responses, which may be confused with global brain activity—for example, if subjects unintentionally hold their breath while performing a brief finger-tapping exercise or the task evokes a startled response (i.e., raising heart rate). In Figure 14.8, we show an example of stimulus-evoked increases in heart rate that were recorded during a 5-s finger-tapping exercise. This heart-rate increase, which is roughly a 0.5% change, has approximately the same timing and duration as the expected cerebral hemodynamic signals, and although these changes are smaller than the focal brain activation changes—usually on the order of a few percent change in optical density—these observations highlight one of the difficulties in optical measurements and the ongoing need for advancement in data analysis methods to deal with these specific aspects of optical.

FIGURE 14.8

Example of a systemic “functional” response recorded from a finger clip pulse-oximeter during a 2-s finger-tapping task. The pulse-rate was recorded on the stationary hand during the motor task. In addition to the cortical activity involved (more...)

Aside from the filtering and analysis methods, which will be discussed later in this chapter as methods to reduce the impact of this physiology, in experiments where systemic physiology is believed to pose a confounding problem, it is advisable to collect additional information about systemic changes. For example, heart rate, respiration, and blood pressure can be measured using auxiliary devices to characterize systemic changes. Alternatively, NIRS measurements from short source-detector separations can be used to measure local fluctuations in the skin layers above the brain activation region.

In the following sections, we will present examples of both NIRS and DOT experiments to emphasize these experimental considerations.

14.4.1. Design Example 1: NIRS Baseline Recording

Many efforts have been made to better understand background physiological fluctuations with hopes of improving the data analysis methods to remove these nuisance variables from functional brain mapping studies. Figure 14.9 shows the observed amplitude modulations from a single subject during two different baseline conditions (in optical density units). The top trace—with data gathered at 830 nm from approximately location C3 in the international 10–20 system—shows 40 s of eyes-closed baseline while the subject was seated upright. The bottom trace is from the same subject, again eyes closed baseline but while lying supine. Heart pulsations are clear in both records, while in the upright case an additional high-amplitude oscillation of ~0.1 Hz appears. The frequency of this periodicity corresponds to the Mayer wave—a systemic blood pressure oscillation that is more prominent when standing or sitting than when lying down [27,125]. In recent years, several studies of resting state (baseline) physiological fluctuations have turned away from considering these fluctuations as simply noise and have begun to investigate information about cerebral physiology within these baseline signals. For example, Franceschini et al. [124] have used optical measurements of resting physiology to examine asymmetry between hemispheres. In addition, several groups have begun to use NIRS to explore normal vascular physiology, such as vasomotion or autonomic regulation [124,126–128]. We believe that these approaches will be continually developed in future studies and offer great promise for the use of optical imaging in a clinical setting.

FIGURE 14.9

Results from 40 s of recording during an eyes-closed resting condition in a subject while in a sitting position (top) and supine (bottom). Optical density changes observed from the 830-nm laser are shown. In both traces, heart pulsations are clearly evident. (more...)

14.4.2. Design Example 2: Functional NIRS

While studies of spontaneous physiology are beginning to appear, to date the majority of NIRS and DOT studies have examined evoked hemodynamic responses to cued functional tasks. In Figure 14.10, we show an example of an evoked hemodynamic response to a train of pulses (4 Hz for 1-s duration) measured using optical imaging during mild electrical stimulation of the median nerve. For the MBLL analysis we assumed that the differential path-length factor was equal to 6 at each wavelength and a partial volume correction of 60 was used. Figure 14.10a shows the spatial distribution of oxy- (red), deoxy- (blue), and total-hemoglobin (green) changes based on the optical measurements from the probe shown. Figure 14.10b shows the evoked response to a single stimulation and a train of stimuli from region-of-interest analysis (p < 0.05) on seven subjects. These responses show the typical hyperemic (increased blood flow) response that is often observed for functional responses. This response corresponds to the typical “positive BOLD” signal observed in fMRI. In this response, both oxyhemoglobin and total-hemoglobin are observed to increase due to increased regional blood flow and volume in the activated brain region. These changes are the result of neurovascular coupling— the action of neurons and neurotransmittors to dilate blood vessels around the brain activation site (reviewed in Reference [129]). Increases in arterial blood flow causes deoxyhemoglobin levels to decrease—a phenomena termed the washout effect. Increased neuronal activity is also associated with increased cerebral oxygen metabolism (CMRO₂). Although an increase in CMRO₂ removes more oxygen from the blood—converting oxyhemoglobin into deoxyhemoglobin—the increases in blood flow generally overcompensate for this effect, and the net result is a usual increase in blood oxygen saturation (for a review see Reference [130]). Figure 14.10 shows this typical hyperemic response from a median nerve electrical stimulation study.

FIGURE 14.10

The hemodynamic response to train of pulses (4 Hz for a 1-s duration) to the median nerve measured using NIRS. Panel (a) and (b) show the spatial distribution of hemodynamic responses and region-of-interest average (p < 0.05 defined for oxyhemoglobin). (more...)

14.4.3. Design Example 3: Tomographic Optical Imaging

In Figure 14.11, we show results from a functional brain study using a tomographic probe. In this study by Joseph et al. [45], an overlapping (tomographic) optical probe was used to record optical signals from the motor cortex in subjects as they preformed a brief finger-tapping task. In this figure, we show reconstructed images from the first nearest-neighbor, second-nearest-neighbor, and overlapping sets of measurements. The overlapping (tomographic) measurements provide the most uniform spatial coverage and most accurately reproduce the spatial maps of brain activation when compared to fMRI.

FIGURE 14.11

In this figure, we show a time-multiplexing scheme for of our optical tomography system (panel A). The system switches between three combinations of light source and detectors arranged on the head (panels B and D). This provides a more uniform spatial (more...)

14.5. ANALYSIS AND INTERPRETATION OF OPTICAL STUDIES

While conceptually very similar to other methods like functional MRI and EEG, the analysis of optical signals requires several additional considerations of some of the unique features of this data, including signal processing.

14.5.1. Methods for the Reduction of Motion

Subject motion can be a detrimental source of noise in the NIRS analysis. While the best opportunities to deal with such motion artifacts occur during the data acquisition and reflect subject compliance, experimental design, and the experimenter’s expertise in positioning and securing the head probe, motion artifacts will undoubtedly still affect many experiments. Motion is particularly common in dealing with infant or animal subjects, where subject compliance is more difficult to control. In general, motion artifacts appear as large jumps in the amplitude of optical signals and are often several orders of magnitude larger then the measurement variance expected from other physiological fluctuations within the head including activation. This feature can be used to help detect regions affected by motion artifacts by simple visual inspection or outlier analysis (for example, by examination of the studentized residual error of the data after functional analysis or using a similar regression analysis model). Subject movement may also move the head probe causing a change in the optode couplings to the head. This can be a major issue since it requires a renormalization of the measured signal or, in a worse case, could mean that measurements represent a different brain region. These errors can greatly affect the quantitative accuracy of optical experiments. Once identified, temporal regions affected by motion artifacts can be excluded from further analysis by either removing these time points from the model used for the functional analysis, or by excluding the entire data file or subject. Filtering methods such as autoregressive models or adaptive filtering methods such as Weiner or Kalman filters have also been suggested for reducing the impact of motion artifacts [131,132]. Since many of the subsequent response estimation, filtering, and other signal processing methods will not work properly if motion is present in the data, motion correction is often done as a early step in data analysis. However, it is important to have an objective criterion for removing data from a study.

Since motion artifacts arise from the physical displacement of the optical probe from the surface of the subject’s head, motion artifacts are statistically covariant across multiple source-detector pairs and independent of the optical wavelength. Such a motion artifact will typically demonstrate a large change of signal in all or at least a large subset of the source-detector channels. The use of principal component analysis (PCA) to remove motion artifacts was first demonstrated by Wilcox and coworkers where it was shown to reduce motion artifacts in a study of infant brain function [133]. The principal components of the data covariance are calculated using singular value decomposition. Since motion artifacts will typically have a highly covariant data structure, motion artifacts may be captured in the first (strongest) eigenvector components using PCA. This allows these artifacts to be subtracted from the time-series of the optical density measurements. Using this filter, motion artifacts can be removed from data by reducing the spatial covariance of the set of measurements [133]. An example of this PCA filter is demonstrated in Figure 14.12. Here, we show an example data set taken as part of a NIRS study on object recognition in infants [133]. The raw data, shown in Figure 14.12A (top) demonstrates very strong motion artifacts, which dominate the optical signals. In addition, these motion artifacts were covariant across the majority of the NIRS probe, making analysis of this data virtually impossible. Using this PCA filter to remove the first and second principal components, the motion artifact is greatly reduced (shown in panels B and C).

FIGURE 14.12

In this figure, we show an example of motion artifact removal by principal component (PCA) filtering. The raw data shown in panel (A, top) is experimental data measured in the inferior temporal cortex in an infant. This data was dominated by several motion (more...)

This motion correction algorithm exploits the property of a motion artifact to have an intense and correlated effect over a large area of the NIRS probe. This is a general characteristic of motion, but not entirely encompassing of all motion artifacts. This algorithm works best for reducing motion that affects areas much larger than the functional region. For a small number of source-detector pairs or a probe with a spatial coverage comparable to the extent of the functional region, principal component analysis can negatively affect the estimate of the hemodynamic response. To ensure that this does not occur, statistical effects analysis can be used to estimate the number of components that maximize the probability of functional activity.

14.6. METHODS FOR THE REDUCTION OF PHYSIOLOGICAL NOISE

Once optical signals have been checked and corrected for motion artifacts, there will still be several other sources of noise in the data. Physiological systemic noise arising from the cardiac cycle, respiration, systemic blood pressure, and Mayer wave fluctuations are a common component of signal interference in NIRS measurements [31,119,120,124,127,134,135]. This is enhanced by the oversensitivity of NIRS to superficial tissue. In fact, the absence of cardiac and other fluctuations can be a sign that data quality is low (perhaps due to poor coupling of the optical probe with the head) and often indicates that functional analysis will not be possible. Background physiological signals have been explored and may reveal interesting details about vascular physiology [136]. However, in most studies, attempts are made to reduce background physiological signals, and several techniques will be discussed in the following sections.

14.6.2. Principal Component Analysis

Principal component analysis has also been used in optical studies to remove systemic physiology based on the spatial and temporal structure of these signals as first described by Zhang and coworkers [135]. Similar to motion artifacts, systemic fluctuations are covariant between the NIRS measurements from different regions of the head. Thus, the reduction of this covariance can be used to filter physiology of systemic origin from the measurements. In comparison to the motion correction, for this PCA filter, the spatial covariance may be calculated from a separate baseline (rest state) data set, rather than from the functional data itself. By using a separate baseline time course to derive the spatial covariance associated with the systemic physiology, this allows better distinction from the evoked hemodynamic functional response. As this systemic signal interference causes different variation in oxy- and deoxyhemoglobin, the spatial covariance should calculated independently for each hemoglobin species or from each optical wavelength.

In Figure 14.13, we demonstrate how the reduction of spatial covariance allows the better localization hemodynamic activation associated with a functional motor task by the removal of global, systemic changes. This data, previously described in Franceschini and coworkers [124], demonstrates that, in addition to the regional cerebral changes, functional tasks can result in global, systemic changes as well. During the finger-tapping task, as demonstrated in Figure 14.9, changes in heart rate, respiration, and blood pressure resulted in a task-associated, global hemodynamic response. These changes result in the appearance of a “functional activation” over nearly the entire head of the subject, as shown in Figure 14.13A. Using the PCA filter to project the strongest principal components calculated from a separate baseline recording, the resulting filtered response was localized to the contralateral motor cortex.

FIGURE 14.13

NIRS measurements are often sensitive to systemic fluctuations arising from blood pressure changes, respiration, or the cardiac cycle. As a complication, this physiology may change during the performance of intense stimulus tasks, such as motor activity. (more...)

14.6.5. HOMER—A Graphical Interface for Functional Optical Signals

In order to analyze our functional NIRS and DOT experiments, we have developed a Matlab-based graphic interface called HOMER for hemodynamic optically measured evoked response. The HOMER program is based on flexible data architecture that allows data to be imported with limited preparatory processing from virtually any currently existing commercial NIRS system. The analysis of this data can be tailored to allow the users complete versatility to specify the details of their experiment, including NIRS probe geometry and layout, optical wavelength selection, stimulus design, and other instrumental and acquisition parameters. The structure of the HOMER program is based on three levels of data processing. (1) At the first level, individual data files are represented. This level contains the data for a single experimental run, for example, a single acquisition scan. Scan-level processing allows the data from a single experimental run to be interactively visualized, time-series data filtered, and analyzed for functional responses. For a typical experiment, multiple such data files will be recorded within a single experimental session. (2) The session level represents the second organizing level in HOMER. This level includes the joint processing of one or more individual data files for a single experimental subject, recorded during a single session. At the session level, response averages and effects analysis can be calculated from multiple data runs. Statistical tests can be preformed to investigate the significance of functional changes (i.e., the f-test and t-test). In addition, images of the changes in optical absorption or hemoglobin concentrations can be reconstructed from these session averages. Session-level data should have consistent optical probe placement and geometry to allow analysis between scans. (3) At the final level of processing, multisession or group information is represented. This allows group averaging from multiple experimental subjects or multiple sessions from the same subject. Data processing at this level is allowed for region-of-interest comparisons to avoid intersubject registration issues. At all these three levels, data analysis, visualization, and data export are provided to Matlab or ASCII formats.

As shown in Figure 14.14, the layout of the HOMER program is based around an interactive graphical display of the NIRS or DOT probe, shown in the upper right. The user specifies this probe geometry within the data file imported into HOMER as described in the text. By selecting source (displayed as “x”) or detector (“o”) positions on this probe layout, the user navigates through the display of their data. Data shown in Figure 14.14 are from Reference [156].

FIGURE 14.14

The layout of the HOMER program is based around an interactive graphical display of the NIRS probe, shown in the upper right (B). The user specifies this probe geometry within the data file imported into HOMER. By selecting source (displayed as “x”) (more...)

14.7. MULTIMODAL AND VALIDATION STUDIES

In the early 1990s, functional magnetic resonance imaging methods began to appear, providing whole-brain imaging of blood flow [157,158] and the blood oxygen level dependent (BOLD) signals [159–163] associated with changes in blood flow and oxygen metabolism within the brain. Shortly after this development, NIRS was shown to have sensitivity similar to fMRI in a series of comparison studies [164–167]. Optical measurements of brain activation arise from similar hemodynamic contrast as functional MRI and thus, these two signals can be used to cross-validate one another and to explore assumptions about these technologies, such as vascular sensitivity. Concurrent multimodal measurements can also be used to explore the effects of intersubject and intertrial variation in the evoked hemodynamic response, since both modalities measure common physiological fluctuations at the same time, which can be used to separate measurement and physiological noise.

A first step toward cross-validating fMRI and optical techniques is to examine simultaneously acquired time courses from spatially coregistered locations. To date a number of groups have accomplished this as recently reviewed by Steinbrink and colleagues [168]. Figure 14.15 shows an example of a comparison between optical and fMRI signals from Huppert et al. [169] In this study, we preformed simultaneous near-infrared spectroscopy (NIRS) along with BOLD (blood oxygen level dependant) and blood flow measured using ASL (arterial spin labeling)-based fMRI during an event-related motor activity in human subjects in order to compare the temporal dynamics of the hemodynamic responses recorded in each method. The functional MRI BOLD signal represents changes in blood oxygenation and deoxy-hemoglobin concentration [170], and thus is expected to correlate to the optical measurement of deoxy-hemoglobin. Both are venous weighted measurements. Likewise, blood flow measured by ASL is an arterial weighted measurement [171] and expected to correlate to the arterial weighted oxy-hemoglobin changes. Multimodal measurements allow us to examine the validity of the biophysical models underlying each modality and, as a result, gain greater insight into the hemodynamic responses to neuronal activation. Concurrent measurements also allow us to examine physiological variability in the evoked response. In Figure 14.15, we show that the fMRI measured BOLD response is more correlated with the NIRS measure of deoxy-hemoglobin (R = 0.98; p < 10⁻²⁰) than with oxy-hemoglobin (R = 0.71) or total hemoglobin (R = 0.53). This result was predicted from the theoretical grounds of the BOLD response and is in agreement with several previous works [168,172–175]. This data has also allowed us to examine more detailed measurement models of the fMRI signal and comment on the roles of the oxygen saturation and blood volume contributions to the BOLD response. In addition, we found high correlation between the NIRS measured total hemoglobin and ASL measured cerebral blood flow (R = 0.91; p < 10⁻¹⁰) and oxy-hemoglobin with flow (R = 0.83; p < 10⁻⁵) as predicted by the biophysical models. Finally, we noted a significant amount of cross modality, correlated, intersubject variability in amplitude change, and time-to-peak of the hemodynamic response. The observed covariance in these parameters between subjects was in agreement with hemodynamic models, and provided further support that fMRI and NIRS have similar vascular sensitivity. This observation also implies that much of the intersubject variability in the temporal dynamics of the evoked response is of physiological origin rather then representing measurement or instrument errors. Using concurrent multimodal measurements, this variability can be further explored in future work.

FIGURE 14.15

Simultaneous NIRS and fMRI measurements of blood oxygen level dependant (BOLD) and blood flow using arterial spin labeling (ASL) were recorded during a brief 2-s finger-tapping task using a rapid event-related experimental design. The temporal dynamics (more...)

Once temporal correlation was established, the next step toward combining MRI and optical signals is to examine spatial correspondences in the diffuse optical image and the fMRI. However, as pointed out in the discussion of optical image reconstruction, the reconstruction of optical spatial maps for this comparison can be challenging due to the ill-poised nature of the inverse problem. This makes direct comparison of optical and MR images difficult because care must be taken to account for possible artifacts of the reconstruction process. Recently, Toronov et al. [172] and Joseph et al. [45] both described spatial comparisons between fMRI and optical. In the work by Joseph et al., tomographic imaging was used to improve spatial resolutions of the optical data [45]. For the most part, these spatial comparisons are qualitative, since the optical inverse problem cannot be strictly solved. To work around this problem, Sassaroli et al. [176], Huppert et al. [177], and Toronov et al. [173] proposed to use the calculated optical measurement sensitivity to weight the fMRI signal, thereby creating a forward projection of the high-spatial resolution MR image onto the optical source-detector measurement space. This approach allowed optical and MRI signals to be quantitatively compared by testing the MR image as a possible solution to the optical image problem.

In Figure 14.16, we show a spatial comparison between optical and fMRI. Using a model of light propagation through the tissue, which was estimated from the anatomical MRI images from each of the subjects, we calculated a weighted average of the underlying fMRI data that represented the optical measurements expected based on the fMRI images. We used this approach to compare the optical and fMRI spatial and temporal maps, and to examine the consistency of the fMRI data as a solution to the optical imaging problem. Like our earlier temporal comparison, we found statistically better spatial correlation between the fMRI-BOLD response and the optically measured deoxy-hemoglobin (R = 0.71; p = 1 × 10⁻⁷) than between the BOLD and oxy-or total hemoglobin measures (R = 0.38; p = 0.04 | 0.37; p = 0.05, respectively). Similarly, we found that the correlation between the ASL measured blood flow and optically measured total and oxy-hemoglobin is stronger (R = 0.73; p = 5 × 10⁻⁶ and R = 0.71; p= 9 × 10⁻⁶, respectively) than the flow to deoxy-hemoglobin spatial correlation (R = 0.26; p = 0.10).

FIGURE 14.16

The spatial profiles of optical and fMRI signals are compared for 10 subjects. Monte Carlo simulations (described in the text) were used to generate the sensitivity profiles for each optical measurement pair and were used to generate spatially weighted (more...)

One of the current limitations to optical methods is the process of image reconstruction and the mathematically ill-posed inverse problem. Several groups have demonstrated that is it possible to anatomical information provided by the MRI to obtain more accurate estimates of where the light has sampled within the head, using the Monte Carlo method described earlier [56,178–180]. To see how this synergy develops, consider the following sequence of events. First, we acquire anatomical MRI scans that (1) help localize our optical probes with respect to the underlying brain anatomy, and (2) can be used to segment the head into a variety of tissue types (typically, air, scalp, skull, cerebral spinal fluid, and white and gray matter). The segmented head can then be used for Monte Carlo studies in photon diffusion to determine the DOT instrument’s sensitivity to various tissue types and depths. Knowledge of the underlying tissue structure can also be used to limit the location of the reconstructed functional activation to the cortex of the brain [181]. These estimates can also be used as a model of the optical changes, which would allow more accurate quantitation of the Hb and HbO₂ changes during functional challenge. In this way, combined MRI and optical recordings have the potential to provide more changes found there—something that is impossible with either technique alone [142].

14.8. SUMMARY

The rapidly growing literature from the past 30 years clearly demonstrates the ability of near-infrared spectroscopy to noninvasively measure cerebral hemodynamic changes in response to stimuli. Furthermore, the literature is growing in support of the unique ability of NIRS to measure metabolic (cytochrome oxidase) and neuronal (fast optical scattering changes) signals. NIRS is unique as a neuro-monitoring/imaging technique as it is the only method which can potentially measure hemodynamic, metabolism, and neuronal signals simultaneously. Furthermore, the technology is relatively inexpensive and portable, allowing studies of brain activation in populations not easily studied in the past, such as babies and children, and freely behaving paradigms.

REFERENCES

1.

Jobsis FF. Noninvasive, infrared monitoring of cerebral and myocardial oxygen sufficiency and circulatory parameters. Science. 1977;198:1264–1267. [PubMed: 929199]

2.

Chance B, et al. Comparison of time-resolved and -unresolved measurements of deoxy-hemoglobin in brain. Proc Natl Acad Sci USA. 1988;85:4971–4975. [PMC free article: PMC281669] [PubMed: 3393526]

3.

Chance B. Optical method. Annu Rev Biophys Biophys Chem. 1991;20:1–28. [PubMed: 1867711]

4.

Cope M, et al. Methods of quantitating cerebral near infrared spectroscopy data. Adv Exp Med Biol. 1988;222:183–189. [PubMed: 3129910]

5.

Wray S, et al. Characterization of the near infrared absorption spectra of cytochrome aa3 and haemoglobin for the non-invasive monitoring of cerebral oxygenation. Biochim Biophys Acta. 1988;933:184–192. [PubMed: 2831976]

6.

McCormick PW, et al. Noninvasive cerebral optical spectroscopy for monitoring cerebral oxygen delivery and hemodynamics. Crit Care Med. 1991;19:89–97. [PubMed: 1986896]

7.

Benaron DA, et al. Noninvasive methods for estimating in vivo oxygenation. Clin Pediatr (Phila). 1992;31:258–273. [PubMed: 1582091]

8.

Elwell CE, et al. Quantification of adult cerebral hemodynamics by near-infrared spectroscopy. J Appl Physiol. 1994;77:2753–2760. [PubMed: 7896617]

9.

Gibson AP, et al. Recent advances in diffuse optical imaging. Phys Med Biol. 2005;50:R1–43. [PubMed: 15773619]

10.

von Siebenthal K, et al. Near-infrared spectroscopy in newborn infants. Brain Dev. 1992;14:135–143. [PubMed: 1514651]

11.

Elwell CE, et al. Measurement of CMRO2 in neonates undergoing intensive care using near infrared spectroscopy. Adv Exp Med Biol. 2005;566:263–268. [PubMed: 16594161]

12.

Delpy DT, et al. Cerebral monitoring in newborn infants by magnetic resonance and near infrared spectroscopy. Scand J Clin Lab Invest Suppl. 1987;188:9–17. [PubMed: 2835812]

13.

Weber P, et al. Cerebral hemodynamic changes in response to an executive function task in children with attention-deficit hyperactivity disorder measured by near-infrared spectroscopy. J Dev Behav Pediatr. 2005;26:105–111. [PubMed: 15827461]

14.

Soul JS, du Plessis AJ. New technologies in pediatric neurology. Near-infrared spectroscopy. Semin Pediatr Neurol. 1999;6:101–110. [PubMed: 10404564]

15.

Adelson PD, et al. The use of near infrared spectroscopy (NIRS) in children after traumatic brain injury: A preliminary report. Acta Neurochirurgica. 1998;71:250–254. [PubMed: 9779198]

16.

Mehagnoul-Schipper DJ, et al. Reproducibility of orthostatic changes in cerebral oxygenation in healthy subjects aged 70 years or older. Clin Physiol. 2001;21:77–84. [PubMed: 11168300]

17.

Kwee IL, Nakada T. Dorsolateral prefrontal lobe activation declines significantly with age—Functional NIRS study. J Neurol. 2003;250:525–529. [PubMed: 12736729]

18.

Hock C, et al. Near infrared spectroscopy in the diagnosis of Alzheimer’s disease. Ann N Y Acad Sci. 1996;777:22–29. [PubMed: 8624087]

19.

Miyai I, et al. Cortical mapping of gait in humans: A near-infrared spectroscopic topography study. NeuroImage. 2001;14:1186–1192. [PubMed: 11697950]

20.

Igawa M, et al. Activation of visual cortex in REM sleep measured by 24-channel NIRS imaging. Psychiatry Clin Neurosci. 2001;55:187–188. [PubMed: 11422835]

21.

Arridge SR, et al. Reconstruction methods for infra-red absorption imaging. Proc. SPIE. 1991;1431:204–215.

22.

Sassaroli A, Fantini S. Comment on the modified Beer-Lambert law for scattering media. Phys Med Biol. 2004;49:N255–257. [PubMed: 15357206]

23.

Kocsis L, et al. The modified Beer-Lambert law revisited. Phys Med Biol. 2006;51:N91–98. [PubMed: 16481677]

24.

Duncan A, et al. Optical pathlength measurements on adult head, calf and forearm and the head of the newborn infant using phase resolved optical spectroscopy. Phys Med Biol. 1995;40:295–304. [PubMed: 7708855]

25.

van der Zee P, et al. Experimentally measured optical pathlengths for the adult head, calf and forearm and the head of the newborn infant as a function of inter optode spacing. Adv Exp Med Biol. 1992;316:143–153. [PubMed: 1288074]

26.

Fantini S, et al. Non-invasive optical monitoring of the newborn piglet brain using continuous-wave and frequency-domain spectroscopy. Phys Med Biol. 1999;44:1543–1563. [PubMed: 10498522]

27.

Kohl M, et al. Determination of the wavelength dependence of the differential path-length factor from near-infrared pulse signals. Phys Med Biol. 1998;43:1771–1782. [PubMed: 9651039]

28.

Hiraoka M, et al. A Monte Carlo investigation of optical pathlength in inhomogeneous tissue and its application to near-infrared spectroscopy. Phys Med Biol. 1993;38:1859–1876. [PubMed: 8108489]

29.

Yoshitani K, et al. Effects of hemoglobin concentration, skull thickness, and the area of the cerebrospinal fluid layer on near-infrared spectroscopy measurements. Anesthesiology. 2007;106:458–462. [PubMed: 17325503]

30.

Strangman G, et al. Factors affecting the accuracy of near-infrared spectroscopy concentration calculations for focal changes in oxygenation parameters. NeuroImage. 2003;18:865–879. [PubMed: 12725763]

31.

Boas DA, et al. Diffuse optical imaging of brain activation: Approaches to optimizing image sensitivity, resolution, and accuracy. NeuroImage. 2004;23(Suppl 1):S275–288. [PubMed: 15501097]

32.

Sato H, et al. Practicality of wavelength selection to improve signal-to-noise ratio in near-infrared spectroscopy. NeuroImage. 2004;21:1554–1562. [PubMed: 15050579]

33.

Matcher SJ, et al. Performance comparison of several published tissue near-infrared spectroscopy algorithms. Anal Biochem. 1995;227:54–68. [PubMed: 7668392]

34.

Ishimaru A. Wave Propagation and Scattering in Random Media. Academic Press; New York: 1978.

35.

Flock ST, et al. Monte Carlo modeling of light propagation in highly scattering tissue—I: Model predictions and comparison with diffusion theory. IEEE Trans Biomed Eng. 1989;36:1162–1168. [PubMed: 2606490]

36.

Flock ST, et al. Monte Carlo modeling of light propagation in highly scattering tissues—II: Comparison with measurements in phantoms. IEEE Trans Biomed Eng. 1989;36:1169–1173. [PubMed: 2606491]

37.

Haskell RC, et al. Boundary conditions for the diffusion equation in radiative transfer. J Opt Soc Am A Opt Image Sci Vis. 1994;11:2727–2741. [PubMed: 7931757]

38.

Furutsu K, Yamada Y. Diffusion approximation for a dissipative random medium and the applications. Phys Rev E Stat Phys Plasmas Fluids Relat Interdiscip Topics. 1994;50:3634–3640. [PubMed: 9962417]

39.

Durduran T, et al. Does the photon-diffusion coefficient depend on absorption? J Opt Soc Am A Opt Image Sci Vis. 1997;14:3358–3365. [PubMed: 9392901]

40.

Fukui Y, et al. Monte Carlo prediction of near-infrared light propagation in realistic adult and neonatal head models. Appl Opt. 2003;42:2881–2887. [PubMed: 12790436]

41.

Boas D, et al. Three dimensional Monte Carlo code for photon migration through complex heterogeneous media including the adult human head. Opt. Express. 2002;10:159–170. [PubMed: 19424345]

42.

Wang L, et al. MCML—Monte Carlo modeling of light transport in multi-layered tissues. Comput Methods Programs Biomed. 1995;47:131–146. [PubMed: 7587160]

43.

Dale AM. Optimal experimental design for event-related fMRI. Hum Brain Mapp. 1999;8:109–114. [PMC free article: PMC6873302] [PubMed: 10524601]

44.

Boas DA, et al. Improving the diffuse optical imaging spatial resolution of the cerebral hemodynamic response to brain activation in humans. Opt Lett. 2004;29:1506–1508. [PubMed: 15259728]

45.

Joseph DK, et al. Design and validation of a time division multiplexed continuous wave diffuse optical tomography (DOT) system optimized for brain function imaging. Appl Opt. 2006;45:814208151.

46.

Custo A, et al. Effective scattering coefficient of the cerebral spinal fluid in adult head models for diffuse optical imaging. Appl Opt. 2006;45:4747–4755. [PubMed: 16799690]

47.

Zeff BW, et al. Retinotopic mapping of adult human visual cortex with high-density diffuse optical tomography. Proc Natl Acad Sci USA. 2007;104:12169–12174. [PMC free article: PMC1924577] [PubMed: 17616584]

48.

Barbour RL, et al. Optical tomographic imaging of dynamic features of dense-scattering media. J Opt Soc Am A Opt Image Sci Vis. 2001;18:3018–3036. [PubMed: 11760200]

49.

Arridge SR. Optical tomography in medical imaging. Inverse Problems. 1999;15:14–93.

50.

Farrell TJ, et al. A diffusion theory model of spatially resolved, steady-state diffuse reflectance for the noninvasive determination of tissue optical properties in vivo. Med Phys. 1992;19:879–888. [PubMed: 1518476]

51.

Natterer F. The Mathematics of Computerized Tomography. Society for Industrial and Applied Mathematics; Philadelphia: 2001.

52.

Benaron DA, Stevenson DK. Optical time-of-flight and absorbance imaging of biologic media. Science. 1993;259:1463–1466. [PubMed: 8451643]

53.

Cubeddu R, et al. Time-resolved imaging on a realistic tissue phantom: mu(s)’ and mu(a) images versus time-integrated images. Appl Opt. 1996;35:4533–4540. [PubMed: 21102872]

54.

Grosenick D, et al. Time-resolved imaging of solid phantoms for optical mammography. Appl Opt. 1997;36:221–231. [PubMed: 18250662]

55.

Selb J, et al. Improved sensitivity to cerebral hemodynamics during brain activation with a time-gated optical system: Analytical model and experimental validation. J Biomed Opt. 2005;10:11013. [PubMed: 15847579]

56.

Ntziachristos V, et al. MRI-guided diffuse optical spectroscopy of malignant and benign breast lesions. Neoplasia New York. 2002;4:347–354. [PMC free article: PMC1661680] [PubMed: 12082551]

57.

Gratton E, et al. Measurements of scattering and absorption changes in muscle and brain. Philos Trans R Soc Lond B Biol Sci. 1997;352:727–735. [PMC free article: PMC1691951] [PubMed: 9232861]

58.

Jiang HB, et al. Simultaneous Reconstruction of Optical-Absorption and Scattering Maps in Turbid Media from near-Infrared Frequency-Domain Data. Opt Lett. 1995;20:2128–2130. [PubMed: 19862273]

59.

Pogue BW, Patterson MS. Frequency-domain optical absorption spectroscopy of finite tissue volumes using diffusion theory. Phys Med Biol. 1994;39:1157–1180. [PubMed: 15552104]

60.

Pogue B, et al. Instrumentation and design of a frequency-domain diffuse optical tomography imager for breast cancer detection. Optics Express. 1997;1:391. [PubMed: 19377563]

61.

Chance B, et al. Phase measurement of light absorption and scatter in human tissue. Rev Scientific Instruments. 1998;69:3457–3481.

62.

Nioka S, et al. Human brain functional imaging with reflectance CWS. Oxygen Transport to Tissue Xix. 1997;428:237–242. [PubMed: 9500053]

63.

Siegel AM, et al. Design and evaluation of a continuous-wave diffuse optical tomography system. Opt Express. 1999;4:287–298. [PubMed: 19396285]

64.

Maki A, et al. Spatial and temporal analysis of human motor activity using noninvasive NIR topography. Med Phys. 1995;22:1997–2005. [PubMed: 8746704]

65.

Arridge SR, Lionheart WRB. Nonuniqueness in diffusion-based optical tomography. Opt Lett. 1998;23:882–884. [PubMed: 18087373]

66.

American National Standard for Save Use of Lasers. ANSI Z1361-1993. American National Standards Institute; New York: 1993.

67.

Heekeren HR, et al. Cerebral haemoglobin oxygenation during sustained visual stimulation—A near-infrared spectroscopy study. Philos Trans R Soc Lond B Biol Sci. 1997;352:743–750. [PMC free article: PMC1691960] [PubMed: 9232863]

68.

Meek JH, et al. Regional changes in cerebral haemodynamics as a result of a visual stimulus measured by near infrared spectroscopy. Proc Biol Sci. 1995;261:351–356. [PubMed: 8587877]

69.

Ruben J, et al. Haemoglobin oxygenation changes during visual stimulation in the occipital cortex. Adv Exp Med Biol. 1997;428:181–187. [PubMed: 9500046]

70.

Sakatani K, et al. Cerebral blood oxygenation changes induced by auditory stimulation in newborn infants measured by near infrared spectroscopy. Early Hum Dev. 1999;55:229–236. [PubMed: 10463787]

71.

Chen S, et al. Auditory-evoked cerebral oxygenation changes in hypoxic-ischemic encephalopathy of newborn infants monitored by near infrared spectroscopy. Early Hum Dev. 2002;67:113–121. [PubMed: 11893442]

72.

Zaramella P, et al. Brain auditory activation measured by near-infrared spectroscopy (NIRS) in neonates. Pediatr Res. 2001;49:213–219. [PubMed: 11158516]

73.

Franceschini MA, et al. Hemodynamic evoked response of the sensorimotor cortex measured noninvasively with near-infrared optical imaging. Psychophysiology. 2003;40:548–560. [PMC free article: PMC3786740] [PubMed: 14570163]

74.

Obrig H, et al. Cerebral oxygenation changes during motor and somatosensory stimulation in humans, as measured by near-infrared spectroscopy. Adv Exp Med Biol. 1996;388:219–224. [PubMed: 8798815]

75.

Obrig H, et al. Cerebral oxygenation changes in response to motor stimulation. J Appl Physiol. 1996;81:1174–1183. [PubMed: 8889751]

76.

Colier WN, et al. Human motor-cortex oxygenation changes induced by cyclic coupled movements of hand and foot. Exp Brain Res. 1999;129:457–461. [PubMed: 10591917]

77.

Hirth C, et al. Non-invasive functional mapping of the human motor cortex using near-infrared spectroscopy. Neuroreport. 1996;7:1977–1981. [PubMed: 8905707]

78.

Kleinschmidt A, et al. Simultaneous recording of cerebral blood oxygenation changes during human brain activation by magnetic resonance imaging and near-infrared spectroscopy. J Cereb Blood Flow Metab. 1996;16:817–826. [PubMed: 8784226]

79.

Kameyama M, et al. Frontal lobe function in bipolar disorder: A multichannel near-infrared spectroscopy study. NeuroImage. 2006;29:172–184. [PubMed: 16125979]

80.

Nagamitsu S, et al. Prefrontal cerebral blood volume patterns while playing video games-A near-infrared spectroscopy study. Brain Dev. 2006 [PubMed: 16481139]

81.

Okamoto M, et al. Prefrontal activity during taste encoding: An fNIRS study. NeuroImage. 2006;47(2):220–232. [PubMed: 16473020]

82.

Sumitani S, et al. Hemodynamic changes in the prefrontal cortex during mental works as measured by multi channel near-infrared spectroscopy (NIRS). J Med Invest. 2005;52(Suppl):302–303. [PubMed: 16366522]

83.

Sato H, et al. Temporal cortex activation during speech recognition: An optical topography study. Cognition. 1999;73:B55–66. [PubMed: 10585521]

84.

Adelson PD, et al. Noninvasive continuous monitoring of cerebral oxygenation periictally using near-infrared spectroscopy: A preliminary report. Epilepsia. 1999;40:1484–1489. [PubMed: 10565573]

85.

Sokol DK, et al. Near infrared spectroscopy (NIRS) distinguishes seizure types. Seizure. 2000;9:323–327. [PubMed: 10933986]

86.

Steinhoff BJ, et al. Ictal near infrared spectroscopy in temporal lobe epilepsy: A pilot study. Seizure. 1996;5:97–101. [PubMed: 8795123]

87.

Watanabe E, et al. Noninvasive cerebral blood volume measurement during seizures using multichannel near infrared spectroscopic topography. J Biomed Opt. 2000;5:287–290. [PubMed: 10958614]

88.

Buchheim K, et al. Decrease in haemoglobin oxygenation during absence seizures in adult humans. Neurosci Lett. 2004;354:119–122. [PubMed: 14698453]

89.

Watanabe E, et al. Focus diagnosis of epilepsy using near-infrared spectroscopy. Epilepsia. 2002;43(Suppl 9):50–55. [PubMed: 12383281]

90.

Fallgatter AJ, et al. Loss of functional hemispheric asymmetry in Alzheimer’s dementia assessed with near-infrared spectroscopy. Brain Res Cogn Brain Res. 1997;6:67–72. [PubMed: 9395850]

91.

Hanlon EB, et al. Near-infrared fluorescence spectroscopy detects Alzheimer’s disease in vitro. Photochem Photobiol. 1999;70:236–242. [PubMed: 10461462]

92.

Liebert A, et al. Bed-side assessment of cerebral perfusion in stroke patients based on optical monitoring of a dye bolus by time-resolved diffuse reflectance. NeuroImage. 2005;24:426–435. [PubMed: 15627584]

93.

Orihashi K, et al. Near-infrared spectroscopy for monitoring cerebral ischemia during selective cerebral perfusion. Eur J Cardiothorac Surg. 2004;26:907–911. [PubMed: 15519181]

94.

Terborg C, et al. Bedside assessment of cerebral perfusion reductions in patients with acute ischaemic stroke by near-infrared spectroscopy and indocyanine green. J Neurol Neurosurg Psychiatry. 2004;75:38–42. [PMC free article: PMC1757454] [PubMed: 14707304]

95.

Thiagarajah JR, et al. Noninvasive early detection of brain edema in mice by near-infrared light scattering. J Neurosci Res. 2005;80:293–299. [PubMed: 15765520]

96.

Chen WG, et al. Hemodynamic assessment of ischemic stroke with near-infrared spectroscopy. Space Med Med Eng (Beijing). 2000;13:84–89. [PubMed: 11543057]

97.

Nemoto EM, et al. Clinical experience with cerebral oximetry in stroke and cardiac arrest. Crit Care Med. 2000;28:1052–1054. [PubMed: 10809281]

98.

Saitou H, et al. Cerebral blood volume and oxygenation among poststroke hemiplegic patients: Effects of 13 rehabilitation tasks measured by near-infrared spectroscopy. Arch Phys Med Rehabil. 2000;81:1348–1356. [PubMed: 11030500]

99.

Vernieri F, et al. Near infrared spectroscopy and transcranial Doppler in monohemispheric stroke. Eur Neurol. 1999;41:159–162. [PubMed: 10202248]

100.

Eschweiler GW, et al. Left prefrontal activation predicts therapeutic effects of repetitive transcranial magnetic stimulation (rTMS) in major depression. Psychiatry Res. 2000;99:161–172. [PubMed: 11068197]

101.

Matsuo K, et al. Alteration of hemoglobin oxygenation in the frontal region in elderly depressed patients as measured by near-infrared spectroscopy. J Neuropsychiatry Clin Neurosci. 2000;12:465–471. [PubMed: 11083163]

102.

Okada F, et al. Dominance of the “nondominant” hemisphere in depression. J Affect Disord. 1996;37:13–21. [PubMed: 8682974]

103.

Matsuo K, et al. Decreased cerebral haemodynamic response to cognitive and physiological tasks in mood disorders as shown by near-infrared spectroscopy. Psychol Med. 2002;32:1029–1037. [PubMed: 12214784]

104.

Suto T, et al. Multichannel near-infrared spectroscopy in depression and schizophrenia: Cognitive brain activation study. Biol Psychiatry. 2004;55:501–511. [PubMed: 15023578]

105.

Fallgatter AJ, Strik WK. Reduced frontal functional asymmetry in schizophrenia during a cued continuous performance test assessed with near-infrared spectroscopy. Schizophr Bull. 2000;26:913–919. [PubMed: 11087023]

106.

Okada F, et al. Impaired interhemispheric integration in brain oxygenation and hemodynamics in schizophrenia. Eur Arch Psychiatry Clin Neurosci. 1994;244:17–25. [PubMed: 7918697]

107.

Kubota Y, et al. Prefrontal activation during verbal fluency tests in schizophrenia—A near-infrared spectroscopy (NIRS) study. Schizophr Res. 2005;77:65–73. [PubMed: 16005386]

108.

Shinba T, et al. Near-infrared spectroscopy analysis of frontal lobe dysfunction in schizophrenia. Biol Psychiatry. 2004;55:154–164. [PubMed: 14732595]

109.

Watanabe A, Kato T. Cerebrovascular response to cognitive tasks in patients with schizophrenia measured by near-infrared spectroscopy. Schizophr Bull. 2004;30:435–444. [PubMed: 15279058]

110.

Wyatt JS, et al. Quantification of cerebral oxygenation and haemodynamics in sick newborn infants by near infrared spectrophotometry. Lancet. 1986;2:1063–1066. [PubMed: 2877225]

111.

Chance B, Williams GR. The respiratory chain and oxidative phosphorylation. Adv Enzymol Relat Subj Biochem. 1956;17:65–134. [PubMed: 13313307]

112.

Lockwood AH, et al. Effects of acetazolamide and electrical stimulation on cerebral oxidative metabolism as indicated by the cytochrome oxidase redox state. Brain Res. 1984;308:9–14. [PubMed: 6089955]

113.

Wong-Riley MT. Cytochrome oxidase: An endogenous metabolic marker for neuronal activity. Trends Neurosci. 1989;12:94–101. [PubMed: 2469224]

114.

Wobst P, et al. Linear aspects of changes in deoxygenated hemoglobin concentration and cytochrome oxidase oxidation during brain activation. NeuroImage. 2001;13:520–530. [PubMed: 11170817]

115.

Salzberg BM, Obaid AL. Optical studies of the secretory event at vertebrate nerve terminals. J Exp Biol. 1988;139:195–231. [PubMed: 2850336]

116.

Ogawa S, et al. An approach to probe some neural systems interaction by functional MRI at neural time scale down to milliseconds. Proc Natl Acad Sci USA. 2000;97:11026–11031. [PMC free article: PMC27142] [PubMed: 11005873]

117.

Rector DM, et al. A focusing image probe for assessing neural activity in vivo. J Neurosci Methods. 1999;91:135–145. [PubMed: 10522832]

118.

Gratton G, et al. Shades of gray matter: Noninvasive optical images of human brain responses during visual stimulation. Psychophysiology. 1995;32:505–509. [PubMed: 7568645]

119.

Toronov V, et al. Near-infrared study of fluctuations in cerebral hemodynamics during rest and motor stimulation: Temporal analysis and spatial mapping. Med Phys. 2000;27:801–815. [PubMed: 10798703]

120.

Obrig H, et al. Spontaneous low frequency oscillations of cerebral hemodynamics and metabolism in human adults. NeuroImage. 2000;12:623–639. [PubMed: 11112395]

121.

Tanosaki M, et al. Variation of temporal characteristics in human cerebral hemodynamic responses to electric median nerve stimulation: A near-infrared spectroscopic study. Neurosci Lett. 2001;316:75–78. [PubMed: 11742719]

122.

Fallgatter AJ, et al. Prefrontal hypooxygenation during language processing assessed with near-infrared spectroscopy. Neuropsychobiology. 1998;37:215–218. [PubMed: 9648131]

123.

Okada F, et al. Region-dependent asymmetrical or symmetrical variations in the oxygenation and hemodynamics of the brain due to different mental stimuli. Brain Res Cogn Brain Res. 1995;2:215–219. [PubMed: 7580403]

124.

Franceschini MA, et al. Diffuse optical imaging of the whole head. J Biomed Opt. 2006;11:054007. [PMC free article: PMC2637816] [PubMed: 17092156]

125.

Taylor JA, et al. Low-frequency arterial pressure fluctuations do not reflect sympathetic outflow: Gender and age differences. Am J Physiol. 1998;274:H1194–1201. [PubMed: 9575922]

126.

Diamond SG, et al. Physiological system identification with the Kalman filter in diffuse optical tomography. Med Image Comput Comput Assist Interv Int Conf Med Image Comput Comput Assist Interv. 2005;8:649–656. [PubMed: 16686015]

127.

Diamond SG, et al. Dynamic physiological modeling for functional diffuse optical tomography. NeuroImage. 2006;30:88–101. [PMC free article: PMC2670202] [PubMed: 16242967]

128.

von Siebenthal K, et al. Cyclical fluctuations in blood pressure, heart rate and cerebral blood volume in preterm infants. Brain Dev. 1999;21:529–534. [PubMed: 10598053]

129.

Edvinsson L, Krause DN. Cerebral Blood Flow and Metabolism. Lippincott: Williamns and Wilkins; 2002.

130.

Buxton RB, et al. Modeling the hemodynamic response to brain activation. NeuroImage. 2004;23 (Suppl 1):S220–233. [PubMed: 15501093]

131.

Matthews F, et al. Hemodynamics for Brain-Computer Interfaces. IEEE Signal Processing Magazine. 2008 January:87–94.

132.

Izzetoglu M, et al. Motion artifact cancellation in NIR spectroscopy using Wiener filtering. IEEE Trans Biomed Eng. 2005;52:934–938. [PubMed: 15887543]

133.

Wilcox T, et al. Using near-infrared spectroscopy to assess neural activation during object processing in infants. J Biomed Opt. 2005;10:11010. [PMC free article: PMC1769348] [PubMed: 15847576]

134.

Strangman G, et al. Non-invasive neuroimaging using near-infrared light. Biol Psychiatry. 2002;52:679–693. [PubMed: 12372658]

135.

Zhang Y, et al. Eigenvector-based spatial filtering for reduction of physiological interference in diffuse optical imaging. J Biomed Opt. 2005;10:11014. [PubMed: 15847580]

136.

Baillet S, et al. Combined MEG and EEG source imaging by minimization of mutual information. IEEE Trans Biomed Eng. 1999;46:522–534. [PubMed: 10230131]

137.

Buxton RB, et al. Dynamics of blood flow and oxygenation changes during brain activation: The balloon model. Magn Reson Med. 1998;39:855–864. [PubMed: 9621908]

138.

Obata T, et al. Discrepancies between BOLD and flow dynamics in primary and supplementary motor areas: Application of the balloon model to the interpretation of BOLD transients. NeuroImage. 2004;21:144–153. [PubMed: 14741651]

139.

Mandeville JB, et al. Evidence of a cerebrovascular postarteriole windkessel with delayed compliance. J Cereb Blood Flow Metab. 1999;19:679–689. [PubMed: 10366199]

140.

Deneux T, Faugeras O. Using nonlinear models in fMRI data analysis: Model selection and activation detection. NeuroImage. 2006 [PubMed: 16844388]

141.

Boas DA, et al. Can the cerebral metabolic rate of oxygen be estimated with near-infrared spectroscopy? Phys Med Biol. 2003;48:2405–2418. [PubMed: 12953906]

142.

Huppert TJ. PhD Thesis. Harvard University; 2007.

143.

Riera J, et al. Fusing EEG and fMRI based on a bottom-up model: Inferring activation and effective connectivity in neural masses. Philos Trans R Soc Lond B Biol Sci. 2005;360:1025–1041. [PMC free article: PMC1854929] [PubMed: 16087446]

144.

Wan X, et al. The neural basis of the hemodynamic response nonlinearity in human primary visual cortex: Implications for neurovascular coupling mechanism. NeuroImage. 2006;32:616–625. [PubMed: 16697664]

145.

Ciftci K, et al. Constraining the general linear model for sensible hemodynamic response function waveforms. Medical and biological engineering and computing. 2008 [PubMed: 18427851]

146.

Jasdzewski G, et al. Differences in the hemodynamic response to event-related motor and visual paradigms as measured by near-infrared spectroscopy. NeuroImage. 2003;20:479–488. [PubMed: 14527608]

147.

Zhang Y, et al. A haemodynamic response function model in spatio-temporal diffuse optical tomography. Phys Med Biol. 2005;50:4625–4644. [PubMed: 16177494]

148.

Cohen-Adad J, et al. Activation detection in diffuse optical imaging by means of the general linear model. Med Image Anal. 2007;11:616–629. [PubMed: 17643341]

149.

Friston KJ, et al. Characterizing evoked hemodynamics with fMRI. NeuroImage. 1995;2:157–165. [PubMed: 9343598]

150.

Friston KJ, et al. Event-related fMRI: Characterizing differential responses. NeuroImage. 1998;7:30–40. [PubMed: 9500830]

151.

Cox RW. AFNI: Software for analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res. 1996;29:162–173. [PubMed: 8812068]

152.

Josephs O, Henson RN. Event-related functional magnetic resonance imaging: Modelling, inference and optimization. Philos Trans R Soc Lond B Biol Sci. 1999;354:1215–1228. [PMC free article: PMC1692638] [PubMed: 10466147]

153.

Friston KJ. Experimental Design and Statistical Issues. 2000:33–58.

154.

Wager TD, Nichols TE. Optimization of experimental design in fMRI: A general framework using a genetic algorithm. NeuroImage. 2003;18:293–309. [PubMed: 12595184]

155.

Cannestra AF, et al. Refractory periods observed by intrinsic signal and fluorescent dye imaging. J Neurophysiol. 1998;80:1522–1532. [PubMed: 9744956]

156.

Hoge RD, et al. Simultaneous recording of task-induced changes in blood oxygenation, volume, and flow using diffuse optical imaging and arterial spin-labeling MRI. NeuroImage. 2005;25:701–707. [PubMed: 15808971]

157.

Belliveau JW, et al. Magnetic resonance imaging mapping of brain function. Human visual cortex. Invest Radiol. 1992;27 (Suppl 2):S59–65. [PMC free article: PMC4097384] [PubMed: 1468876]

158.

Kim SG. Quantification of relative cerebral blood flow change by flow-sensitive alternating inversion recovery (FAIR) technique: Application to functional mapping. Magn Reson Med. 1995;34:293–301. [PubMed: 7500865]

159.

Belliveau JW, et al. Functional mapping of the human visual cortex by magnetic resonance imaging. Science. 1991;254:716–719. [PubMed: 1948051]

160.

Kwong KK, et al. Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc Natl Acad Sci USA. 1992;89:5675–5679. [PMC free article: PMC49355] [PubMed: 1608978]

161.

Ogawa S, et al. Intrinsic signal changes accompanying sensory stimulation: Functional brain mapping with magnetic resonance imaging. Proc Natl Acad Sci USA. 1992;89:5951–5955. [PMC free article: PMC402116] [PubMed: 1631079]

162.

Frahm J, et al. Dynamic MR imaging of human brain oxygenation during rest and photic stimulation. J Magn Reson Imaging. 1992;2:501–505. [PubMed: 1392241]

163.

Bandettini PA, et al. Time course EPI of human brain function during task activation. Magn Reson Med. 1992;25:390–397. [PubMed: 1614324]

164.

Hoshi Y, Tamura M. Dynamic multichannel near-infrared optical imaging of human brain activity. J Appl Physiol. 1993;75:1842–1846. [PubMed: 8282640]

165.

Villringer A, et al. Near infrared spectroscopy (NIRS): A new tool to study hemodynamic changes during activation of brain function in human adults. Neurosci Lett. 1993;154:101–104. [PubMed: 8361619]

166.

Okada F, et al. Gender- and handedness-related differences of forebrain oxygenation and hemodynamics. Brain Res. 1993;601:337–342. [PubMed: 8431784]

167.

Boas DA, et al. Scattering and wavelength transduction of diffuse photon density waves. Physical Review. E. Statistical Physics, Plasmas, Fluids, and Related Interdisciplinary Topics. 1993;47:R2999–R3002. [PubMed: 9960455]

168.

Steinbrink J, et al. Illuminating the BOLD signal: Combined fMRI-fNIRS studies. Magn Reson Imaging. 2006;24:495–505. [PubMed: 16677956]

169.

Huppert TJ, et al. A temporal comparison of BOLD, ASL, and NIRS hemodynamic responses to motor stimuli in adult humans. NeuroImage. 2006;29:368–382. [PMC free article: PMC2692693] [PubMed: 16303317]

170.

Norris DG. Principles of magnetic resonance assessment of brain function. J Magn Reson Imaging. 2006;23:794–807. [PubMed: 16649206]

171.

Buxton RB. Quantifying CBF with arterial spin labeling. J Magn Reson Imaging. 2005;22:723–726. [PubMed: 16261574]

172.

Toronov V, et al. Investigation of human brain hemodynamics by simultaneous near-infrared spectroscopy and functional magnetic resonance imaging. Med Phys. 2001;28:521–527. [PubMed: 11339749]

173.

Toronov VY, et al. A spatial and temporal comparison of hemodynamic signals measured using optical and functional magnetic resonance imaging during activation in the human primary visual cortex. NeuroImage. 2007;34(3):1136–1148. [PMC free article: PMC2752293] [PubMed: 17134913]

174.

MacIntosh BJ, et al. Transient hemodynamics during a breath hold challenge in a two part functional imaging study with simultaneous near-infrared spectroscopy in adult humans. NeuroImage. 2003;20:1246–1252. [PubMed: 14568493]

175.

Toronov V, et al. The roles of changes in deoxyhemoglobin concentration and regional cerebral blood volume in the fMRI BOLD signal. NeuroImage. 2003;19:1521–1531. [PubMed: 12948708]

176.

Sassaroli A, et al. Spatially weighted BOLD signal for comparison of functional magnetic resonance imaging and near-infrared imaging of the brain. NeuroImage. 2006;33:505–514. [PubMed: 16945553]

177.

Huppert T, et al. A quantitative spatial comparison of diffuse optical imaging with BOLD- and ASL-based fMRI. J Biomed Opt. 2006;11:064018. [PMC free article: PMC2670188] [PubMed: 17212541]

178.

Barbour RL, et al. MRI-guided optical tomography: Prospectives and computation for a new imaging method. IEEE Comput. Sci. Eng. 1995;2:63–77.

179.

Pogue BW, Paulsen KD. High-resolution near-infrared tomographic imaging of the rat cranium by use of a priori magnetic resonance imaging structural information. Opt Lett. 1998;23:1716–1718. [PubMed: 18091894]

180.

Intes X, et al. Diffuse optical tomography with physiological and spatial a priori constraints. Phys Med Biol. 2004;49:N155–163. [PubMed: 15272687]

181.

Boas DA, Dale AM. Simulation study of magnetic resonance imaging-guided cortically constrained diffuse optical tomography of human brain function. Appl Opt. 2005;44:1957–1968. [PubMed: 15813532]