1. Introduction

The development of multi-channel instruments based on near-infrared spectroscopy (NIRS) will provide a noninvasive and simple way for studying brain and muscle functions. Maki et al. introduced optical topography as a practical technique for investigation of brain function based on continuous-wave spectroscopy (NIR_CWS) [1]. This method, utilizing arrays of light sources and detectors, enables two-dimensional functional imaging of brain oxygenation and has been clinically applied to the investigation of brain function [1–3].

Multi-channel instruments based on NIRS have also been used to analyze temporal and spatial changes in muscle oxygenation, perform topographic imaging of muscle oxygenation, evaluate cooperative contraction of muscles and investigate hemodynamic changes in muscles in response to different exercises [4–15]. The results of these studies are potentially useful for studying muscle performance in sports science and for examining the restoration of muscle function in rehabilitation medicine. Among these instruments, instruments based on NIR_CWS have been applied to functional imaging of lower extremity muscles, and regional differences in muscle oxygenation in the calf and thigh muscles during various exercises were observed [7–9, 15]. These studies demonstrated the importance and feasibility of imaging techniques based on NIRCWS for studying muscle function. Despite having the advantage of being simple in design, instruments based on NIR_CWS provide only relative values of muscle oxygenation from rest.

Measurement of absolute values of muscle oxygenation is important because it enables direct and quantitative comparison of measurement results among exercising muscles and between different subjects. Instruments based on phase modulation spectroscopy (NIR_PMS) and time-resolved spectroscopy (NIR_TRS) that enable measurement of absolute values have been used to perform single-point measurement [16], multi-point measurement [11, 13, 14] and functional imaging of muscle oxygenation and tissue oxygen saturation [10, 12]. For example, Maris et al. reported the first in vivo imaging of muscle oxygenation in the forearm using NIR_PMS, in which a pair of source-detector light guides was mechanically moved by an X-Y-Z scanner [10]. A multi-channel real-time instrument based on NIR_PMS that can be package as a portable unit was first designed by Fantini et al. [11] and it was utilized by Wolf et al. for imaging of muscle oxygenation in patients with peripheral vascular disease [12]. Instruments based on NIR_PMS and NIR_TRS require sophisticated instrumentation and are mostly designed for clinical measurements at the bedside such as monitoring of brain function [17, 18] or detection of breast tumors [19, 20].

Instruments based on spatially resolved spectroscopy (NIR_SRS), which also enable measurement of absolute values, assuming that the reduced scattering coefficient of muscle is known and constant, have the advantages of portability, simplicity in design and low cost compared to instruments based on NIR_PMS and NIR_TRS. The NIRSRS approach is also relatively insensitive to motion artifacts compared to single distance measurement instruments and it subtracts the influence of superficial layers of tissue [21]. Although single-channel instruments based on NIR_SRS are now commercially available and have been used to investigate temporal changes in O₂ saturation in the brain, forearm muscle and calf muscle [22–24], there has been no report on the development of imaging instruments based on NIR_SRS.

The purpose of this study was to construct an imaging instrument based on NIR_SRS that enables temporal and spatial analyses of muscle oxygenation. Its performance was evaluated using tissue-equivalent phantoms and in vivo measurements.

2. Materials and methods

2.1 Development of an instrument for imaging of muscle oxygenation based on NIRSRS

An imaging instrument based on NIR_SRS that enables temporal and spatial analyses of muscle oxygenation was designed. At present, the instrument can be connected to a maximum of 32 compact and separate-type optical probes. Using the time-multiplexing technique, a sampling rate of 250 ms is achieved when 16 probes are used. As shown in Fig. 1(a), the probe was fitted with a light-emitting diode (LED) (95010, Optrans Corp., Japan) that consists of six LED elements: two elements of 830 nm (FWHM=35 nm) and four elements of 770 nm (FWHM=30 nm) in peak wavelengths. The peak power of the LEDs is 2 mW. Silicon photodiodes (S2386-18K and S2386-45K, Hamamatsu Photonics, Japan) with high sensitivity in the near-infrared wavelength range and low dark current were used. To increase the accuracy of measurement, photodiodes were placed at 20 mm, 25 mm and 32 mm from the LED to measure the slope of intensity of backscattered light from tissue. In addition, taking into account the steep decrease in intensity of backscattered light with increase in pathlength, photodiodes with total effective areas of 1.2 mm2, 2.4 (1.2 mm²×2) mm² and 17.9 mm² were placed at 20 mm, 25 mm and 32 mm, respectively, from the LED.

As shown in Fig. 1(b), the LED, photodiodes, and small-outline package low-noise operational amplifiers (Op-amp) (OP177, Analog Devices, USA) for current-to-voltage (I–V) converters were mounted on a three-layered printed circuit board (PCB). They were molded in room temperature vulcanizing (RTV) white silicone rubber (KE-17, Shin-etsu Chemical, Japan). The surface layer that is in direct contact with the skin was made of black silicone rubber with a frosted surface to absorb diffusing light from the skin surface. The black silicone rubber was prepared by adding black toner powder of a laser printer to the white silicone rubber.

A block diagram of the imaging instrument is shown in Fig. 2(a). The instrument consists of a primary control box connected to a personal computer (PC) via a PC-card, probe-connecting boxes and optical probes. The data acquisition PC-card (DAQCard-6024E, National Instruments, USA), consisting of digital input/output channels (DIO), digital-to-analog converters (DAC), and analog-to-digital converters (ADC), functions as an interface between the imaging instrument and the PC. The primary control box and probe-connecting boxes consist of the following circuits. (i) A multiplexer (MPX) array in the primary control box selects the corresponding wavelengths in the LED and the probe-connecting boxes. The MPX array in the probe-connecting box selects the corresponding probes. This selection is controlled by the DIO. (ii) The output of the DAC determines the illumination intensities of the two-wavelength LED. The light intensity of the LED was adjusted prior to the start of measurement to obtain an appropriate signal level of backscattered light from tissue. (iii) An I–V converter, placed just behind the photodiode (Fig. 1(b)), converts the current signal from the photodiode to a corresponding voltage level. The signal is then fed to the primary control box and is sent to the ADC. The total gain from the I–V converter output to the ADC is 48–54 dB, depending on the distance of the photodiodes from the LED.

As shown in Fig. 2(b), to reduce the number of wires, the eight probes are connected to a probe-connecting box (0.25 kg) and the latter is connected to the primary control box (2.0 kg) using flexible shielded-cable. The use of the wearable probe-connecting box enables measurements in which mobilization over a certain distance (~5 m) is required.

 

Fig. 1. A separation-type optical probe of the imaging instrument based on spatially resolved spectroscopy. (a) Top view (LED: light-emitting diode) and (b) cross-sectional view of a probe (PCB: printed circuit board). The operational amplifiers (Op-amp) were placed just behind the photodiodes to reduce external noise.

Download Full Size | PDF

 

Fig. 2. An imaging instrument based on spatially resolved spectroscopy. (a) A block diagram of the instrument. A probe-connecting box can be connected to eight probes (1 set), MPX: multiplexer; LED: light-emitting diode; DAC: digital-to-analog converter; DIO: digital input/output channel; ADC: analog-to-digital converter. (b) Appearance of the instrument. Only two probe-connecting boxes and eight probes are shown. The separation-type probes provide flexibility in their arrangement on different sites, enabling simultaneous measurements within a muscle or between several exercising muscles.

Download Full Size | PDF

The instrument control, data acquisition and user-interface programs were coded by Microsoft Visual C++ 6.0. Two types of program are available to meet different measurement requirements: (i) an imaging program that displays changes in spatial distribution of muscle oxygenation when multiple probes are utilized for measurement (temporal resolution =250 ms) and (ii) a tracing program that shows the temporal profile of muscle oxygenation when a single probe is used (temporal resolution =50 ms). The temporal resolution includes the time for data acquisition and that for calculation of oxygenation.

2.2 Algorithm for calculation of muscle oxygenation and tissue oxygen saturation

Conventional NIR_SRS algorithms are based on the assumption that tissue is homogeneous. However, the presence of overlying tissue such as a fat layer decreases the measurement sensitivity since the absorption coefficient µ_a of the fat layer is about one order of magnitude smaller than that of muscle [25]. Therefore, an algorithm for correcting the influence of adipose tissue for NIR_SRS measurements was utilized [26]. We analyzed the influence of the fat layer using Monte Carlo simulation for NIR_SRS, the model of which consisted of three layers, skin, fat and muscle layers, and obtained the relationship between slope of intensity of backscattered light S and µ_a of muscle for various fat layer thicknesses. Absorption coefficients µ_a of 0.02 and 0.002 mm^-1 and µ′_s of 1.3 and 1.2 mm^-1 were used for the skin [27] and fat [28], respectively. According to the theoretical diffusion model of NIR_SRS with a single layer [29], µ_a can be expressed as a quadratic function of S as follows, assuming that ρµ_eff≫1 and µ_a+µ′_s≈µ′_s :

where I_α and I_β are the intensity of light detected by the photodiodes located at distances α and β from the LED, respectively, ρ is the distance between α and β, µ_eff is the attenuation coefficient and µ′_s is the reduced scattering coefficient of muscle. From (1) and (2), the following quadratic expression can be obtained:

where a, b and c are constants. This relationship was also able to be applied to the results of Monte Carlo simulation using the three-layered model. The constants a, b and c vary with ρ, fat layer thickness and µ′_s of muscle and they were obtained by performing curve-fitting to the S-µ_a curves. For instance, for a fat layer thickness of 3 mm, ρ=12 mm and µ′_s=0.5 mm^-1, a=12.6, b=-2.59 and c=0.147.

In this study, the intensity of backscattered light measured by the photodiodes located at 20 and 32 mm from the LED was used for calculation of muscle oxygenation. The reduced scattering coefficient µ′_s of muscle was assumed to be constant (µ′_s,830=0.5 mm^-1, µ′_{s, 770}=0.6 mm^-1) during measurements and its value at rest was determined on the basis of our results of an in vivo experiment using an instrument based on NIR_TRS [30]. The algorithm for calculating oxygenated and deoxygenated hemoglobin concentrations was based on conventional two-wavelength spectroscopy. The concentration of hemoglobin also includes that of myoglobin since the absorption spectra of these two chromophores are similar [31]. The absolute concentrations of oxygenated hemoglobin ([HbO₂+MbO₂]), deoxygenated hemoglobin ([Hb+Mb]) and total hemoglobin [total (Hb+Mb)] can be calculated by the following equations:

where εHbO2λ and εHbε are the molar extinction coefficients of HbO2 and Hb at wavelength of λ, respectively (values of the coefficients obtained from Appendix I of Matcher et al. [32]). The subscripts 830 and 770 indicate the wavelengths of the LED. Tissue oxygen saturation (StO2), an important physiological parameter for evaluating the balance of O2 supply and O2 consumption in muscle, was defined as

(8)

2.3 Evaluation of measurement accuracy using a tissue-equivalent phantom

A tissue-equivalent phantom was constructed to validate the measurement accuracy of StO₂ measured using the imaging instrument based on NIR_SRS. A 20×20×20 cm³ acrylic water tank (wall thickness =3 mm, refractive index =1.49) was used. The water tank was filled with a solution of 130 ml 0.5% Intralipid (Intralipid 20%, Terumo Corp., Japan) suspended in 5200 ml phosphate-buffered saline (PBS) (µ′_s=0.5 mm^-1) [33], into which washed red blood cells and NaHCO₃ were added to produce a phantom with a final hemoglobin concentration of 0.05 mM (about 1% hematocrit) and pH of about 7.4. The water tank was placed in a temperature-controlled reservoir to maintain the phantom temperature at 37°C. The O₂ saturation level of the phantom was sequentially varied by bubbling approximately 100% N₂ or 100% O₂, with a small amount of CO₂ added to maintain the partial pressure of CO₂ at about 40 mmHg. A small amount of yeast was added during the deoxygenation process.

An optical probe of the imaging instrument was placed on the surface of the phantom for continuous measurements of S_tO₂ of the phantom. The probe was slightly submerged in the phantom to ensure good contact between the probe and the solution. The O₂ saturation (O₂SAT) value measured using a blood gas analyzer (BGA) (348 pH/BGA, Bayer Healthcare, USA) was used as a reference. The solution of the phantom was periodically sampled for measurement of O₂SAT using the BGA. The relationship between S and µ_a to calculate S_tO₂ of the phantom was obtained by Monte Carlo simulation described in section 2.2, assuming a semi-infinite medium without skin and fat layers. When ρ=12 mm and µ′_s=0.5 mm-1, a=7.72, b=-0.987, and c=0.019. The experiment was performed in a darkroom.

2.4 Calibration, instrument stability and absorption linearity

As described in section 2.1, the sensitivity of the photodiodes of an optical probe varies due to differences in their effective active area and gain of amplifiers. Calibration of the sensitivity of the photodiodes was performed using a stable light source during construction of the probes. Using the signal level of detected light of the photodiode placed at 20 mm from the LED as a reference, calibration coefficients were determined for the photodiodes placed at 25 and 32 mm from the LED. The average disparity in measurement sensitivity of the corresponding photodiodes between different probes was less than 3%. Therefore, the coefficients for calibrating the measurement sensitivity between different probes were not used in actual measurement. In addition, a solid phantom was also constructed for periodic maintenance of the instrument and for examination of instrument stability, as described below.

For the examination of instrument stability, an optical probe of the imaging instrument was firmly placed on the surface of a solid phantom. Measurement was performed for 3 hours, starting from the time the instrument was turned on. The phantom was constructed with reference to the procedures reported by Firbank et al. [34]. Near-infrared absorbing dye (Projet 900NP, FUJIFILM Imaging Colorants Ltd., Japan) and white pigment (Pica-ace, Kurachi Ltd., Japan) were added as an absorber and a scatterer, respectively, to a two-part epoxy resin (Craft Resin Z-1 resin with hardener, Nissin Resin Co. Ltd., Japan) to form a solid phantom (diameter =12.5 cm, height =7.0 cm, weight =900 g). The concentration of the absorber was 5.0×10^-4 wt% and that of the scatterer was 0.1 wt%. Reduced scattering coefficients µ′_s of the phantom (µ′_s,830=0.755 mm^-1, µ′_{s, 770}=0.851 mm^-1) were determined using NIR_TRS (C4334, Hamamatsu Photonics, Japan). These values were applied to the calculation of µ_a of the phantom using the algorithm described in section 2.2 (µ_{a, 830}=0.035 mm^-1, µ_{a, 770}=0.022 mm^-1).

Absorption linearity was also examined with a liquid phantom using the acrylic water tank described in section 2.3. The acrylic water tank was filled with a solution of 130 ml 0.5% Intralipid (Intralipid 20%, Terumo Corp., Japan) suspended in 5200 ml distilled water (µ′_s=0.5 mm^-1) [33], into which diluted Indian ink (100 times) (INK-30-DR, PILOT Corp., Japan) was added in steps of 2.5 or 5 ml. The concentration of the ink solution was varied in the range of 4.7×10^-4 to 9.3×10^-3 %. The experiment was repeated four times. Both experiments were performed in a darkroom and the room temperature was maintained at 25°C throughout the experiments.

2.5 Imaging of intra-muscular S_tO₂

A healthy male subject (age: 23 years, height: 168 cm, weight: 60 kg) participated in this measurement. Informed consent was obtained from the subject prior to measurement. The subject performed 30-s continuous isometric knee extension exercises at 20%, 40% and 70% maximum voluntary contraction (MVC). Before the start of each exercise, the subject remained in a stable sitting position on a customized chair with both knees flexed at 90° of flexion. A leg brace was firmly strapped to the ankle of the right leg and the connector of the brace was connected to a strain gauge to measure the extension force at the ankle. The strain gauge was connected to a strain amplifier (N5901, NEC, Japan) and the output signal from the amplifier was displayed on an oscilloscope that was placed in front of the subject to provide visual feedback so that the subject could monitor and maintain the level of extension force during measurements.

The spatial distribution of S_tO₂ in the rectus femoris (RF) of the right leg was measured. As shown in Fig. 3(a), sixteen optical probes were placed on the skin surface over the RF. The probes were first affixed on the RF with surgical tape (Transpore, 3M HealthCare, Japan). A long and flexible black bandage (Aquwrap, Aida Corp., Japan) was then wrapped around the thigh and over the probes to prevent displacement of the probes during exercise. Care was taken to ensure that no excessive pressure was applied to prevent blockage of blood flow to the muscle. Measurements began with the 20% MVC exercise and were followed by measurements with the 40% MVC and 70% MVC exercises. A 30-minute recovery interval was allocated between each measurement. During the rest interval, the probes and bandage were not removed in order to maintain the consistency of probe location and bandage pressure throughout measurements.

2.6 Multi-point measurement of inter-muscular S_tO₂

The same subject participated in this measurement. As shown in Fig. 3(b), four probes were placed on the skin surfaces over the RF, the vastus lateralis (VL) and the vastus medialis (VM) of the right leg to measure the S_tO₂ in these muscles. The subject performed (i) continuous isometric knee extension (KE) and (ii) continuous isometric knee extension with leg press action (KELP). During KE, both knees were flexed at 90° (the angle at which the knee is fully extended being taken as 0°). During KELP, the knee angle of the right leg was maintained at about 60° of flexion [35] and the knee of the left leg was flexed at 90°. Both exercises were performed for 30 s at 70% MVC. The recovery time allocated between the two exercises was 30 min. For the measurements described in sections 2.5 and 2.6, the thickness of the overlying fat tissue for the measurement site was measured and its effect was corrected by the algorithm described in section 2.2. The thickness was measured using a B-mode ultrasound diagnostic apparatus with an 8-MHz ultrasound probe (SSA-320A, Toshiba Corp., Japan).

 

Fig. 3. Arrangement of probes on measurement sites. (a) Arrangement of probes on the skin surface over the rectus femoris (RF) of the right leg and (b) on the skin surfaces over the RF, the vastus lateralis (VL) and the vastus medialis (VM) of the right leg. Intervals between center of measurement of the VL and center of the patella were about 10 cm and 15 cm in circumferential and longitudinal directions, respectively, whereas those between the VM and center of the patella were about 5 cm and 10 cm, respectively. The probes were placed on the belly of each muscle. Distances between the probes were about 2 cm and 5 cm in lateral and longitudinal directions, respectively.

Download Full Size | PDF

3. Results

3.1 Evaluation of measurement accuracy using a tissue-equivalent phantom

As shown in Fig. 4, a significant correlation (intraclass correlation coefficient =0.99, p<0.05) was found between the O₂ saturation values measured using the imaging instrument based on NIRSRS and those measured using the BGA. Statistical analysis was performed using SPSS 14.0J for Windows (SPSS Inc., USA). The total hemoglobin concentration measured using the imaging instrument was (mean ± SD) 0.047±0.002 mM, and this value agrees well with the concentration of the phantom.

3.2 Instrument stability and absorption linearity

The change in µ_a over a 3-hour measurement period is shown in Fig. 5(a). The values were plotted after normalization to the average values obtained in the last 15 min of the measurement period. The coefficient of variation (SD/mean) in the measured µ_a values is less than 0.6% for both wavelengths. Typical changes in µ_a as a function of ink concentration are shown in Fig. 5(b). At both wavelengths, µ_a is found to increase linearly as a function of ink concentration. The average coefficient of variation is less than 0.8% for both wavelengths.

 

Fig. 4. Simultaneous measurement of O₂ saturation by the imaging instrument based on NIR_SRS and by a blood gas analyzer (BGA), S_tO₂: tissue oxygen saturation; O₂SAT: O₂ saturation. The line is least-squares fit to data.

Download Full Size | PDF

 

Fig. 5. (a) Test of instrument stability on a stable solid phantom and (b) measurement of absorption linearity as a function of ink concentration for a representative probe. (a) The variation in absorption coefficient µ_a as a function of time was normalized to the average values obtained in the last 15 min of the measurement period. (b) Points represent the average values and the error bars represent the SD (n=4). The solid line (830 nm, r²=0.99) and broken line (770 nm, r²=0.99) are linear least-squares fits to data.

Download Full Size | PDF

3.3 Imaging of intra-muscular S_tO₂

Typical temporal changes in S_tO₂ are shown in Figs. 6(a) and 6(b). The data shown in Fig. 6(a) were obtained in the proximal region of the RF (probe 3, as indicated in Fig. 3(a)) and those shown in Fig. 6(b) were obtained in the distal region (probe 14). The values measured in the proximal and distal regions were not significantly different. S_tO₂ values decreased from 70% at rest to values ranging from 20% to 40% during exercise, depending on the intensity of exercise.

Images showing spatial distribution of S_tO₂ are presented in Fig. 6(c). The images were generated by applying linear interpolation to the measurement points. At rest (0 s), S_tO₂ values (spatial means ± SD) of 16 probes in the RF during the 20%, 40% and 70% MVC exercises, 65±3%, 65±3% and 67±3%, respectively, were not significantly different. However, the RF region was clearly imaged during exercise as decreases in S_tO₂. At the end of exercise (30 s), S_tO₂ values were 43±17%, 30±19% and 22±15% for the 20%, 40% and 70% MVC exercises, respectively. After exercise, slower recovery of S_tO₂ to the rest level was observed for the 40% MVC and 70% MVC exercises, because of larger O₂ debt, than for 20% MVC. At 60 s (30 s after the end of exercise), S_tO₂ values in the RF for the 20%, 40% and 70% MVC exercises were 70±3%, 56±11% and 52±11%, respectively.

The arrangement and geometry of the probes allowed simultaneous measurement of 16 sites over a skin area of about 100 cm². Spatial resolution, determined by the distance between adjacent probes, was about 2 cm and 5 cm in lateral and longitudinal directions, respectively.

 

Fig. 6. Examples of temporal and spatial changes in tissue oxygen saturation (S_tO₂) in the rectus femoris (RF) during isometric knee extension exercise (indicated by Exerc) for 30 s at 20%, 40% and 70% maximum voluntary contraction (MVC). The data shown in (a) were obtained from the proximal region of the RF (probe 3, as indicated in Fig. 3(a)) and those shown in (b) were obtained from the distal region (probe 14). The black arrowheads indicate the times that the S_tO₂ images shown in (c) were generated. (c) Selected imaging results of the spatial distribution of S_tO₂ in the RF during exercise at 20% (left panels), 40% (middle panels) and 70% MVC (right panels). From top to bottom, images obtained at 0 s (start of exercise) and at 10 s, 30 s (end of exercise) and 60 s after the start of exercise are shown. The pseudo-colored bar shows S_tO₂ values from 0% to 80% ((Media 1), 0.77 MB: reproduced at 5 times faster than its original speed) ((Media 2)large version: 3.90 MB).

Download Full Size | PDF

3.4 Multi-point measurement of inter-muscular S_tO₂

Typical temporal changes in S_tO₂ during KE and KELP are shown in Figs. 7(a) and 7(b), respectively. These data were obtained in the lateral side of the proximal region of the VL, RF and VM (probes 1, 5 and 9, as indicated in Fig. 3(b)). In the case of KE (Fig. 7(a)), S_tO₂ value in the RF rapidly decreased from 70% at rest to about 10% and reached a plateau about 10 s after the start of exercise, whereas S_tO₂ values in the VL and VM decreased at a lower rate and were significantly higher than that in the RF at the end of exercise. In contrast, in the case of KELP (Fig. 7(b)), S_tO₂ values in all muscles decreased to around 20% and reached a plateau 10–15 s after the start of exercise.

Images showing spatial distribution of S_tO₂ are shown in Fig. 7(c). The arrangement and geometry of the probes allowed simultaneous measurement of four sites within each muscle. Differences in S_tO₂ values among the muscles, though not clearly shown in results obtained at the onset of exercise (0 s) (top panels), were clearly visible in the results obtained during exercise at 10 s (middle panels) and 30 s (bottom panels). Notable decreases in S_tO₂ values were detected in the RF during both exercises. As shown in the left panels of Fig. 7(c) (KE), S_tO₂ values in the VL and VM decreased at a slower rate and were higher than S_tO₂ in the RF during exercise (10 s). At the end of exercise (30 s), S_tO₂ values (spatial means ± SD) of four probes in the VL, RF and VM were 50±11%, 27±10% and 40±8%, respectively. In comparison, as shown in the right panels of Fig. 7(c) (KELP), S_tO₂ values in the VL, RF and VM were 34±8%, 28±9% and 27±6%, respectively, at the end of exercise.

 

Fig. 7. Examples of inter-muscular temporal and spatial changes in tissue oxygen saturation (S_tO₂) in the vastus lateralis (VL), rectus femoris (RF) and vastus medialis (VM). The data shown in (a) were obtained during continuous isometric knee extension (KE) for 30 s (indicated by Exerc) at 70% maximum voluntary contraction (MVC). The data shown in (b) were obtained during KE with leg press action (KELP) for 30 s (indicated by Exerc) at 70% MVC. The data shown in (a) and (b) were obtained from the lateral side of the proximal region of the muscles (probes 1, 5 and 9, as indicated in Fig. 3(b)). The black arrowheads indicate the times that the selected S_tO₂ images shown in (c) were generated. (c) Selected imaging results of the spatial distribution of S_tO₂ in the VL, RF and VM during KE (left panels) and during KELP (right panels). From top to bottom, images obtained at 0 s (start of exercise) and at 10 s, 30 s (end of exercise) after the start of exercise are shown. The pseudo-colored bar shows S_tO₂ values from 0% to 80%.

Download Full Size | PDF

4. Discussion

Unlike NIR_CWS, NIR_SRS enables measurement of the absolute values of hemoglobin concentration and S_tO₂ under the assumption that variation in µ′_s is much smaller than that of µ_a during measurements. The average rest values of S_tO₂ obtained in the present study (64–75%) (Figs. 6 and 7) agree well with those measured using other NIRS instruments that enable simultaneous measurements of absolute values of µ′_s and µ_a, such as instruments based on (i) NIR_TRS: 71±2%, 74±1% [13] and 60% [16] (estimated from Fig. 1 of Hamaoka et al. [16]) and (ii) NIR_PMS: 65±5% [14]. Our rest S_tO₂ values also agree reasonably well with those measured using an instrument based on broadband NIR_CWS: 66±4% [36].

Unlike measurement of brain oxygenation, measurement of muscle oxygenation will yield both the concentrations of hemoglobin and myoglobin because muscle tissues contain myoglobin and the absorption spectrum of myoglobin is similar to that of hemoglobin [31]. Seiyama et al. and Mancini et al. showed that most of the NIRS signal comes from hemoglobin [37, 38]. However, Tran et al. demonstrated by using ¹H-nuclear magnetic resonance (NMR) and NIRS that the NIRS signals were strongly correlated with deoxygenation of myoglobin [39]. These contradictory results indicate the need for additional studies to elucidate the relative contributions of hemoglobin and myoglobin to the NIRS signal. Although we cannot estimate the relative contributions of hemoglobin and myoglobin to the NIRS signal from our S_tO₂ values, it should be stressed that quantitative measurement of muscle oxygenation using NIRS is assured because the molar absorption coefficient ε of myoglobin is one-fourth of that of hemoglobin [31] since ε is proportional to the number of heme groups. Therefore, the total concentration of heme groups as the O₂ carriers, including myoglobin, can be quantified as an equivalent molar concentration of hemoglobin. The contribution of other O₂-dependent absorbing chromophores such as mitochondrial cytochrome c oxidase to the NIRS light attenuation signal is small (about 2–5%) [40] and its influence is considered negligible in this study. Also, the absorption by skin melanin and fat is less than 5% of the signal if adequate source-detector separation (>20 mm) is used [40, 41].

One benefit of our algorithm to calculate S_tO₂, described in section 2.2, is that a massive amount of numerical calculation is not required and it is thus appropriate for real-time measurement. In addition, the influence of overlying tissue such as a fat layer can be corrected, though measurement of the thickness of fat layer is needed. Using this algorithm, the construction of an imaging instrument based on NIR_SRS that utilizes simple and inexpensive electronic circuits was possible. The disadvantage of our method is that knowledge of µ′_s and fat layer thickness is required beforehand. Another approach would be to apply the multi-distance method based on NIRPMS [11, 42]. The major benefits of this method are that it enables estimation of both µ_a and µ′_s and that measurement of the thickness of overlying tissue is not needed. The disadvantage of this multi-distance method is that it is difficult to confirm whether or not the influence of the overlying tissue has been eliminated for each individual [42].

To the best of our knowledge, this is the first study in which spatial distribution of S_tO₂ within a muscle during exercise was imaged using an imaging instrument based on NIR_SRS. As shown in Fig. 6(c), exercise intensity-dependent (20%, 40% and 70% MVC) dynamic changes in S_tO₂ in the RF region during exercise and recovery were clearly imaged. In addition, as shown in Fig. 7(c), changes in S_tO₂ among the VL, RF and VM muscles in response to variation in the exercise protocol (KE and KELP) were observed. At the end of KE, the level of S_tO₂ in the RF was lower than that in the VL and VM. On the other hand, at the end of KELP, the lowest levels of S_tO₂ in the three muscles were almost the same. The VL, RF and VM demonstrated different O₂ kinetics in response to variation in exercise protocol and this could be attributed to the differences in relative contribution of the muscles to force generation in exercise [43, 44]. Since functional imaging or multi-point measurement of S_tO₂ within a muscle or between different exercising muscles provides important physiological information, our instrument has potential applications in the fields of (i) sports medicine for study of cooperative contraction of muscles during whole body exercise, such as evaluation of the changes in O₂ supply and O₂ consumption capability of several muscles during and after a training program and (ii) pathophysiology for detection of regional differences in S_tO₂, such as measurement of lower extremity muscles of subjects with peripheral vascular disease [12]. The lowest level of S_tO₂ in this study is consistent with those measured using NIR_TRS (12-min arterial occlusion) [16] and NIR_SRS (5-min arterial occlusion) [45]. The lower limit of about 20% saturation suggest that O₂ saturation of myoglobin is relatively high even in a state of hypoxia, whereas hemoglobin is almost fully deoxygenated, as inferred from the dissociation curves of hemoglobin and myoglobin. In addition, a relatively small amount of arterial blood of about 100% saturation might raise the lower limit. The mechanism of the existence of this lower limit is unclear [16] and further studies are needed to clarify it.

The strength of our instrument is that a high temporal resolution of measurement was achieved and this is one of the most distinct advantages of NIR_SRS compared to NIR_TRS and NIR_PMS. Despite employing the time-multiplexing technique, a sampling period of 50–250 ms was achieved. This temporal resolution was sufficiently fast for measurement of the dynamic changes in S_tO₂ during exercise (Figs. 6 and 7). A fast temporal resolution is essential for study of muscle function because it could also be applied to measure local O₂ consumption that increases significantly (about 10 times more than the rest value) and rapidly (within a few seconds) during intermittent isometric exercises at high intensity (greater than 50% MVC), as shown by Hamaoka et al. [46].

The limitations of our instrument are as follows (i) insufficient spatial resolution and (ii) the assumption that µ′_s of muscle is constant during measurement. Although there was an inevitable limitation of spatial resolution (2–5 cm), it can be further improved by decreasing the distance between the light source and detectors. However, the measurement sensitivity will be deteriorated because more light will propagate through the shallow regions (skin and fat layer) instead of the region of interest (muscle), which is situated at a deeper region. The practical limit of the spatial resolution was about 1.5 cm, taking the following factors into consideration: (i) measurement depth and (ii) mean fat layer thickness (about 5 mm: based on our measurements of fat layer thicknesses of the VL, RF and VM in 16 healthy male subjects in a separate study). Therefore, measurement sensitivity rather than spatial resolution was given priority in the design of our probes. One approach to increase the spatial resolution would be to design a probe that consists of a mesh of sources and detectors instead of discrete single sensors as used in this study. This design would enable one light source to be measured by several detectors simultaneously, thus resulting in higher resolution. However, a new design for a flexible mesh-type probe that can accommodate the deformation of muscle during exercise is needed.

On the other hand, although we assumed a constant µ′_s during measurement, the difference in S_tO₂ estimation is relatively small even if our assumed µ′_s values were varied by about twice the original values. For example, S_tO₂ values at rest (0 s) in the RF during the 70% MVC exercise (Fig. 6(c)) were (i) 67±3%, assuming that µ′_s,830=0.5 mm^-1 and µ′_s,770=0.6 mm^-1 (original assumption) and (ii) 61±4%, assuming that µ′_s,830=0.9 mm^-1 and µ′_s,770=1.0 mm^-1. Our assumption is supported by the results of Torricelli et al. [13] and Hamaoka et al. [16] obtained using NIR_TRS that showed ±10% changes in rest µ′_s and exercise µ′_s values of the calf during dynamic plantar flexion exercise [13] and of the forearm during arterial occlusion [16] (estimated from Fig. 8 of [13] and Fig. 1 of [16]). The disparity in S_tO₂ due to errors (such as inter-subject variability) in the assumption of optical properties of adipose tissue is a few percent [26], but the influence of skin is negligible because a long source-detector separation (20 and 32 mm) was used in this study [40, 41].

In this study, LEDs were used as light sources, but the influence of the broad wavelength width (about 30 nm) of LEDs on the accuracy of measurements was not sufficiently investigated in this study. However, the results of simultaneous measurements of O₂ saturation by our instrument and by a blood gas analyzer (Fig. 4) and an absorption linearity test as function of ink concentration (Fig. 5(b)) indirectly show the validity of the use of LEDs. Fantini et al. investigated the effects of wavelength width of LEDs on single-wavelength equations using simulation and showed that the error in estimation of µa and µ′s is only a few percent [11]. The use of monochromatic light sources such as laser diodes (LDs) (spectral bandwidth ~5 nm) is an alternative approach if fiber optic guides from LDs to optical probes could be handled easily in measurements of muscle during exercise.

5. Conclusions

An imaging instrument based on NIR_SRS (spatially resolved spectroscopy) was developed for temporal and spatial analyses of muscle oxygenation, and the instrument can be connected to 32 compact and separate-type optical probes to provide flexibility in their arrangement on different measurement sites. Its measurement accuracy was evaluated using a tissue-equivalent phantom. In addition, imaging of tissue oxygen saturation (S_tO₂) in the RF (rectus femoris) and multi-point measurement of S_tO₂ in the RF, VL (vastus lateralis) and VM (vastus medialis) were performed, and the changes in S_tO₂ in response to increase in exercise intensity and to variation in exercise protocol were clearly imaged. The results show that the instrument is potentially useful for applications in areas of sports medicine such as monitoring of the improvement in muscle function among several muscles during a training program.