Near infrared imaging

From Scholarpedia
David Boas and Maria Angela Franceschini (2009), Scholarpedia, 4(4):6997. doi:10.4249/scholarpedia.6997 revision #91552 [link to/cite this article]
Jump to: navigation, search

Near Infrared Spectroscopy and Imaging (NIRS) uses near infrared light between 650 and 950 nm to non-invasively probe the concentration and oxygenation of hemoglobin in the brain, muscle and other tissues and is used e.g. to detect changes induced by brain activity, injury, or disease. In brain research it complements functional magnetic resonance imaging (fMRI) by providing measures of both oxygenated and deoxygenated hemoglobin concentrations and by enabling studies in populations of subjects with experimental paradigms that are not amenable to fMRI.

Contents

Spectroscopic Origin of NIRS Signal

The dominant absorbers in the human body in the visible and near infrared wavelengths are oxygenated and deoxygenated hemoglobin and water. The absorption spectra of these chromophores are shown in Figure 1. Visible wavelengths of light are strongly absorbed by hemoglobin, decreasing significantly for the near infrared wavelengths greater than 650 nm. Above 950 nm, water absorption increases significantly. Thus, there is a window between 650 and 950 nm where the absorption of light is small and, despite the strong scattering of light by tissue, the light is able to diffuse several centimeters through the tissue and before it is detected.

Figure 1: This is the absorption spectra of oxygenated and deoxygenated hemoglobin and water for typical concentrations found in the brain.

It was first reported in 1977 (Jobsis 1977) that this near infrared light could diffuse through the intact scalp and skull of an adult human to probe the hemoglobin concentrations in the brain. Cope et al (Cope 1988) proposed a modified Beer-Lambert Law to quantify the absorption changes occurring in the brain by starting with the traditional Beer-Lambert Law to quantify the absorption coefficient within purely absorbing samples and adding a differential path-length factor (DPF) to account for the increased path-length of light through the sample due to the highly scattering tissue. The resulting equation is \(\tag{1} OD = -log \frac{I}{I_o} = \mu_a L D_{PF} + G \)

where \(OD\) is the optical density of the sample as determined from the negative log ratio of the detected intensity of light \(I\) with respect to the incident intensity of light \(I_o\ .\) The \(OD\) is related to the absorption coefficient of the tissue \(\mu_a\) multiplied by the net distance traveled by the light from the source to the detector, \(L\ ,\) scaled by the differential path-length factor \(D_{PF}\ ,\) plus a geometry factor \(G\ .\) \(D_{PF}\) accounts for the extra total distance that light travels through the tissue due to scattering, and \(G\) accounts for light attenuation because of the geometry of the sample. If a small change is induced in \(\mu_a\) between time \(t_1\) and \(t_2\ ,\) then the change in the optical density will be \(\tag{2} \Delta OD = -log \frac{I(t_2)}{I(t_1)} = \Delta \mu_a L D_{PF} \)

negating the need to estimate \(G\ .\) The change in the absorption coefficient is related to the concentration changes in the chromophores by their intrinsic extinction coefficients (Cope 1988).

Generally, to estimate changes in N chromophores, measurements of \(\Delta OD\) are required at N or more wavelengths of light. NIRS is typically used to estimate changes in oxygenated and deoxygenated hemoglobin requiring a minimum of 2 wavelengths. It is common to use only 2 wavelengths, one below the isosbestic point of hemoglobin and one above. The isosbestic point at 800 nm is where the extinction coefficient of oxygenated and deoxygenated hemoglobin is the same. By using wavelengths on either side of this wavelength, we have a measurement that is more sensitive to oxygenated hemoglobin and one which is more sensitive to deoxygenated hemoglobin. To optimize the estimation of the hemoglobin concentration changes in the brain with 2 wavelengths of light, it has been shown that one wavelength should be less than 765 nm and one greater than 830 nm (reviewed in (Boas 2004)). It should be noted that NIRS has the potential to probe the changes in redox state of cytochrome oxidase in the brain (Uludag 2004), but these changes are generally negligible except in pathological brain disorders.

A rigorous treatment of the propagation of light through tissue is more accurately formulated by the radiative transport equation. When the probability for photon scattering is much greater than the probability for absorption, then the radiative transport equation can be reduced to a simple diffusion equation (reviewed in Arridge 1999). From this diffusion equation it is possible to derive the modified Beer-Lambert Law (eq. (1) and eq.(2) and obtain explicit relations for the differential path-length factor \(D_{PF}\) and the geometry term \(G\ .\)

While NIRS is typically performed using instruments that emitted continuous wave light and simply measure the intensity of light propagated through the tissue, it is also possible to perform measurements where the source of light is intensity modulated (between 50 to 500 MHz) or pulsed (typically pulsed on for less than 100 ps) and the detector resolves respectively the phase or temporal delay of the light propagating through the tissue. These measurements are usually called frequency domain or time domain measurements and because they provide direct measurements of photon propagation delay within the tissue as well as the intensity (reviewed in Gibson 2005 and Wolf 2007), it is possible to estimate absolute absorption and scattering properties of the tissue in addition to the changes that can be estimated by continuous-wave NIRS. In addition, it is possible to use near infrared light to measure blood flow non-invasively in the brain using diffuse correlation spectroscopy (Durduran 2004) which exploits the fact that photons experience a Doppler shift in their wavelength when they scatter from moving red blood cells.

Imaging and Tomography

Figure 2: NIRS Sensitivity profile to absorption changes within an adult human head for a source and detector separated by 4.5 cm on the scalp. Contours are drawn for every 10 fold decrease in sensitivity. The scalp, skull, cerebral spinal fluid, and gray and white matter are distinguished by different shades of gray.

A single NIRS measurement is sensitive to a volume of tissue that falls between the source of light entering the tissue and the detector receiving the light that diffuses out of the tissue. The spatial profile of the NIRS volume sensitivity can be calculated by the photon transport and diffusion equations. Figure 2 shows the sensitivity profile of a continuous-wave measurement to absorption changes within an adult human head given a source and detector on the scalp separated by 4.5 cm. The sensitivity profile extends through the scalp and skull into the most superficial cortex. Note that the separation between the source and detector must be large enough so that the detected photons are sensitive to absorption changes in the brain. In adult human subjects it is common to use a 3 cm separation for continuous-wave NIRS measurements although smaller separations can be used. Separations of 1.5 to 3.0 cm are typically used on infants. Note that in adults, insufficient light reaches detectors at separations larger than 5 or 6 cm and thus trans-illumination of the head is not possible which ultimately severely limits depth resolution within the brain.

By performing measurements with multiple sources and detectors distributed over the scalp, one obtains overlapping sensitivity profiles which enable the spatial localization of absorption changes within the brain. This can be formalized by generalizing the modified Beer-Lambert Law for a set of discrete voxels each having a potentially different absorption change \(\tag{3} \Delta OD_i = \Delta \mu_{a,j} L_{i,j} \)

where \(L_{i,j}\) is the effective path-length of light through the \(j^{th}\) voxel for the \(i^{th}\) measurement. Eq. (3) can be written in matrix form as \(y = A x\) where each row of \(y\) and \(A\) represents a measurement and each row of \(x\) represents a voxel. Further, an explicit relation for the \(L_{i,j}\) can be derived from the diffusion equation (Arridge 1999). Given measurements of \(\Delta OD\) from multiple sources and detectors, the inverse problem is to estimate the spatial variation in \(\Delta \mu_a\ .\) Because of the smoothing arising from the highly scattered light, this is an ill-conditioned inverse problem. In addition, there are usually fewer measurements than unknowns making the problem underdetermined. Solutions to eq. (3) can thus only be obtained by using regularization methods that impose priors on the solution. It is common to use Tikhonov regularization to minimize the norm of the solution, but prior information about the spectral, spatial, and temporal behavior of the solution are also being incorporated into the inverse problem.

Mapping Brain Activity

Figure 3: The video displays the real-time evolution of deoxygenated hemoglobin changes during contralateral finger-tapping (reproduced from (Franceschini 2000) with permission).

Neuronal activity results in a complex sequence of events, generally referred to as neurovascular coupling, that produce changes in blood flow and oxygen consumption. An increase in blood flow will be associated with increases in blood volume and will deliver more oxygenated hemoglobin into the downstream capillaries and veins and washout deoxygenated hemoglobin. An increase in oxygen consumption will increase the concentration of deoxygenated hemoglobin and decrease the concentration of oxygenated hemoglobin. The interplay between the blood flow and oxygen consumption during brain activation generally results in an increase in blood volume and oxygenated hemoglobin and a decrease in deoxygenated hemoglobin.

These hemoglobin concentration changes during brain activation were first measured with NIRS in 1993 (Chance 1993, Hoshi 1993, Kato 1993, Villringer 1993). Instruments with arrays of sources and detectors were quickly developed to enable mapping of the hemoglobin changes associated with brain activity (Maki 1995). An example movie of deoxygenated hemoglobin decreases during a simple finger tapping motor stimulation is shown in Figure 3 (reproduced with permission from (Franceschini 2000)). This movie shows that hemoglobin changes are localized over the contra-lateral motor cortex and reveals that responses to single stimulation sequences can be observed. NIRS has been used to investigate a range of applications including mapping of brain activity in healthy subjects, prevention and treatment of seizures, psychiatric concerns, stroke monitoring and rehabilitation, and Alzheimer’s disease (see review by Wolf 2007). Perhaps the most important brain function application areas for NIRS are those that complement fMRI such as studies in infants and children and elderly subjects who are less likely to remain sufficiently still for fMRI, and studies with behavioral paradigms that are not easily performed during fMRI (e.g. social interactions and physical therapy).

Advantages and Disadvantages

Near Infrared Spectroscopy has an advantage of being a portable and inexpensive device for functional neuroimaging comparable to electroencephalography (EEG), as opposed to fMRI, positron emission tomography (PET) and magnetoencephalography (MEG). It doesn’t require severely limiting subject motion. It is a non-invasive measure using laser or incoherent light sources that are typically safer than laser pointers. It can provide quantitative measures of the changes in oxygenated and deoxygenated hemoglobin that can help resolve ambiguities in the effects of the blood flow and oxygen consumption responses to brain activation. It can measure these hemodynamic signals with a temporal resolution of 100 Hz or higher, significantly greater than fMRI or PET aiding in the resolution of the onset time of brain activation and in filtering physiological interference from the brain activation signals of interest. Measurements can bridge from the human level down to microscopic invasive measurements in animals enabling efficient translation of knowledge in both directions. Finally, NIRS is compatible with fMRI, PET, MEG, EEG enabling interference free simultaneous multi-modal studies of brain activation.

NIRS is limited by depth sensitivity which in adult humans typically reaches 5 to 10 millimeters beneath the inner surface of the skull for brain activation. This can be improved with time-domain methods where depth sensitivity increases with delayed detection of the propagating pulse of light. Spatial resolution is limited by the ill-conditioned inverse tomography problem. Most studies do not employ overlapping measurements and thus lateral resolution is approximately equal to the source-detector separation. Lateral spatial resolution can be improved by a factor of 2 or more by incorporating overlapping measurements. Depth resolution within the brain, however, is challenged by the lack of transmission measurements through the head. The limited lateral and depth spatial resolution causes a partial volume issue that results in underestimating changes in the hemoglobin concentrations. Similar to EEG, care must be taken to insure hair does not severely block the transmission of light between the source and detector and the scalp, although conductive paste is not necessary as is the case for EEG. While subjects are less physically constrained than for fMRI, PET, and MEG, care must still be taken to secure the optical probe on the head of the subject to minimize motion artifacts.

Multi-Modal Imaging

Figure 4: The plastic and glass fiber optics used for NIRS do not interfere with simultaneous measures by fMRI.

NIRS measurements are typically made with fiber optic cables that deliver light to and from the head and the instrument. These fiber optic cables thus generally do not interfere with fMRI, EEG, MEG, and PET, enabling simultaneous multi-modal studies to be performed. Figure 4 shows an example of NIRS being used in conjunction with MRI. The advantages of concurrent NIRS-BOLD fMRI studies include: increased spatial resolution for NIRS imaging by using spatial constraints provided by fMRI; better filtering of physiological fluctuations within fMRI by using the higher temporal resolution NIRS measures of the fluctuations; and the additional measures of oxygenated and total hemodynamic changes improves physiological interpretation of the fMRI signals and estimation of the cerebral metabolic rate of oxygen. Simultaneous NIRS-EEG measurements offer artifact-free EEG signal detection, contrary to fMRI-EEG measures, with potential benefits for neurovascular coupling studies. Concurrent NIRS-EEG/MEG measurements will play an important role in the study of neurovascular coupling in humans bridging to more invasive studies in animals.

References

Arridge, S.R. (1999). Optical tomography in medical imaging. Inverse Problems 15, R41-R93.

Boas, D.A., Dale, A.M., and Franceschini, M.A. (2004). Diffuse optical imaging of brain activation: approaches to optimizing image sensitivity, resolution, and accuracy. NeuroImage 23, S275-S288.

Chance, B., Zhuang, Z., UnAh, C., Alter, C., and Lipton, L. (1993). Cognition-activated low-frequency modulation of light absorption in human brain. Proc. Natl. Acad. Sci. U. S. A. 90, 3770-3774.

Cope, M., Delpy, D.T., Reynolds, E.O.R., Wray, S., Wyatt, J., Van der Zee, P. (1988). Methods of quantitating cerebral near infrared spectroscopy data. Adv Exp Med Biol 222, 183-189.

Durduran, T., Yu, G., Burnett, M.G., Detre, J.A., Greenberg, J.H., Wang, J., Zhou, C., Yodh, A.G. (2004). Diffuse optical measurement of blood flow, blood oxygenation, and metabolism in a human brain during sensorimotor cortex activation. Opt Lett 29:1766-1768.

Franceschini, M.A., Toronov, V., Filiaci, M., Gratton, E., Fanini, S. (2000). On-line optical imaging of the human brain with 160-ms temporal resolution. Optics Express 6:49-57.

Gibson, A.P., Hebden, J.C., Arridge, S.R. (2005). Recent advances in diffuse optical imaging. Phys Med Biol 50:R1-43

Hoshi, Y., Tamura, M. (1993). Dynamic multichannel near-infrared optical imaging of human brain activity. J Appl Physiol 75:1842-1846

Jobsis, F.F. (1977). Noninvasive infrared monitoring of cerebral and myocardial sufficiency and circulatory parameters. Science 198, 1264-1267.

Kato, T., Kamei, A., Takashima, S., Ozaki, T. (1993). Human visual cortical function during photic stimulation monitoring by means of near-infrared spectroscopy. J. Cereb. Blood Flow Metab. 13, 516-520.

Maki, A., Yamashita, Y., Ito, Y., Watanabe, E., Mayanagi, Y., Koizumi, H. (1995). Spatial and temporal analysis of human motor activity using noninvasive NIR topography. Med. Phys. 22:1997-2005.

Uludag, K., Steinbrink, J., Kohl-Bareis, M., Wenzel, R., Villringer, A., Obrig ,H. (2004). Cytochrome-c-oxidase redox changes during visual stimulation measured by near-infrared spectroscopy cannot be explained by a mere cross talk artefact. Neuroimage 22:109-119.

Villringer, A., Planck, J., Hock, C., Schleinkofer, L., Dirnagl, U. (1993). Near infrared spectroscopy (NIRS): a new tool to study hemodynamic changes during activation of brain function in human adults. Neuroscience Letters 154:101-104.

Wolf, M., Ferrari, M., Quaresima, V. (2007). Progress of near infrared spectroscopy and imaging instrumentation for brain and muscle clinical applications. J. Biomed. Opt. 12: 062104.

Internal references

  • Giovanni Gallavotti (2008) Fluctuations. Scholarpedia, 3(6):5893.
  • William D. Penny and Karl J. Friston (2007) Functional imaging. Scholarpedia, 2(5):1478.
  • Rodolfo Llinas (2008) Neuron. Scholarpedia, 3(8):1490.


See also

fMRI, MRI

Personal tools
Namespaces
Variants
Actions
Navigation
Focal areas
Activity
Toolbox