OSA's Digital Library

Optics Express

Optics Express

  • Editor: C. Martijn de Sterke
  • Vol. 18, Iss. 18 — Aug. 30, 2010
  • pp: 19386–19395
« Show journal navigation

Optimal quantitation of the cerebral hemodynamic response in functional near-infrared spectroscopy

Irina Schelkanova and Vladislav Toronov  »View Author Affiliations


Optics Express, Vol. 18, Issue 18, pp. 19386-19395 (2010)
http://dx.doi.org/10.1364/OE.18.019386


View Full Text Article

Acrobat PDF (1056 KB)





Browse Journals / Lookup Meetings

Browse by Journal and Year


   


Lookup Conference Papers

Close Browse Journals / Lookup Meetings

Article Tools

Share
Citations

Abstract

We have compared cerebral hemodynamic changes measured by near-infrared spectroscopy (NIRS) with simultaneously acquired BOLD fMRI signals during breath hold challenge in humans. The oxy- and deoxyhemoglobin concentration changes were obtained from the same broadband NIRS data using four different quantitation methods. One method used only two wavelengths (690 nm and 830 nm), and three other methods used broadband data with different spectral fitting algorithms. We found that the broadband techniques employing spectral derivatives were significantly superior to the multi-wavelength methods in terms of the correlation with the BOLD signals. In two cases out of six we found that the time courses of the deoxyhemoglobin changes produced by the two-wavelength method were qualitatively inconsistent with the BOLD fMRI signals.

© 2010 OSA

1. Introduction

In recent years many research groups began to use home-made or commercially available NIRS monitors in basic research on brain function and cognition [1

1. A. Gibson and H. Dehghani, “Diffuse optical imaging,” Philos. Transact. A Math. Phys. Eng. Sci. 367(1900), 3055–3072 (2009). [CrossRef] [PubMed]

4

4. G. Pfurtscheller, G. Bauernfeind, S. C. Wriessnegger, and C. Neuper, “Focal frontal (de)oxyhemoglobin responses during simple arithmetic,” Int. J. Psychophysiol. 76(3), 186–192 (2010). [CrossRef] [PubMed]

] and for clinical monitoring of the brain during and after cardiac surgery [5

5. C. Fedorow and H. P. Grocott, “Cerebral monitoring to optimize outcomes after cardiac surgery,” Curr. Opin. Anaesthesiol. 23(1), 89–94 (2010). [CrossRef]

]. Most of these monitors have limited numbers of wavelengths. Monitors from NIRO family (Hamamatsu Photonics, Hamamatsu, Japan) use four wavelengths, while many, such as INVOS (Somanetics, Troy, MI) and ISS oximeters (ISS, Champaign, IL) use only two wavelengths of light. Such a multi-wavelength design based on non-dispersive light detectors, such as photomultipliers, was dictated by the availability and costs of the detectors in 1990. However the emergence of new generations of low-cost portable CCD spectrometers now allows for relatively inexpensive multichannel designs with broadband spectral resolution. Other options to achieve broadband spectral sensitivity are provided by acousto-optical [6

6. M. E. Martin, M. Wabuyele, M. Panjehpour, B. Overholt, R. DeNovo, S. Kennel, G. Cunningham, and T. Vo-Dinh, “An AOTF-based dual-modality hyperspectral imaging system (DMHSI) capable of simultaneous fluorescence and reflectance imaging,” Med. Eng. Phys. 28(2), 149–155 (2006). [CrossRef]

] or liquid crystal tunable filters [7

7. S. C. Gebhart, R. C. Thompson, and A. Mahadevan-Jansen, “Liquid-crystal tunable filter spectral imaging for brain tumor demarcation,” Appl. Opt. 46(10), 1896–1910 (2007). [CrossRef] [PubMed]

]. The goal of this study is to re-evaluate the capabilities of the broadband NIRS approach for cerebral perfusion monitoring in comparison with the multi-wavelength approach. We use a simple broad-band setup to measure cerebral responses to breatholding challenge in adult humans and compare our results with the simultaneously acquired blood oxygen level dependent (BOLD) functional MRI signals, which are indicative of the cerebral deoxyhemoglobin concentration changes ΔHb. Using different spectral analysis techniques we show that the best correlation between the BOLD fMRI and NIRS time courses occurs when ΔHbis obtained using an algorithm based on the spectral derivative fit of the attenuation.

The water and the deoxyhemoglobin spectral second derivative features were used in the past to quantify the chromophore concentration changes in human forearm [8

8. S. J. Matcher, M. Cope, and D. T. Delpy, “Use of the water absorption spectrum to quantify tissue chromophore concentration changes in near-infrared spectroscopy,” Phys. Med. Biol. 39(1), 177–196 (1994). [CrossRef] [PubMed]

] and the absolute deoxyhemoglobin concentration in the neonatal brain [9

9. C. E. Cooper, C. E. Elwell, J. H. Meek, S. J. Matcher, J. S. Wyatt, M. Cope, and D. T. Delpy, “The noninvasive measurement of absolute cerebral deoxyhemoglobin concentration and mean optical path length in the neonatal brain by second derivative near infrared spectroscopy,” Pediatr. Res. 39(1), 32–38 (1996). [CrossRef] [PubMed]

]. In both [8

8. S. J. Matcher, M. Cope, and D. T. Delpy, “Use of the water absorption spectrum to quantify tissue chromophore concentration changes in near-infrared spectroscopy,” Phys. Med. Biol. 39(1), 177–196 (1994). [CrossRef] [PubMed]

] and [9

9. C. E. Cooper, C. E. Elwell, J. H. Meek, S. J. Matcher, J. S. Wyatt, M. Cope, and D. T. Delpy, “The noninvasive measurement of absolute cerebral deoxyhemoglobin concentration and mean optical path length in the neonatal brain by second derivative near infrared spectroscopy,” Pediatr. Res. 39(1), 32–38 (1996). [CrossRef] [PubMed]

] the spectral derivative features were only used to estimate the photon differential pathlengths. In this study we used spectral derivatives in a different fashion, namely we use them to calculate the chromophore concentration changes by fitting the first and the second spectral derivatives of the attenuation.

2. Materials and methods

2.1 NIRS

Near-infrared light was generated by a stabilized fan-cooled AvaLight-HAL Tungsten Halogen Light Source (Avantes Inc., Broomfield, CO) with an adjustable focusing connector to maximize light coupling with the source fiber. The source fibre bundle was made of 30 Thorlabs broadband silica 400 μm core diameter fibers. On the probe side the source fibers were arranged circularly around the location of the detector bundle at a radius of 25 mm. Light was collected using a 3-mm diameter fiber optic bundle (Sunoptic Technologies, FL). The length of all fibres was 6 m so that the source and the detector could be set outside of the MRI scanner room. The detector bundle was connected to a QE65000 cooled spectrometer (Ocean Optics, Dunedin, FL), which had a spectral range between 650 and 1100 nm. The spectrometer output was digitized using the Spectral Suite software (Ocean Optics, Dunedin, FL).

The optical probe was positioned on the left side of the forehead near the hairline. Vitamin E capsules were attached to the probe for the visualization on MR images. A proper positioning against the frontal lobe avoiding sinuses was ensured using anatomical MRI. Spectra were acquired at the sampling rate of one spectrum per second. This sampling rate was selected to ensure that the instrumental noise does not affect physiological data. MRI

Imaging was performed at Sunnybrook Health Sciences Centre using the Achieva 3.0T scanner with SENSE-Head-8 coil. In order to assess the position of the optical probe the T2-weighted, two-dimensional, turbo spin echo (T2W TSE; repetition time/echo time 3000/80 ms, 1-mm slice thickness, no gap, 22-cm field of view, 0.43 × 0.43 × 1.00 mm voxel size,5.25-min acquisition time) anatomical images were acquired. Functional images were taken during 360 seconds by EPI technique with T2*-weighted protocol (FE-EPI sequence, TR = 1000 ms, TE = 35 ms, flip angle = 90 deg). The in-plane resolution was 1.72 × 1.72 mm2 (FoV = 22 cm at 128 × 128 pixel2), and 15 axial slices, each 4 mm thick.

2.2 Subjects and protocol

Breath holding was used because it generates blood CO2 which increases BOLD signal as CO2 acts as a cerebral vasodilator [10

10. B. J. MacIntosh, L. M. Klassen, and R. S. Menon, “Transient hemodynamics during a breath hold challenge in a two part functional imaging study with simultaneous near-infrared spectroscopy in adult humans,” Neuroimage 20(2), 1246–1252 (2003). [CrossRef] [PubMed]

]. Vasodilation causes increased cerebral blood flow which washes the deoxyhemoglobin out of the cerebral capillary bed thus increasing both the cerebral blood volume and oxygenation [11

11. L. P. Safonova, A. Michalos, U. Wolf, M. Wolf, D. M. Hueber, J. H. Choi, R. Gupta, C. Polzonetti, W. W. Mantulin, and E. Gratton, “Age-correlated changes in cerebral hemodynamics assessed by near-infrared spectroscopy,” Arch. Gerontol. Geriatr. 39(3), 207–225 (2004). [CrossRef] [PubMed]

]. Six young healthy adult females (19-29 years old) were placed into the scanner and audibly cued to perform a breath hold at the end of expiration with voluntary resumption of breathing after holding about 15-20 seconds. The exercise was repeated five times with 1 minute intervals. Research has been performed according to Sunnybrook Guidelines applicable to MRI studies on healthy volunteers and to Ryerson University 2008-003-01 Research Ethics protocol.

2.3 Modeling

We assumed that changes in the normalized attenuation (ΔA) of near-infrared light in scattering tissue were governed by the Modified Beer Lambert Law:
ΔA(λ,t)=1DP(λ)ln(I0(λ)It(λ))=εHbO2(λ)ΔHbO2(t)+εHb(λ)ΔHb(t)
(1)
where DP(λ) is the differential pathlength factor (DPF), I0(λ)is the average intensity, It(λ)is the instantaneous intensity, εHbO2and εHbare the oxy- and deoxy-hemoglobin extinction coefficients, respectively, and ΔHbO2and ΔHbare the instantaneous hemoglobin concentration deviations from their average levels. We used the DPF dependence on the wavelength measured in [12

12. L. M. Klassen, B. J. MacIntosh, and R. S. Menon, “Influence of hypoxia on wavelength dependence of differential pathlength and near-infrared quantification,” Phys. Med. Biol. 47(9), 1573–1589 (2002). [CrossRef] [PubMed]

] at normal fraction of inspired oxygen.

We used four different methods based on Eq. (1) to recover ΔHbO2(t)and ΔHb(t)from our spectral data.

  • 1) For two isolated wavelengths Eq. (1) reduces to a set of two linear equations with constant coefficients. We solved this system of equations for the wavelengths of 690 nm and 830 nm which are used in many bi-wavelength instruments. To reduce the noise the data were averaged within 5 nm wavebands centered at either 690 or 830 nm, respectively.
  • 2) For the quasi-continuum of wavelengths between 690 nm and 900 nm we found instantaneous ΔHbO2and ΔHbusing the general linear model (GLM) fit (glmfit function of MATLAB) of the instantaneous absorbance ΔA(λ,t).
  • 3) To find ΔHbO2and ΔHbwe also used the same GLM fit with the first spectral derivative of the instantaneous absorbance ΔA(λ,t)λ .
  • 4) We used the GLM fit of the second spectral derivative of the instantaneous absorbance2ΔA(λ,t)λ2 to obtain ΔHb only, since the second derivative of the oxy-hemoglobin extinction εHbO2(λ)is almost zero compared to that forεHb(λ).

The time series of the oxy- and deoxyhemoglobin changes obtained using the above four methods will be further referred to as the type 1-4 signals and the correlations of those with BOLD will be referred to as type the 1-4 correlations. We will also use the notations ΔHb14 to denote the deoxyhemoglobin signals obtained by Methods 1-4.

In all methods using broadband data (2-4) the parameters of glmfit were the same. Namely, the normal error distribution was assumed and the constant offset was set off. We found that the inclusion of the optional constant offset always produced unrealistically small ΔHb2(t) changes (less than 0.01 μM). Without using the offset in glmfit all ΔHb(t) amplitudes were always within 0.5-1 μM. In Methods 3 and 4 the spectra were smoothed using smooth function with a 3 nm span in order to reduce the effect of the spectral noise. This size of the span produced ΔHb3,4(t)andΔHbO22,3(t) of similar amplitudes to those of ΔHb1,2(t)andΔHbO21,2(t).

2.4 Noise analysis

In order to test how ΔHb signals produced by different methods were affected by the instrumental noise we have performed the power spectrum analysis of the noise and physiological signals. The instrumental noise was acquired on a tissue-like phantom (ISS, Champaign, IL) using the same instrumentation and acquisition settings as those used for the physiological acquisitions. Due to the optical properties of the phantom the spectral intensity values of the phantom signal were in the same range of values as the for the in-vivo signals. Figure 1(a)
Fig. 1 (a) Temporal power spectra of noise for all four types ofΔHb. Numbers 1 through 4 in the legend correspond to the type of the signal. The bar shows the spectral band of the signal due to repeated breath holdings. (b) Normalized power spectral densities of noise (acquired on the phantom) and cross-subject average signal (acquired on a subject during exercise) at 0.017 Hz.
compares the power spectra of all four types of ΔHbobtained from the phantom data. Although a much smaller number of photons contributed to the type 1 signal than to other signals, from Fig. 1(a) one can see that the noise influence was strongest for the type 3 and 4 signals. This was both due to the relatively low sampling frequency and due to the higher spectral noise in the first and second order spectral derivatives of the absorbance spectra than in the absorbance itself. However, as Fig. 1(b) shows, during the in-vivo measurements the instrumental noise was insignificant compared to the physiological signals. Figure 1(b) compares the temporal power spectra of the raw data acquired on the phantom and in-vivo. The plotted power spectral densities correspond to the frequency of the breath holding repetitions (0.017 Hz) as functions of the wavelength. In order to quantitatively compare the signal and the noise power spectra both spectra were normalized to the corresponding time-averaged spectral intensities. From Fig. 1(b) one can see that in the range between 690 and 900 nm the signal to noise ratio was much greater than one. In particular, it was close to four near 690 nm and close to 60 near 800 nm. This confirms that none of the four types of derived hemodynamic responses were significantly affected by the instrumental noise.

2.5 Comparison between NIRS and BOLD MRI

Although a number of studies (see for example [13

13. V. Toronov, A. Webb, J. H. Choi, M. Wolf, L. Safonova, U. Wolf, and E. Gratton, “Study of local cerebral hemodynamics by frequency-domain near-infrared spectroscopy and correlation with simultaneously acquired functional magnetic resonance imaging,” Opt. Express 9(8), 417–427 (2001), http://www.opticsinfobase.org/oe/viewmedia.cfm?URI=oe-9-8-417&seq=0. [CrossRef] [PubMed]

]) have revealed that BOLD signal most closely corresponds to the negative of ΔHb(t), we compared BOLD signals both with the changes in the deoxyhemoglobin and oxyhemoglobin concentrations obtained by the above four methods. For this we used the correlation analysis tools implemented in AFNI software [14

14. A. F. N. I. Main Page, http://afni.nimh.nih.gov/afni

]. Before applying the correlation analysis the fMRI data sets were preprocessed to correct for motion artifacts and slice timing differences and to remove non-brain regions. Then both fMRI and NIRS time series were filtered using the digital filter with the pass band between 0.01 and 0.02 Hz corresponding to the breath hold repetition frequency. The Butterworth filter was designed using the MATLAB signal processing tool and applied to the data time series using the filtfilt function to prevent temporal shifts of filtered signals. Then the correlation coefficients between –ΔHb(t), ΔHbO2(t) and the voxel BOLD signals were computed for voxels within the volume of the brain interrogated by the optical channel. Each such a volume of interrogation included approximately 1000 fMRI voxels. Following the approach used in [15

15. A. Sassaroli, B. deB Frederick, Y. Tong, P. F. Renshaw, and S. Fantini, “Spatially weighted BOLD signal for comparison of functional magnetic resonance imaging and near-infrared imaging of the brain,” Neuroimage 33(2), 505–514 (2006). [CrossRef] [PubMed]

] and [16

16. T. J. Huppert, R. D. Hoge, A. M. Dale, M. A. Franceschini, and D. A. Boas, “Quantitative spatial comparison of diffuse optical imaging with blood oxygen level-dependent and arterial spin labeling-based functional magnetic resonance imaging,” J. Biomed. Opt. 11(6), 064018 (2006). [CrossRef]

] we determined the volume of interrogation for each measurement using the voxel photon-hitting density Pn defined by Eq. (1) in [15

15. A. Sassaroli, B. deB Frederick, Y. Tong, P. F. Renshaw, and S. Fantini, “Spatially weighted BOLD signal for comparison of functional magnetic resonance imaging and near-infrared imaging of the brain,” Neuroimage 33(2), 505–514 (2006). [CrossRef] [PubMed]

]. Only voxels with Pn >0.1∙max(Pn) were included into the interrogation volume. The average correlation coefficients and the average BOLD signals were computed using the equation
X¯=vxnWn
(2)
where Wn = Pn/Σn Pn, xnwas the voxel correlation or BOLD intensity and X¯ was the corresponding volume average quantity. For the analysis of the statistical significance of the average correlation the t-test was applied to the Wn -weighted voxel correlations.

3. Results

Figure 2(a)
Fig. 2 (a) Time courses of –ΔHb(t)for one of six subjects. Numbers 1 through 4 in the legend correspond to four different methods to obtainΔHb(t). The vertical lines show the beginning of each breath hold. (b) Volume-average BOLD signal (c) Time courses of ΔHbO2(t)for the same measurement.
shows the time courses of –ΔHb(t) (negative of deoxyhemoglobin concentration change) for one of six subjects. Numbers 1 through 4 in the legend correspond to four different methods to obtainΔHb(t). Figure 2(b) shows the BOLD signal averaged over the volume near the optical probe (see Fig. 3
Fig. 3 BOLD-NIRS Correlation map. Red color corresponds to high positive correlation, and blue color corresponds to high negative correlation. The arrows show the positions of the light source and detector.
). In Fig. 2(a) one can see that although all curves showed similar amplitudes of changes and a certain degree of synchronicity with the exercise, their time courses were visibly different. While green and blue curves in Fig. 2(a) were well in phase with the BOLD signal shown in Fig. 2(b) (the correlation values were 0.72 and 0.56, respectively, see Table 1

Table 1. Volume-averaged correlation coefficients for all subjects: three types of ΔHbO2(t) and four types of –ΔHb(t). The corresponding confidence intervals were all close to ± 0.01

table-icon
View This Table
), the red curve was anti-correlated with BOLD (correlation value of −0.43), and the black curve was not well-correlated with BOLD (correlation value of 0.23). However, all three ΔHbO2(t) shown in Fig. 2(c) were well correlated with the average BOLD signal (correlations of 0.75 and higher for this subject). We did not observe significant differences in the amplitudes of neither ΔHbO2(t)nor ΔHb(t) obtained by different methods.

Figure 3 shows one slice of the functional image obtained by correlating ΔHb3(t)with the voxel BOLD signals. The voxels shown by colors were within the volume interrogated by the light channel. Red color corresponded to high positive correlation, and blue color corresponded to high negative correlation. One can see that although most of voxels showed positive correlation between BOLD and ΔHb3(t), some voxels showed negative correlations.

4. Discussion

We have computed cerebral oxy- and deoxyhemoglobin responses to breath hold challenge from the same broadband near-infrared data using four different methods. The first method used only two wavelengths, and three other methods used broadband data with different spectral fitting algorithms. The ΔHbO2(t)signals were always well synchronized with the BOLD signal. Unlike ΔHbO2(t), in some cases the behaviors of ΔHb1(t) and ΔHb2(t)were qualitatively different from the time courses of the volume-average BOLD signals and of ΔHb3(t) and ΔHb4(t). The latter were always consistent with the time course of the average BOLD signals. At a group average level all four methods were qualitatively consistent with BOLD in terms of the sign of the correlation. However, the group average correlations for ΔHb3(t) and ΔHb4(t) were higher than for ΔHb1(t)andΔHb2(t), and the corresponding confidence intervals did not overlap. One should emphasize that methods 3 and 4 were essentially broadband ones as they used spectral derivatives of the absorbance changes.

The methods that used the spectral derivatives were more prone to temporal instrumental noise because the fitting procedures converted the spectral noise into the temporal one. However the same methods produced –ΔHb(t) signals that were better correlated with the BOLD signals than two other methods. The highest group correlation was obtained using the first spectral derivative and the worst correlation was produced by the direct spectral absorbance fit (Method 2). Both Methods 2 and 3 used the same mathematical fitting algorithm, i.e. the GLM fit, but were applied either to the absorbance change (Method 2) or to the spectral derivative of it (Methods 3). The reason for such a contrast behavior of these two similar methods should was twofold. The first part of the reason was physiological and consisted in the opposite directions of changes in the oxy- and deoxyhenoglobin concentrations due to the washout effect [17

17. L. P. Safonova, A. Michalos, U. Wolf, M. Wolf, D. M. Hueber, J. H. Choi, R. Gupta, C. Polzonetti, W. W. Mantulin, and E. Gratton, “Age-correlated changes in cerebral hemodynamics assessed by near-infrared spectroscopy,” Arch. Gerontol. Geriatr. 39(3), 207–225 (2004). [CrossRef] [PubMed]

]. The increases in the oxyhemoglobin concentration caused the entire absorbance spectrum to increase but with greater effect in the longer wavelength half of the used spectral band (wavelengths greater than 800 nm). The decreases in the deoxyhemoglobin concentration cased the opposite effect on the absorbance but mostly in the short-wavelength part of the spectrum (wavelengths from 700 to 800 nm). Since the concentration of the deoxyhemoglobin in the brain was much lower than the concentration of the oxyhemoglobin, the relative changes in the former were also much smaller than in the latter. Therefore the overall rise of the absorption across the spectral band 700-900 nm was much stronger than the differences between the short- and the long-wavelength halves of the band. These differences in the effects of the oxy- and deoxyhemoglobin changes on the absorbance spectrum could result in a poor quantitation of ΔHb(t)by Methods 1 and 2 in the cases when signal distortions due to physiological or motion artifacts occurred. These distortions did not affect significantly ΔHbO2(t)because of the larger amplitude of ΔHbO2(t)changes. However, the effect of ΔHb(t)andΔHbO2(t)on the spectral derivatives of the absorbance was better balanced than that on the absorbance itself because the differentiation eliminated the homogeneous changes across the spectrum and magnified the effect of the deoxyhemoglobin concentration changes due to the strong feature in the first- and second-order differential extinction spectra of deoxihemoglobin near 760 nm (see Fig. 6 in [8

8. S. J. Matcher, M. Cope, and D. T. Delpy, “Use of the water absorption spectrum to quantify tissue chromophore concentration changes in near-infrared spectroscopy,” Phys. Med. Biol. 39(1), 177–196 (1994). [CrossRef] [PubMed]

]).

The second differential extinction spectrum of the deoxyhemoglobin had even stronger feature near 760 nm than the first differential spectrum and therefore in terms of ΔHb(t) Method 4 theoretically could provide even better result than Method 3. However, since the second differential also magnified the spectral noise, the Method 3 provided the best performance. The two-wavelength Method 1 worked slightly better than the broadband Method 2 because the algorithm differences. The algorithm of Method 1 was based not on the fitting of the absorbance changes spectrum but rather on the solution of the linear system of equations for the absorption changes at 690 nm and 830 nm. At these two wavelengths the difference between the extinction spectra of HHb and HbO2 were high and in particular at the 690 nm the HbO2 extinction was minimal so that the effect of HbO2 changes at 690 nm was much smaller than the effect of HHb changes. However, both Method 3 and 4 outperformed Method 1 in terms of ΔHb(t)quantitation.

In [18

18. J. Virtanen, T. Noponen, and P. Meriläinen, “Comparison of principal and independent component analysis in removing extracerebral interference from near-infrared spectroscopy signals,” J. Biomed. Opt. 14(5), 054032 (2009). [CrossRef] [PubMed]

] the principal and the independent component analyses (PCA and ICA, respectively) were used to clean the distorted hemodynamic signals measured by a two-wavelength system at 30 mm source-detector separations during breath holding. Both PCA and ICE produced better signals in cases of slight distortions, i.e. when typical exercise-synchronized increases in ΔHbO2(t) and decreases in ΔHb(t) were noticeable in non-cleaned data. However, these methods would not improve the behavior of “bad” signals such as the red curve in Fig. 2, a in our case unless more channels with different source-detector separations were used. Nevertheless, if the underlying reasons for a bad performance of Methods 1 and 2 were the data artifacts such as the hemodynamic fluctuations in the scalp, there is a chance that the cerebral signals acquired using two-wavelength methods can be cleaned if the distortions were isolated, for example by applying advanced signal processing techniques to the data acquired at short and long source-detector distances. On the other hand, our results show that our broadband Method 3 at least provides a single-channel alternative to PCA and ICA which require multi-channel measurements.

5. Conclusions and future work

We have compared cerebral hemodynamic signals obtained using near-infrared spectroscopy (NIRS) with simultaneously acquired BOLD fMRI signals during breatholding challenge. The oxy- and deoxyhemoglobin concentration changes were obtained from the same broadband NIRS data using four different quantitation methods. One method used only two wavelengths, and three other methods used broadband data with different spectral fitting algorithms. We have found that broadband techniques using spectral derivative algorithms were superior over the multi-wavelength methods in studies of cerebral hemodynamic responses to stimuli in humans.

In our future work we plan to apply the independent and principal component analyses to the broadband data acquired at multiple source-detector distances such that the contribution of the scalp can be separated from the cerebral signals. This should clarify whether the advanced signal processing techniques applied to the multi-distance data will be sufficient to resolve the problems of the two-wavelength method or the broadband approach is required to ensure the correct measurement of cerebral deoxyhemoglobin changes.

Acknowledgements

This research has been supported by the Ryerson University Faculty of Engineering, Architecture and Science Dean’s Research Fund. The authors are also thankful to Sunnybrook Health Sciences Centre for providing MRI.

References and links

1.

A. Gibson and H. Dehghani, “Diffuse optical imaging,” Philos. Transact. A Math. Phys. Eng. Sci. 367(1900), 3055–3072 (2009). [CrossRef] [PubMed]

2.

S. Perrey, “Non-invasive NIR spectroscopy of human brain function during exercise,” Methods 45(4), 289–299 (2008). [CrossRef] [PubMed]

3.

M. Calderon-Arnulphi, A. Alaraj, and K. V. Slavin, “Near infrared technology in neuroscience: past, present and future,” Neurol. Res. 31(6), 605–614 (2009). [CrossRef] [PubMed]

4.

G. Pfurtscheller, G. Bauernfeind, S. C. Wriessnegger, and C. Neuper, “Focal frontal (de)oxyhemoglobin responses during simple arithmetic,” Int. J. Psychophysiol. 76(3), 186–192 (2010). [CrossRef] [PubMed]

5.

C. Fedorow and H. P. Grocott, “Cerebral monitoring to optimize outcomes after cardiac surgery,” Curr. Opin. Anaesthesiol. 23(1), 89–94 (2010). [CrossRef]

6.

M. E. Martin, M. Wabuyele, M. Panjehpour, B. Overholt, R. DeNovo, S. Kennel, G. Cunningham, and T. Vo-Dinh, “An AOTF-based dual-modality hyperspectral imaging system (DMHSI) capable of simultaneous fluorescence and reflectance imaging,” Med. Eng. Phys. 28(2), 149–155 (2006). [CrossRef]

7.

S. C. Gebhart, R. C. Thompson, and A. Mahadevan-Jansen, “Liquid-crystal tunable filter spectral imaging for brain tumor demarcation,” Appl. Opt. 46(10), 1896–1910 (2007). [CrossRef] [PubMed]

8.

S. J. Matcher, M. Cope, and D. T. Delpy, “Use of the water absorption spectrum to quantify tissue chromophore concentration changes in near-infrared spectroscopy,” Phys. Med. Biol. 39(1), 177–196 (1994). [CrossRef] [PubMed]

9.

C. E. Cooper, C. E. Elwell, J. H. Meek, S. J. Matcher, J. S. Wyatt, M. Cope, and D. T. Delpy, “The noninvasive measurement of absolute cerebral deoxyhemoglobin concentration and mean optical path length in the neonatal brain by second derivative near infrared spectroscopy,” Pediatr. Res. 39(1), 32–38 (1996). [CrossRef] [PubMed]

10.

B. J. MacIntosh, L. M. Klassen, and R. S. Menon, “Transient hemodynamics during a breath hold challenge in a two part functional imaging study with simultaneous near-infrared spectroscopy in adult humans,” Neuroimage 20(2), 1246–1252 (2003). [CrossRef] [PubMed]

11.

L. P. Safonova, A. Michalos, U. Wolf, M. Wolf, D. M. Hueber, J. H. Choi, R. Gupta, C. Polzonetti, W. W. Mantulin, and E. Gratton, “Age-correlated changes in cerebral hemodynamics assessed by near-infrared spectroscopy,” Arch. Gerontol. Geriatr. 39(3), 207–225 (2004). [CrossRef] [PubMed]

12.

L. M. Klassen, B. J. MacIntosh, and R. S. Menon, “Influence of hypoxia on wavelength dependence of differential pathlength and near-infrared quantification,” Phys. Med. Biol. 47(9), 1573–1589 (2002). [CrossRef] [PubMed]

13.

V. Toronov, A. Webb, J. H. Choi, M. Wolf, L. Safonova, U. Wolf, and E. Gratton, “Study of local cerebral hemodynamics by frequency-domain near-infrared spectroscopy and correlation with simultaneously acquired functional magnetic resonance imaging,” Opt. Express 9(8), 417–427 (2001), http://www.opticsinfobase.org/oe/viewmedia.cfm?URI=oe-9-8-417&seq=0. [CrossRef] [PubMed]

14.

A. F. N. I. Main Page, http://afni.nimh.nih.gov/afni

15.

A. Sassaroli, B. deB Frederick, Y. Tong, P. F. Renshaw, and S. Fantini, “Spatially weighted BOLD signal for comparison of functional magnetic resonance imaging and near-infrared imaging of the brain,” Neuroimage 33(2), 505–514 (2006). [CrossRef] [PubMed]

16.

T. J. Huppert, R. D. Hoge, A. M. Dale, M. A. Franceschini, and D. A. Boas, “Quantitative spatial comparison of diffuse optical imaging with blood oxygen level-dependent and arterial spin labeling-based functional magnetic resonance imaging,” J. Biomed. Opt. 11(6), 064018 (2006). [CrossRef]

17.

L. P. Safonova, A. Michalos, U. Wolf, M. Wolf, D. M. Hueber, J. H. Choi, R. Gupta, C. Polzonetti, W. W. Mantulin, and E. Gratton, “Age-correlated changes in cerebral hemodynamics assessed by near-infrared spectroscopy,” Arch. Gerontol. Geriatr. 39(3), 207–225 (2004). [CrossRef] [PubMed]

18.

J. Virtanen, T. Noponen, and P. Meriläinen, “Comparison of principal and independent component analysis in removing extracerebral interference from near-infrared spectroscopy signals,” J. Biomed. Opt. 14(5), 054032 (2009). [CrossRef] [PubMed]

OCIS Codes
(170.0170) Medical optics and biotechnology : Medical optics and biotechnology
(170.2655) Medical optics and biotechnology : Functional monitoring and imaging

ToC Category:
Medical Optics and Biotechnology

History
Original Manuscript: June 8, 2010
Revised Manuscript: July 29, 2010
Manuscript Accepted: August 18, 2010
Published: August 27, 2010

Virtual Issues
Vol. 5, Iss. 13 Virtual Journal for Biomedical Optics

Citation
Irina Schelkanova and Vladislav Toronov, "Optimal quantitation of the cerebral hemodynamic response in functional near-infrared spectroscopy," Opt. Express 18, 19386-19395 (2010)
http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-18-18-19386


Sort:  Author  |  Year  |  Journal  |  Reset  

References

  1. A. Gibson and H. Dehghani, “Diffuse optical imaging,” Philos. Transact. A Math. Phys. Eng. Sci. 367(1900), 3055–3072 (2009). [CrossRef] [PubMed]
  2. S. Perrey, “Non-invasive NIR spectroscopy of human brain function during exercise,” Methods 45(4), 289–299 (2008). [CrossRef] [PubMed]
  3. M. Calderon-Arnulphi, A. Alaraj, and K. V. Slavin, “Near infrared technology in neuroscience: past, present and future,” Neurol. Res. 31(6), 605–614 (2009). [CrossRef] [PubMed]
  4. G. Pfurtscheller, G. Bauernfeind, S. C. Wriessnegger, and C. Neuper, “Focal frontal (de)oxyhemoglobin responses during simple arithmetic,” Int. J. Psychophysiol. 76(3), 186–192 (2010). [CrossRef] [PubMed]
  5. C. Fedorow and H. P. Grocott, “Cerebral monitoring to optimize outcomes after cardiac surgery,” Curr. Opin. Anaesthesiol. 23(1), 89–94 (2010). [CrossRef]
  6. M. E. Martin, M. Wabuyele, M. Panjehpour, B. Overholt, R. DeNovo, S. Kennel, G. Cunningham, and T. Vo-Dinh, “An AOTF-based dual-modality hyperspectral imaging system (DMHSI) capable of simultaneous fluorescence and reflectance imaging,” Med. Eng. Phys. 28(2), 149–155 (2006). [CrossRef]
  7. S. C. Gebhart, R. C. Thompson, and A. Mahadevan-Jansen, “Liquid-crystal tunable filter spectral imaging for brain tumor demarcation,” Appl. Opt. 46(10), 1896–1910 (2007). [CrossRef] [PubMed]
  8. S. J. Matcher, M. Cope, and D. T. Delpy, “Use of the water absorption spectrum to quantify tissue chromophore concentration changes in near-infrared spectroscopy,” Phys. Med. Biol. 39(1), 177–196 (1994). [CrossRef] [PubMed]
  9. C. E. Cooper, C. E. Elwell, J. H. Meek, S. J. Matcher, J. S. Wyatt, M. Cope, and D. T. Delpy, “The noninvasive measurement of absolute cerebral deoxyhemoglobin concentration and mean optical path length in the neonatal brain by second derivative near infrared spectroscopy,” Pediatr. Res. 39(1), 32–38 (1996). [CrossRef] [PubMed]
  10. B. J. MacIntosh, L. M. Klassen, and R. S. Menon, “Transient hemodynamics during a breath hold challenge in a two part functional imaging study with simultaneous near-infrared spectroscopy in adult humans,” Neuroimage 20(2), 1246–1252 (2003). [CrossRef] [PubMed]
  11. L. P. Safonova, A. Michalos, U. Wolf, M. Wolf, D. M. Hueber, J. H. Choi, R. Gupta, C. Polzonetti, W. W. Mantulin, and E. Gratton, “Age-correlated changes in cerebral hemodynamics assessed by near-infrared spectroscopy,” Arch. Gerontol. Geriatr. 39(3), 207–225 (2004). [CrossRef] [PubMed]
  12. L. M. Klassen, B. J. MacIntosh, and R. S. Menon, “Influence of hypoxia on wavelength dependence of differential pathlength and near-infrared quantification,” Phys. Med. Biol. 47(9), 1573–1589 (2002). [CrossRef] [PubMed]
  13. V. Toronov, A. Webb, J. H. Choi, M. Wolf, L. Safonova, U. Wolf, and E. Gratton, “Study of local cerebral hemodynamics by frequency-domain near-infrared spectroscopy and correlation with simultaneously acquired functional magnetic resonance imaging,” Opt. Express 9(8), 417–427 (2001), http://www.opticsinfobase.org/oe/viewmedia.cfm?URI=oe-9-8-417&seq=0 . [CrossRef] [PubMed]
  14. A. F. N. I. Main Page, http://afni.nimh.nih.gov/afni
  15. A. Sassaroli, B. deB Frederick, Y. Tong, P. F. Renshaw, and S. Fantini, “Spatially weighted BOLD signal for comparison of functional magnetic resonance imaging and near-infrared imaging of the brain,” Neuroimage 33(2), 505–514 (2006). [CrossRef] [PubMed]
  16. T. J. Huppert, R. D. Hoge, A. M. Dale, M. A. Franceschini, and D. A. Boas, “Quantitative spatial comparison of diffuse optical imaging with blood oxygen level-dependent and arterial spin labeling-based functional magnetic resonance imaging,” J. Biomed. Opt. 11(6), 064018 (2006). [CrossRef]
  17. L. P. Safonova, A. Michalos, U. Wolf, M. Wolf, D. M. Hueber, J. H. Choi, R. Gupta, C. Polzonetti, W. W. Mantulin, and E. Gratton, “Age-correlated changes in cerebral hemodynamics assessed by near-infrared spectroscopy,” Arch. Gerontol. Geriatr. 39(3), 207–225 (2004). [CrossRef] [PubMed]
  18. J. Virtanen, T. Noponen, and P. Meriläinen, “Comparison of principal and independent component analysis in removing extracerebral interference from near-infrared spectroscopy signals,” J. Biomed. Opt. 14(5), 054032 (2009). [CrossRef] [PubMed]

Cited By

Alert me when this paper is cited

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.

Figures

Fig. 1 Fig. 2 Fig. 3
 

« Previous Article  |  Next Article »

OSA is a member of CrossRef.

CrossCheck Deposited