Near-IR spectroscopy (NIRS), which uses light in the NIR range (700 to 900 nm), enables detection of changes in the oxygenation state of hemoglobin (Hb) and water in biological tissues. Several studies have shown the usefulness of NIRS for noninvasive measurement of cerebral blood volume and oxygenation in infants.1 2 3 4 5 6 7
A number of approaches for measuring cerebral Hb concentration and oxygen saturation using several types of NIRS systems have been proposed. These approaches include (1) continuous wave spectroscopy (CWS), by which only the relative value of Hb can be estimated;8 (2) full spectral spectroscopy (FSS), by which the NIR full spectrum of the NIR range can be measured;9 10 11 12 (3) time-resolved spectroscopy (TRS), by which the transit time of each photon through the tissue of interest can be measured;13 14 15 16 17 18 (4) phase-modulated spectroscopy (PMS), by which amplitude signals for phase, intensity and depth of modulation after passage can be measured;19 20 21 22 23 24 25 and (5) spatially resolved spectroscopy, by which the slope of light attenuation versus distance is determined at a distant point from the source.26
A new TRS system that is portable and has a high data acquisition rate was recently reported.18 27 28 A TRS device enables quantitative analysis of light absorption and scattering in tissue using the photon diffusion theory. The absorption coefficient (μa) represents the physiological state, particularly the Hb concentration and oxygen saturation, and the reduced scattering coefficient (μs ′) represents the structural change in tissue. However, for application to tissue measurements, there is the problem concerning how to determine the background absorption for estimation of Hb in the head. Hueber et al.24 reported the use of fixed weighted averages of SaO 2 and SvO 2 for estimation of background tissue absorption using PMS. However, an exact background absorption has not been reported.
In this paper, we first investigate whether our TRS measurements could be used to accurately measure Hb concentration and oxygen saturation in an in vitro model such as the brain. Next, we apply this method to newborn piglets during hypoxic loading. In this experiment, we use two different methods to determine Hb concentration and oxygen saturation in the head using μa values measured at three wavelengths. One method uses a water-only background absorption to determine the Hb concentration, and we investigate the relationships of cerebral Hb oxygenation with arterial and sagittal sinus venous Hb oxygen saturation at various concentrations of inspired oxygen using a piglet hypoxic model. Another method uses background absorption from the piglet head with blood exchange by fluorocarbon (FC), and we reestimate cerebral Hb concentration for comparison to that with water-only background absorption calculations. Then the reliability of the instruments is evaluated. The results of optical path length obtained in this study were previously published.27
Materials and Methods
In Vitro Experimental Protocol
To provide a model for in vivo measurement of Hb in the piglet brain, we filled a 1.000-mL glass beaker with 0.1-M phosphate buffer (pH 7.4) containing 0.7% (w/vol) dry yeast in 1% intralipid and glucose (2% w/vol) to produce a solution with a light scattering condition equivalent to that of blood-perfused brain tissue. A magnetic stirrer was used to prevent particle settling. The measurement receiver fiber was set at a distance of 30 mm from the light source fiber, and the fibers were immersed directly in the suspension. The oxygen content was also monitored at the same time with a conventional gas analyzer (ABL5, Radiometer, Copenhagen, Denmark). In the first stage of the experiment, human blood was added gradually to the beaker to achieve blood volume ranging from 0 to 100 μM, and the resulting changes in the reemission profiles were recorded. Each model was bubbled with oxygen. In the second stage of the experiment, we used a model solution with a 80-μM Hb suspension, and the oxygen concentration in the solution was reduced to simulate anoxia in the brain. A mixture of O 2 and N 2 gases was prepared and bubbled through the suspension to manipulate the oxygen saturation of Hb from 0 to 100%.
Once μa had been determined at three wavelengths, the oxyHb and deoxyHb concentrations were calculated from the absorption coefficients of oxyHb and deoxyHb using the same law as that described in Sec. 2.4 with the assumption that background absorption is due to only 100% (by volume) water and that the refractive index of water is 1.3.
Animal Study Protocol
Eleven newborn piglets, less than 48 h old and weighing 1.9 to 2.4 kg, were each anesthetized with an intramuscular injection of sodium pentobarbital (2 mg/kg). The umbilical vein and artery of each piglet were cannulated. After cannulation, each piglet was paralyzed with pancronium bromide at an initial dose of 0.1 mg/kg followed by infusion at 0.1 mg/kg h−1 and then anesthetized with fentanyl citrate at an initial dose of 10 μg/kg followed by infusion at 5 μg/kg h−1. The animals were then intubated and mechanically ventilated with an infant ventilator. A 24-gauge cannula was inserted into the sagittal sinus through a burr hole in the anterior fontanel for venous blood sampling.
The umbilical artery was used for blood pressure monitoring and arterial blood sampling. Infusion of maintenance solution was continued at a rate of 4 mL/kg h−1 via the umbilical vein. Respiration and acid-base balance were checked by arterial blood gas analysis at each step of FI O 2 . A negative base excess lower than 5.0 mmol/L, which was caused by hypoxia, was corrected as much as possible by sodium bicarbonate infusion to maintain pH between 7.3 and 7.5, thus minimizing the shift in the Hb dissociation curve. The hematocrit of arterial blood was in the normal range, and mean arterial pressure was maintained above 60 mm Hg except near zero FI O 2 . Arterial Hb oxygen saturation (SaO 2) and sagittal sinus venous Hb oxygen saturation (SvO 2) were measured using a blood co-oximeter (OSM-3, Radiometer, Copenhagen, Denmark). To obtain aerobic and anaerobic conditions, FI O 2 was changed by mixing pure nitrogen and oxygen gases with continuous monitoring by an oxygen electrode. We first set FI O 2 at 27% and then at 100% to obtain hyperaerobic conditions. Then, to obtain anaerobic conditions, FI O 2 was decreased in steps from 100 to 4%. The hair on the scalp of the parietal region was removed with an epilating agent. Two optical fibers were brought into contact with the skin on the head of each piglet. Throughout the experiment, rectal temperature was monitored and maintained between 38.0 and 39.0 °C using a heated-water mattress. At the end of the experiment, animals were sacrificed with intravenous pentobarbital (100 mg/kg). The protocols for animal care were in compliance with institutional guidelines.
Near-IR TRS System and Analysis
We used a portable three-wavelength TRS system (TRS-10, Hamamatsu Photonics K.K., Japan) and attached a probe to the head of each piglet with a light source fiber and light detector fiber separation of 30 mm. In the TRS system, a time-correlated single-photon-counting technique is used for detection. The system is controlled by a computer through a digital I/O interface consisting of a three-wavelength (761, 795, and 835 nm) picosecond light pulser (PLP) as the pulsed light source, a photon-counting head for single photon detection, and signal-processing circuits for time-resolved measurement. The PLP emits NIR light with a pulse duration of about 100 ps and an average power of at least 150 μW at each wavelength at a repetition of 5 MHz, and the input light power to the subject was around 20 μW because of the loss in an incident optical system of the TRS system.
The light from the PLP is sent to a subject by a source fiber with a length of 3 m, and the photon reemitted from the subject is collected simultaneously by a detector fiber bundle with a length29 of 3 m. The light source fiber was a graded index (GI) type single fiber (GC200/250L Fujikura, Japan) with a numerical aperture (NA) of 0.25 and a core diameter 200 μm, and the light detector fiber was a bundle fiber with a diameter of 3 mm and NA of 0.21. Finally, a set of histograms of the photon flight time, which is called a reemission profile, is recorded. One temporal reemission profile includes 1024 time channels spanning about 10 ns with a time step of about 10 ps. In this study, the emerging light was collected over a period of 3 s to exceed at least 1000 photon counts at the peak channel of the reemission profiles, and the procedure was repeated 30 times for each measurement.
The instrumental response was measured with the input fiber placed opposite the receiving fiber through a neutral density filter. The instrumental response of the TRS system was around 150 ps FWHM at each wavelength.
The mean optical path length was calculated from the difference between the center of gravity of the measured reemission profile and that of the instrumental response. It was assumed30 that the value for the refractive index of brain tissue is 1.4.
The reemission profiles observed at each measurement point were fitted by the photon diffusion equation proposed by Patterson et al.13 in order to calculate the values of absorption coefficient (μa) and reduced scattering coefficient (μs ′) of the head at wavelengths18 of 761, 795, and 835 nm. In the fitting procedure, a nonlinear least-squares fitting method based on the Levenberg-Marquardt method was used. This is a method of iterative improvement. In each iterative calculation, the function from the photon diffusion equation in reflectance mode, which was convoluted with the instrumental response, was fitted to the observed reemission profile. The calculation regions were determined to include the observed profile data and the data of 600 channels were included into the fit.
Estimation of Absolute Values of oxyHb and deoxyHb with Water-Only Background Absorption
After determination of the values of μa at three wavelengths, the oxyHb and deoxyHb concentrations were calculated from the absorption coefficients of oxyHb and deoxyHb using the following equations with the assumption that background absorption is due only to 85% (by volume) water.9 11ɛλ nm x is the extinction coefficient at λ nm, and [oxyHb] and [deoxyHb] are the concentration.
First, the water absorption was subtracted from μa at each of the three wavelengths, and then the values of oxyHb and deoxyHb were estimated using the least-squares fitting method. The absorption coefficients for oxyHb, deoxyHb, and water shown in Table 1 were used.
|Absorption coefficients for oxyHb, deoxyHb, water, and absorption of the piglet head with blood exchange transfusion by FC.|
|oxyHb (mM−1 cm−1)||deoxyHb (mM−1 cm−1)||Water (cm−1)||Piglet Head with Blood Exchange|
Transfusion by FC (cm−1)
The cerebral total hemoglobin (total Hb) concentration and cerebral Hb oxygen saturation (SO 2) were calculated as follows:
Estimation of Absolute Values of oxyHb and deoxyHb with Absorption of the Piglet Head with Blood Exchange Transfusion by FC
For measurement of background absorption of the head without Hb, we used a piglet for blood exchange transfusion by FC. First, partial blood exchange was done by saline and then blood was replaced with perfluorotributylamine emulsion (FC43 Emulsion, Green Cross Corp., Japan) as an oxygen carrier. We used 500 mL of FC and the hematocrit was lowered to 0.5%. After replacing the blood with FC, oxygenated head TRS measurements were done while the piglet breathed 100% O 2. The EEG showed normal activity during the period of ventilation with 100% O 2. We used the μa of piglet head with blood exchange transfusion by FC at three wavelengths as the background absorption value. The values at 761, 795, and 835 nm were estimated to be 0.1190, 0.1008, and 0.1260, respectively (Table 1). The following equations were used for estimation of oxyHb and deoxyHb concentrations:μa λ nm background is the background absorption derived from the piglet head with blood exchange transfusion by FC. First, the absorption of the head with blood exchange transfusion by FC (abs of the head with BET by FC) was subtracted from μa at each of the three wavelengths, and then the values of oxyHb and deoxyHb were estimated using the least-squares fitting method.
A Statview J 4.5 package for the Macintosh computer was used for statistical analysis. Values obtained at different levels of FI O 2 were compared from values at FI O 2 of 21% by Wilcoxon’s signed-rank test. The level of statistical significance was set at a probability of p<0.05 for all tests. Results at FI O 2 levels in the range 10 to 100% were obtained from measurements in 11 piglets, results at FI O 2 of 8 and 6% were obtained from 10 piglets and results at FI O 2 of 4% were obtained from 8 piglets. Values are expressed as means ± standard deviation (SD).
In Vitro Experiment Results
Figure 1(a) shows the relationship between the actual blood volume and total Hb concentration of the model as measured by TRS. The relationship between the actual blood volume (X) and Hb concentration measured by TRS (Y) was Y=0.843X+2.21 (r=0.998, P<0.001). We can see that the relationship between these two parameters is linear over the experimental range of Hb concentrations. Figure 1(b) shows the relationship between the values of Hb oxygen saturation in the 80-μM Hb suspension measured by a conventional gas analyzer or TRS. The relationship between the values of Hb oxygen saturation measured by a conventional gas analyzer (X) and TRS (Y) was Y=0.960X+0.889 (r=0.991, P<0.001). Linear regression analysis demonstrated that a significant correlation existed between the respective values determined by the two methods. These results indicated that our TRS method enables accurate and noninvasive measurement of Hb content and oxygen saturation in a highly scattered medium such as the piglet brain.
Values of μs ′ and μa
In Fig. 2, the values of μs ′ measured concurrently at corresponding FI O 2 levels are plotted against FI O 2 . At FI O 2 of 21%, the values of μs ′ of the newborn piglet brain at 761, 795, and 835 nm were estimated to be 8.70±1.23, 8.33±0.91, and 8.24±1.11/cm, respectively. There were no significant differences between the values of μs ′ for each FI O 2 level in the range of 4 to 100% at all three wavelengths.
In Fig. 3, the values of μa ′ measured concurrently at corresponding FI O 2 levels are plotted against FI O 2 . The values of μa at 761 and 795 nm at FI O 2 levels below 15% were higher than the values at FI O 2 of 21% at each wavelength. When FI O 2 was decreased from 21 to 15%, the values of μa at 761 and 795 nm increased from 0.227±0.021 to 0.259±0.024/cm and from 0.189±0.016 to 0.203±0.013/cm, respectively. The values of μa at 835 nm at FI O 2 levels below 10% were higher than those at an FI O 2 level of 21%. When FI O 2 was decreased from 21 to 10%, the value of μa at 835 nm increased from 0.223±0.020 to 0.251±0.025/cm.
Concentration of oxyHb and deoxyHb with Water-Only Background Absorption
As shown in Fig. 4(a), the relationship between oxyHb and deoxyHb changed reciprocally in a mirror-image manner when FI O 2 was changed. Total Hb showed an increase of 31.3% when FI O 2 was decreased from 21 to 6%, indicating that cerebral blood volume increased during hypoxia. The values of oxyHb, deoxyHb, and total Hb at FI O 2 of 21% were 53.5±8.1, 33.4±5.8, and 86.9±8.1 μM, respectively.
In Fig. 5(a), levels of SO 2, SaO 2, and SvO 2, which were measured concurrently at corresponding FI O 2 levels, are plotted against FI O 2 . At FI O 2 of 21%, the mean SO 2 of newborn piglets was calculated to be 61.5±6.3%. The values of SaO 2 and SvO 2 were simultaneously determined to be 92.1±3.76 and 40.1±6.4%, respectively. The contributions of SaO 2 and SvO 2 to SO 2 were 41.1 and 58.9%, respectively. However, values of SO 2 at FI O 2 in the range of 10 to 4% were higher than those of SaO 2.
The SaO 2 values also showed a positive linear relationship with SO 2 [Fig. 6(a)]. The relationship between SaO 2 and SO 2 was SO 2=37.9×10−2×SaO 2+27.5 (r=0.862, P<0.001). From this regression equation, the values of SaO 2 and SO 2 were estimated to be 44.3% when SO 2 was larger than SaO 2 during a hypoxic condition.
As shown in Fig. 7(a), there was a positive linear relationship between SvO 2 and SO 2. The relationship between SvO 2 and SO 2 was SO 2=69.3×10−2×SvO 2+31.8 (r=0.841, P<0.001).
Concentration of oxyHb and deoxyHb with Absorption of the Piglet Head with Blood Exchange Transfusion by Fluorocarbon
As shown in Fig. 4(b), the relationship between oxyHb and deoxyHb changed reciprocally in a mirror-image manner when FI O 2 was changed. The values of oxyHb, deoxyHb, and total Hb at FI O 2 of 21% were 25.5±7.3, 18.6±5.8, and 44.1±7.4 μM, respectively. The values of oxyHb, deoxyHb, and total Hb are significantly lower at each FI O 2 level than those in the case of water-only background absorption.
In Fig. 5(b), levels of SO 2, SaO 2, and SvO 2, which were measured concurrently at corresponding FI O 2 levels, are plotted against FI O 2 . At FI O 2 of 21%, the mean SO 2 of newborn piglets was calculated to be 57.6±12.3%. The contributions of SaO 2 and SvO 2 to SO 2 were 33.7 and 66.3%, respectively. SO 2 values were between those of SaO 2 and SvO 2 at all levels of FI O 2 . SO 2 values are not significantly different from SO 2 values using the water-only background at FI O 2 in the range of 15 to 100%, but the values are lower than the values using the water-only background at FI O 2 in the range of 4 to 12%.
The SaO 2 values also showed a positive linear relationship with SO 2 [Fig. 6(b)]. The relationship between SaO 2 and SO 2 was SO 2=59.0×10−2×SaO 2+4.9 (r=0.843, P<0.001). As shown in Fig. 7(b), there was a positive linear relationship between SvO 2 and SO 2. The relationship between SvO 2 and SO 2 was SO 2=108.6×10−2×SvO 2+11.5 (r=0.829, P<0.001).
Cerebral Hb content and Hb oxygen saturation in infants have already been estimated using NIRS in several studies. However, the values estimated in those studies cannot be compared easily because the same phantom models, animal models, and experimental protocols were not used in those studies. Although the ultimate purpose of this study was to estimate SO 2 and Hb content in infants, cerebral Hb content and SO 2 have already been estimated by NIRS in previous studies using an experimental animal model, i.e., the newborn piglet, which approximates infants. The NIRS used in those studies can be divided into three categories: (1) FSS (Refs. 11 and 12), (2) PMS (Refs. 20, 22, 23, 25, and 31), and (3) TRS (Ref. 17). Therefore, we used newborn piglets with hypoxia for our study to investigate the validity of NIR TRS. This model is easy to reproduce and is useful for comparison with results obtained by using other methods.
In the in vitro study, the use of intralipids and blood phantom showed that qualitative measurements of Hb concentrations and of oxygen saturation could be made under the same conditions as those in an in vivo study. The Hb concentration used was approximately the same as the Hb molar concentration in the human brain. Even if the dispersion of light is the same as that for a human brain, quantitative analysis of Hb concentration and oxygenation is considered possible. In practice, however, there is a substantial difference between this model and noninvasive measurement of head structure: the head has a multiplex structure, and blood vessels have an inhomogenous structure.
The application of an analytical solution of the diffusion equation assumes a homogeneous medium, but the piglet head consists of multilayers including the scalp, skull, cerebrospinal fluid (CSF) layer, gray matter, and white matter. The optical fibers were attached to the skin. Therefore, the obtained optical properties are not those of only the brain but the average of those of the multiple layers.
Kienle et al.32 33 investigated analytical solutions to the two-layered optical phantom model using the PMS method. They demonstrated that a superficial layer of 4 mm in thickness (or less) will have only a small influence on the accuracy of the measured optical properties of an underlying thick layer. In the piglets used in our study, the thickness of the scalp and skull layer measured postmortem were about 1.5 and 1.5 mm, respectively, similar to values obtained in an earlier study.24 25 Huber et al.24 and Fantini et al.25 concluded that in the presence of a relatively thin (⩽4 mm) scalp/skull layer (as in the case of the neonatal piglet), those absolute values of Hb concentration and oxygen saturation should be actually representative of the brain. In contrast, the significantly thicker tissue inhomogeneities found in the adult human head have an effect on the absolute optical measurements.
Furthermore, with regard to the effect of the CSF layer on photon migration, Fukui et al.34 estimated light properties from results of Monte Carlo simulation using the neonatal head model. In their model, the thicknesses of the scalp, skull, and CSF layer were about 2, 2, and 1 mm, respectively, almost the same as the values in our piglet model. A photon that has penetrated into the white matter can be identified by a detector 10 mm from the source. The partial DPF for the white matter proportionally increased with the increase in source-detector spacing, whereas that for the gray matter was almost constant at source-detector spacing greater than 30 mm. At source-detector spacing of 30 mm, the DPF was almost 4.3 and the values of partial DPF of gray matter and white matter were 1.3 and 0.7, respectively. Fukui et al.34 reported that the intensity sensitive region in the neonatal brain is confined to the gray matter, but the spatial sensitivity profile penetrates into the deeper region of the white matter. Therefore, our TRS measurements of the piglet head enable estimation of Hb of the brain, which is a multilayer structure, and the contribution of the brain to the measurement signal is estimated to be approximately 50%. If the contributions of the scalp, skull, and CSF layers of the measurement signal are assumed to be 10, 20, and 20%, respectively,35 and μa values of the scalp, skull, and CSF at 795 nm are assumed to be 0.18, 0.16, and 0.04/cm, respectively,34 the μa of the brain is estimated to be 0.24/cm. This value is lower than that of the human neonatal brain gray or white matter.32
There were no significant differences between the values of μs ′ at each wavelength for FI O 2 levels in the range of 4 to 100%. These results are similar to those obtained in a study by Zhang et al.,36 showing that scattering changes detected by a frequency-domain oximeter were associated only with asphyxia and death. Yamashita et al.37 reported the results of a preliminary study on light scattering in the piglet brain by using TRS. Their results showed that the values of μs ′ in piglet brains were around 1.3 mm−1 in a state of normoxia and that the value of μs ′ showed a notable decrease after death. Tissue edema and structural changes occur during severe hypoxia, particularly at and after death, and values of μs ′ are thought to change only during structural changes in tissue due to cerebral energy failure. Therefore, there were no significant differences between the values of μs ′ during hypoxia in this study.
The values μa at 761 and 795 nm at FI O 2 levels below 15% were significantly higher than those at FI O 2 of 21% at each wavelength because deoxyHb increased during hypoxia. The values of μa at 835 nm at FI O 2 levels below 10% were higher than those at an FI O 2 of 21%. This is because of an increase in both oxyHb and deoxyHb absorption, indicating an increase in cerebral blood volume during hypoxia.
In this paper, at levels of FI O 2 in the range of 10 to 4%, the values of SO 2 using a water-only background were higher than those of SaO 2 [Fig. 4(a)]. These results are thought to be mainly due to the effect of background absorption in tissue.24 In the brain, Hb, water, and cytochrome c oxidase are known chromophores in the NIR region,38 but the concentrations vary between individuals and the in vivo absorption spectra of cytochromes are difficult to measure precisely. It is also likely that there are other important chromophores, such as fats, that make a significant contribution to non-Hb absorption. Indeed, Hueber et al.24 reported that using the frequency-domain multidistance method in the piglet head, SO 2 is greater than SaO 2 under the condition of extreme hypoxia with the assumption of a water-only background. They estimated the background tissue absorption with the assumption that SO 2 is equivalent to a mixture of 50% SaO 2 and 50% SvO 2 during normoxia and mild hypoxia to improve the estimation of brain Hb.
In this paper, we estimated the background absorption from the piglet head with blood exchange transfusion by FC. Using this absorption, the Hb values were lower than those using a water-only background, especially at a low level of FI O 2 . The SO 2 values obtained using abs of the head with BET by FC were between the values of SaO 2 and SvO 2. These values were more accepted from a physiological point of view.
The mean SO 2 in 11 newborn piglets at FI O 2 of 21% was calculated to be 58% using FC blood exchanged piglet head absorption. The values of SaO 2 and SvO 2 were simultaneously determined to be 92 and 40%, respectively. The contributions from arterial blood and venous blood were estimated to be 34 and 66%, respectively. The ratio of the contribution of arterial blood to that of venous blood is almost the same as the ratios reported by Brun et al.39 and Kusaka et al.12 Thus, the validity of measurement in vivo can be verified, even in an inhomogenous piglet head. Table 2 shows some values of cerebral Hb content and SO 2 in newborn piglets measured by using NIRS. Previously reported values of10 11 17 22 23 24 25 SO 2 in newborn piglets range from 48 to 84%, a range similar to our results.
|Values of cerebral Hb oxygen saturation (SO 2) and Hb concentration (total Hb) in the piglet head in a state of normoxia measured by using near-infrared spectroscopy.|
|Age||SO 2 (%)||Total Hb (μM)||References|
|FSS||7||less than 24 h||65.4±3.5||37.5±9.1||Springett et al.11|
|FSS||5||5 to 6 weeks||48.4±3.2||Sakamoto et al.10|
|FSS||8||less than 24 h||63±4||Kusaka et al.12|
|PMS||60||5 to 10 days||68±5||Kurth et al.23|
|PMS||12||1 to 2 days||84||Du et al.22|
|PMS||4||58.5±2||63±14||Hueber et al.24|
|PMS||3||11±1 days||60||42||Fantini et al.25|
|TRS||6||2 to 5 days||138±31||Kurth and Uher16|
|TRS||1||2 weeks||65||110||Yamashita et al.17|
|TRS||11||less than 48 h||57.6±12.3||44.1±7.4||Ijichi et al. (this paper)|
|FSS, full spectral spectroscopy; PMS, phase-modulated spectroscopy; TRS, time-resolved spectroscopy.|
In conclusion, the results of this study demonstrated that our NIR TRS method can be used to monitor cerebral Hb oxygen saturation and Hb content in a piglet model, especially using abs of the head with BET by FC. However, more studies are required for estimation of true background absorption and for developing and validating the instrument for clinical use.
We thank Mr. M. Oda, Mr. Y. Yamashita, and Hamamatsu Photonics, K.K., for their technical support in this study. This study was supported by grants-in-aid for scientific research (C), Nos. 13671140, 14571051, and 15591159, and encouragement of young scientists, Nos. 13770619 and 14770571, from the Ministry of Education, Culture, Sports, Science and Technology of Japan.