SPIE Proceedings [SPIE SPIE BiOS - San Francisco, California, USA (Saturday 2 February 2013)] Optical Tomography and Spectroscopy of Tissue X - Near-infrared spectroscopy system with

Near-infrared spectroscopy system with non-contact source and detector for in vivo multi-distance measurement of deep biological tissue

Tsukasa Funane*1, 2, Hirokazu Atsumori1, Masashi Kiguchi1, Yukari Tanikawa3, and Eiji Okada2

1Hitachi, Ltd., Central Research Laboratory, Hatoyama, Saitama 350-0395, Japan 2Keio University, Department of Electronics and Electrical Engineering, Yokohama 223-8522, Japan

3National Institute of Advanced Industrial Science and Technology, Ibaraki 305-8564, Japan

ABSTRACT

A non-contact near-infrared spectroscopy (NIRS) scanning system with a phosphor cell placed on the skin for in vivo measurement of biological tissue was developed and evaluated. Because the phosphor is excited by the light that propagates in the tissue, and the excitation light is cut by optical filters, the light that propagates in the tissue is selectively detected. The non-contact system was extended to create a scanning system that can flexibly change source positions with a galvano scanner. The optical scanning system was used for non-contact measurement of the human forearm muscle, and the dependence of optical-density change (ΔOD) caused by the upper-arm occlusion and release on source-detector distance was observed. The obtained ΔOD demonstrates the effectiveness of using this system for multi-distance human-forearm measurement. Furthermore, a human forehead was measured with the system. To extract a deep-layer signal, a surface-layer subtraction method with short-distance regression was applied to measured data. On the basis of the correlation with a simultaneously measured laser-Doppler flowmetry signal, it was confirmed that the deep-layer signal was successfully extracted. The extraction result demonstrates that the optical scanning system can be used as a multi-distance NIRS system for measuring the human brain activity at the forehead.

Keywords: Near-infrared spectroscopy (NIRS), muscle, brain, scanning system, biological tissue, non-contact, phosphor, multi-distance

1. INTRODUCTION Near-infrared spectroscopy (NIRS) is a noninvasive technology for measuring changes in cerebral hemodynamics and oxygenation caused by brain activity.1–3 An NIRS system irradiates near-infrared light on the scalp and detects the light that passes through the brain by using a reflection arrangement. The change in cerebral hemodynamics and oxygenation is then estimated on the basis of the change in the intensity of the detected light. NIRS is used for noninvasive measurement of brain functions in medical and industrial applications.4

Whereas NIRS has found a variety of applications and has many advantages over other neuroimaging modalities, such as functional magnetic resonance imaging (fMRI), positron emission tomography (PET), and electroencephalogram (EEG), it has several limitations,5,6 such as the quantification of NIRS data and influence of extracerebral tissue, that have been recently reported by many researchers.

While these limitations should be addressed and investigated further, the limitation focused on herein is the probe arrangement of NIRS, that is, the restriction placed on the system-human interface by the contact probes of the NIRS system. In the case of an NIRS system, the light sources and detectors or light guides/fibers are usually attached to the skin so that the light detectors catch the light that propagates in the tissue rather than skin-reflected light or stray light. If the light source and detector are not in contact with the skin, the detector will catch skin-reflected light or stray light as noise; however, the intensity of such noise is much higher than that of tissue-propagated light because the intensity of the latter is approximately 10−7–10−9 times as high as that of the incident light when source-detector (SD) distance is 30 mm.7 Consequently, the signal-to-noise ratio (SNR) deteriorates. Owing to this SNR reduction, non-contact measurement of changes in cerebral blood volume is difficult. If this problem (SNR reduction) is overcome, a non-

* E-mail: [email protected]

Optical Tomography and Spectroscopy of Tissue X, edited by Bruce J. Tromberg, Arjun G. Yodh, Eva Marie Sevick-Muraca, Proc. of SPIE Vol. 8578, 85782W

© 2013 SPIE · CCC code: 1605-7422/13/$18 · doi: 10.1117/12.2002254

Proc. of SPIE Vol. 8578 85782W-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/25/2013 Terms of Use: http://spiedl.org/terms

contact system would provide more degrees of freedom (with respect to both the system and subject)8 and enable new applications of brain-activity measurement, such as sleep research and long-term brain monitoring.

In this study, a non-contact optical scanning system9 is applied to measurement of human biological tissue. In particular, optical-density change caused by blood-volume change in the human forearm muscle during upper-arm occlusion and relief was measured. The dependence of optical-density change on SD distance was thereby obtained. This result demonstrates the possibility that the optimal SD distance for oxygenation monitoring on the human forearm can be determined by the scanning system. The scanning system was then used to measure human brain activity.

2. HUMAN-MUSCLE MEASUREMENT 2.1 Participant

One male adult (30 years old) participated as a volunteer subject in this study. The subject provided written informed consent after being provided a complete explanation of the study.

2.2 Experimental setup

The forearm of the subject was measured by using a non-contact optical scanning system.9 The measurement principle involves a phosphor layer placed on the skin and optical filters for eliminating excitation light and stray light. As a result of this set-up, neither the emitter probes nor the detector probes of the system are in contact with the skin.10,11 SD distances at which optical-density change (ΔOD) was measured were 10, 13.8, 17.5, 21.3, and 25 mm.

An upper-arm ischemia test was performed on the male subject. The blood flow in the upper arm was manually occluded with an inelastic band (by another person) for less than 60 s, at which point the band was released. ΔOD was measured during the period of constriction and for 60 s after the band was released.

2.3 Results and discussion

The change of blood volume was modeled by the following single exponential function. The function was fitted to the measured ΔOD to obtain the fitting parameters (A, B, C, 1τ , and 2τ ).

( )[ ] ( ) CttBttAtOD +−−+−−−=Δ 2010fit )(exp)(exp1)( ττ , (1)

where t0 indicates the end time of occlusion (t0 = 10 s). The measured ΔOD values and the corresponding fitting lines for each SD distance are shown in Fig. 1, where the upper-arm occlusion was released at t = 10 s.

The obtained ΔOD curves which are proportional to the total hemoglobin (total-Hb) changes calculated by using the molar extinction coefficient of hemoglobin at an isosbestic point of around 800 nm, are consistent with those reported by Niwayama et al. (2006).12 From the error distribution of the measured data, we tested the validity of residual errors between the fitting line (estimated by the least-squares method) and measured ΔOD data.

A chi-square test, with significance level set to 5% (α = 0.05), was performed on the measured ΔOD data. If no significant difference is found, the model is valid under the significance level. If it is found, the model should be reviewed because either the data or the premise of the least-squares method is invalid. When the vector of the residual error (ε) between the vectors of estimated ΔOD value ( fitΔOD ) and measured ΔOD data ( ΔOD ) is defined as

ΔODΔODε fit −≡ , and standard deviation of the residual error is denoted as σ , the chi-square ( 2χ ) is expressed as

22 /σχ εεt≡ . (2)

The expected value of 2χ is equal to the degrees of freedom of the fitting line (d = 25), which is calculated from the

number of data points (30) minus the number of parameters (5), and the measured 2χ obtained by Eq. (2) has a chi-

square distribution. The reduced chi-square 2~χ , expressed as



Fig. 1. ΔOD measured on forearm at SD 10–25 mm before and after occlusion of the upper arm.

d

22~ χ

χ = , (3)

is the chi-square divided by the degrees of freedom. If the probability that the expected reduced chi-square 2~Eχ is larger

than the measured reduced chi-square ( 2~χ ) is expressed as )~~( 22 χχ ≥EP ,

)~~( 22 χχα ≥< EP (4)

becomes the requirement for validating the model. From this measurement, the measured reduced chi-square ( 2~χ ) was

obtained as 1.16 (degrees of freedom: 25) at all SD distances. Since α = 0.05 and )~~( 22 χχ ≥EP > 0.22 ( 2.1~2 =χ , degrees of freedom: 26) >α , the assumed model (Eq. (1)) was validated.

3. MEASUREMENT OF HUMAN BRAIN ACTIVITY 3.1 Participant

One male adult (30 years old) participated as a volunteer subject in this study. The subject provided written informed consent after receiving a complete explanation of the study.

3.2 Experimental setup

To apply a subtraction method to measurement data of human brain activity, the optical scanning system (described in Section 2) was used to measure brain activity at a human forehead. A block diagram for measuring a human forehead using the optical scanning system is shown in Fig. 2. The system, consisting of a galvano mirror, an 808-nm laser diode (LD), and an avalanche photodiode (APD) module, was installed over the subject’s head. The LD was driven by a laser driver connected to a function generator, which sends a 3.3-kHz amplitude-modulated signal to both the laser driver and

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0 10 20 30 40 50 60 70 80 90

ΔOD

Time (s)

SD: 10 mmSD: 13.8 mmSD: 17.5 mmSD: 21.3 mmSD: 25 mm

Occlusion ~ 60 s

Release point



the lock-in amplifier as reference signals. The intensity of the laser’s irradiation power was changed according to SD distance controlled by a controller PC. The PC sends an amplitude-control signal to a general-purpose interface bus (GPIB) controller connected to the function generator, and it also sends a galvano-control signal to a galvano controller connected to the galvano mirror. The output of the APD module is lock-in detected by a lock-in amplifier and sent to an analog-to-digital (A/D) converter. The mirror-angle signal from the galvano controller and the laser Doppler flowmetry (LDF) signals and marker signal from the stimulation presentation PC are also sent to the A/D converter. The output of the A/D converter is recorded by the recorder PC.

A total of five points (channels) were scanned at SD distances of 7, 12.5, 18, 23.5, and 29 mm (0.5-s duration for each). The phosphor cell and an optical filter (InP) were placed on the forehead above the left eyebrow near Fp1 according to the International 10–20 electrode-placement method.13 Two LDF probes were placed near the detection point (positions of the phosphor cell and InP), and microvascular blood flow in the forehead skin was measured by a laser Doppler flowmeter (MicroFlo DSP, Oxford Optronix, UK).

As a cognitive task for inducing brain activation, a verbal working-memory task14,15 was used because it has been reported that the activation region during performance of a verbal working-memory task is larger than that during a spatial working memory task15 and that the amplitude of the NIRS signal during performance of a verbal working-memory task is also larger.14,15 The sequence of the verbal working-memory task is shown in Fig. 3. As for the task, the subject was requested to look at the center of a screen. Each trial starts with a 1.5-s presentation of the target screen, which is followed by a 7-s delay. A probe screen is then presented for 2 s. In the target screen, four Japanese characters (categorized as “hiragana”) are presented. In the probe screen, only a single Japanese character (categorized as “katakana”) is presented. After the probe stimulus is presented, the subject judges whether the character presented in the probe screen has the same sound as any of the four characters presented in the target screen. The trial was repeated 16 times. The task was presented through the “Platform of Stimuli and Tasks” software developed by Hitachi, Ltd. The display for presentation of the task was located in the direction of subject’s feet. The subject saw the display through prism glasses placed over his eyes.

Fig. 2. Block diagram for measuring a human forehead using the optical scanning system. Stim. PC: stimulation-presentation PC; A/D converter: analog-to-digital converter; Lock-in amp.: lock-in amplifier; LD: laser diode; APD: avalanche photodiode. The subject wore prism glasses over his eyes so he could see the PC display in the direction of his feet, and two LDF probes were put on his forehead.

Display

Prismglasses

Scanning system(galvanomirror, LD, & APD)

LaserDopper

Lock‐inamp.

A/Dconverter

Laserdriver

Stim.PC

Markersignal

Functiongenerator

Galvanocontroller

GPIBcontroller

ControllerPC

Optical fiber

Mirror angle

Signal

RecorderPC

RS232C

USB

GPIB

Reference

Subject



+

Fig. 3. Sequence of a verbal working-memory task used for measuring brain activation. The subject was requested to look at the center of the screen during the experiment and memorize the positions of four Japanese characters categorized as “hiragana” in a target screen and judge whether a Japanese character categorized as “katakana” in a probe screen has the same sound as any of the characters in the target screen. The subject has to memorize the four characters for 8.5 s during the display of the target screen and delay screen in each trial.

The experiment was repeated three times. The SD distances at five points changed slightly (within several millimeters) at each setup. Three data sets obtained during the same task were used in the analysis, but only the waveforms from one experiment are shown in the following figures. To calculate the correlation coefficient data between the Hb signals and LDF signals, however, all three data sets were used.

3.3 Results and discussion

Because only one wavelength (808 nm) of light was used in the experiment, the change in the total-Hb was calculated under the assumption that the wavelength is approximately at the isosbestic point of the absorption coefficient of oxy- and deoxy-Hb. When the wavelength is at the isosbestic point, the change in total-Hb can be calculated by using the molar extinction coefficient and the change in optical density as

( ) ,naturalHb

total εODLC Δ

=×Δ (5)

where

196.0)10ln(commonnaturalHb =×= Hbεε mM−1･mm−1 , (6)

Rest16 – 21 s

Delay7 s

Target

Start1 2 16

End

Time

Memory holding

1.5 sProbe0 – 2 s

・・・

+のふ

ほぬヌ



naturalHbε and common

Hbε denote molar extinction coefficients based on the natural logarithm and the common logarithm,

respectively. Change in optical density ( ODΔ ) was calculated by Eq.(7) as

( )baseln IIOD −=Δ , (7)

where Ibase denotes the baseline intensity of detected light, and I denotes the intensity of detected light at a certain time point.

Block-averaged total-Hb data measured at SD distances of 12.5, 18, 23.5, and 29 mm are shown in Fig. 4. The task-related change in total-Hb was obtained at an SD distance of 29 mm. Data obtained at an SD distance of 7 mm were not used because the excitation of the light at a 7-mm SD distance was very close to that of phosphor-holding heavy paper, and a slight body movement can affect the quality of the data obtained at the shortest SD distance. For analysis of the other two experiments, the data obtained at the shortest SD distance were used for a subtraction method.

The estimated deep signal was obtained using a subtraction method with a channel with a shorter SD distance (short channel).16 In the subtraction method, the signal from a short channel is linearly fitted to that from a longer SD distance channel (long channel). The fitting calculation is conducted by minimizing the square sum of differences between long channel and scaled short channel time series.

The deep signal (subtracted data from the total Hb signal) at an SD distance of 29 mm, the shallow signal (scaled data for an Hb signal at an SD distance of 12.5 mm), and the raw total Hb signal at 29 mm (long channel) are shown in Fig. 5(a). An LDF signal for evaluating the performance of the subtraction method is shown in Fig. 5(b).

The correlation coefficients between the LDF signal and the Hb signals (raw, shallow, and deep signals) calculated using the data from three repeated experiments are shown in Fig. 6. The correlation coefficients between the LDF signal and raw data, shallow signal, and deep signal (calculated by using averaged block data for 13 data points) were 0.17 ± 0.24, 0.45 ± 0.12, and −0.04 ± 0.26, respectively.

These results demonstrated that the subtraction method is effective for extracting a task-related deep signal that is less correlated with the LDF signal than the original signal and that the optical scanning system can be used as a multi-distance NIRS system for measuring a human forehead.

Fig. 4. Block-averaged total-Hb data measured at SD distances of 12.5, 18, 23.5, and 29 mm. Solid and broken vertical lines represent start and end times of the memory-holding trials. The subtraction method was performed using total-Hb changes obtained at SD distances of 29 and 12.5 mm.

0 5 10 15 20 25 30

-0.1

-0.08

-0.06

-0.04

-0.02

0

0.02

SD:12.5mmSD:18mmSD:23.5mmSD:29mm

tota

l Hb

chan

ge (m

M・

mm

)

Time (s)



Fig. 5. (a) Deep signal (subtracted data from the total Hb signal) at SD distance of 29 mm, shallow signal (scaled data of Hb signal obtained at SD distance of 12.5-mm), and raw-total Hb signal at SD distance of 29 mm for implementation of the subtraction method; (b) LDF signal used for evaluating performance of the subtraction method. Solid and broken vertical lines represent start and end times of the memory-holding trial.

Fig. 6. Correlation coefficients (mean ± standard deviation) between the LDF signal and the raw data for a long channel, shallow signal (short channel), and deep signal. Error bars indicate the standard deviations between three repeated measurements (16 blocks for each measurement) for a single participant.

4. CONCLUDING REMARKS Optical-density change of human muscle during upper-arm occlusion and relief was measured by a new scanning system. In particular, the dependence of the optical-density change on SD distance was obtained. This result demonstrates that the optimal SD distance for oxygenation monitoring on the human forearm can be determined by the scanning system. Furthermore, the optical scanning system was used for measuring human brain activity at the forehead. The subtraction method (used for extracting deeper-layer signals) was applied to the measurement data. On the basis of the correlation with the LDF signal, it was confirmed that a deep-layer signal with lower correlation than the original signal was successfully extracted by using channels with short (12.5 mm) and long (29 mm) SD distance. This result demonstrates

0 5 10 15 20 25 30-0.05

0

0.05

Time (s)

tota

l Hb

chan

ge (m

Mm

m)

Deep signalShallow signalRaw data (long channel)

(a)to

tal H

bch

ange

(mM・

mm

)

Raw data (long channel)Shallow signal (short channel)Deep signal

0 5 10 15 20 25 30-3

-2

-1

0

1

2

3

Time (s)

LDF

signa

l cha

nge

(arb

. uni

t)

(b)

-0.4

-0.2

0

0.2

0.4

0.6

Raw data(long channel)

Shallow signal(short channel)

Deep signal

Cor

rela

tion c

oeffi

cien

t (vs

. LD

F)

0.17±0.24

0.45±0.12

‐0.04±0.26



that the subtraction method is effective and that the optical scanning system can be used as a multi-distance NIRS system for measuring the human forehead.

If the detector of the non-contact system is replaced with a high-sensitivity charge-coupled-device (CCD) sensor, the degree of freedom at the interface between instruments and human subjects will increase and the application range of this technique will be broadened. For example, the lack of pressure on the skin can be utilized in long-term brain-activity monitoring during sleep, cognitive-state monitoring during automobile driving, and psychological- and physiological-state monitoring during office work. Moreover, this technique may also be used for non-contact monitoring of infants.

REFERENCES

[1] Jöbsis, F. F., “Noninvasive, infrared monitoring of cerebral and myocardial oxygen sufficiency and circulatory parameters,” Science, 198(4323), 1264–1267 (1977).

[2] Hoshi, Y. and Tamura, M., “Detection of dynamic changes in cerebral oxygenation coupled to neuronal function during mental work in man,” Neurosci. Lett., 150(1), 5–8 (1993).

[3] Maki, A., Yamashita, Y., Ito, Y. et al., “Spatial and temporal analysis of human motor activity using noninvasive NIR topography,” Med. Phys. 22(12), 1997–2005 (1995).

[4] Ferrari, M. and Quaresima, V., “A brief review on the history of human functional near-infrared spectroscopy (fNIRS) development and fields of application,” NeuroImage, 63(2), 921–35 (2012).

[5] Koizumi, H., Yamamoto, T., Maki A. et al., “Optical Topography: Practical Problems and New Applications,” Appl. Opt., 42(16), 3054–3062 (2003).

[6] Hoshi, Y., “Functional near-infrared spectroscopy: current status and future prospects,” J. Biomed. Opt., 12(6), 062106 (2007).

[7] Okada, E., Firbank, M., Schweiger, M. et al., “Theoretical and experimental investigation of near-infrared light propagation in a model of the adult head,” Appl. Opt., 36(1), 21–31 (1997).

[8] Mazurenka, M., Jelzow, A., Wabnitz, H. et al., “Non-contact time-resolved diffuse reflectance imaging at null source-detector separation,” Opt. Express, 20(1), 283–290 (2012).

[9] Funane, T., Atsumori, H., Kiguchi, M. et al., “Optical scanning system for light-absorption measurement of deep biological tissue,” Rev. Sci. Instrum., 82(9), 093101 (2011).

[10] Funane, T., Atsumori, H., Suzuki, A. et al., “Noncontact brain activity measurement system based on near-infrared spectroscopy,” Appl. Phys. Lett., 96(12), 123701 (2010).

[11] Funane, T., Atsumori, H., Suzuki, A. et al., “Noncontact optical brain activity measurement system using phosphor placed on skin,” Jpn. J. Appl. Phys., 50(7), 077001 (2011).

[12] Niwayama, M., Murata, H., and Shinohara, S. “Noncontact tissue oxygenation measurement using near-infrared spectroscopy,” Rev. Sci. Instrum., 77(7), 073102 (2006).

[13] Jasper, H. H. “The ten twenty electrode system of the International Federation,” Electroencephalogr. Clin. Neurophysiol., 10, 371–375 (1958).

[14] Aoki, R., Sato, H., Katura, T. et al., “Relationship of negative mood with prefrontal cortex activity during working memory tasks: An optical topography study,” Neurosci. Res., 70(2), 189–96 (2011).

[15] Sato, H., Aoki, R., Katura, T. et al., “Correlation of within-individual fluctuation of depressed mood with prefrontal cortex activity during verbal working memory task: optical topography study,” J. Biomed. Opt., 16(12), 126007 (2011).

[16] Saager, R. B., Telleri, N. L., and Berger, A. J., “Two-detector Corrected Near Infrared Spectroscopy (C-NIRS) detects hemodynamic activation responses more robustly than single-detector NIRS,” NeuroImage, 55(4), 1679–85 (2011).



Documents

SPIE Proceedings [SPIE SPIE BiOS - San Francisco, California, USA (Saturday 2 February 2013)] Optical Tomography and Spectroscopy of Tissue X - Near-infrared spectroscopy system with