2.1.

Instrumentation

For measuring the distribution of times of arrival (DTA) of fluorescence photons, a time-resolved multichannel spectral system based on time-correlated single photon counting (TCSPC) electronics³⁰ was constructed. The diagram of the system is presented in Fig. 1. Fluorescence of the ICG was excited by femtosecond (fs) pulses generated with Ti:Sa laser MaiTai (Spectra Physics, USA) operating at a wavelength of 760 nm with a repetition frequency of 80 MHz.

Fig. 1

The setup for multiwavelength time-resolved fluorescence measurements.

The excitation light was delivered to the tissue with the use of an optical fiber (length 1 m, $\mathrm{NA}=0.28$, diameter 1 mm). The power of light at the source fiber tip was about 60 mW. The optode was constructed in such a way that the power density of the light on the surface of the skin was not higher than $2\mathrm{mW}/{\mathrm{mm}}^2$. The remitted photons were transmitted to the detection system with the use of a fiber bundle (length 1 m, $\mathrm{NA}=0.22$, diameter 3 mm) terminated with a circular tip on the tissue side and line-shaped ($1\times 7\mathrm{mm}$) fitting on the slit of polychromator. For blocking of the excitation light, an optical module was implemented between the tissue and the tip of the detecting fiber bundle. A high-pass filter (RLP793, Thin Films Inc.) was positioned between two lenses focusing the light remitted from the tissue to the fiber bundle. The source fiber and optical detection module were positioned at source-detector separation of $R=3\mathrm{cm}$ and fixed to the surface of the tissue (of physical phantom) with a homemade optode holder using flexible black rubber and Velcro strips.

The fluorescence photons were acquired using an integrated polychromator PML Spec equipped with 16-channel photomultiplier tube detector (Becker & Hickl, Germany). This detection system allowed for simultaneous acquisition of times of arrival of fluorescence photons on 16 spectral channels. The width of spectral channel was approximately 12 nm. The spectral range of 208 nm was adjusted in such a way that the measurement covered wavelengths between 772 and 960 nm. Single photon pulses were counted using a router and a time-correlated single photon counting card SCP830 (Becker & Hickl, Germany). Power supply to the photomultiplier tubes (PMTs) was provided from a DCC-100 card (Becker & Hickl, Germany) which avoided eventual over-illumination of the detectors. Measured distributions of times of arrival of fluorescence photons contained 1024 data points with time channels width of $\mathrm{\Delta }t=9.77\mathrm{ps}$.

Synchronization of the laser pulses with the TCSPC electronics was provided with the use of a reference detecting photodiode (PHD 400). Instrumental response function (IRF) for our experimental setup was measured by positioning the source fiber in front of the detecting bundle with a piece of paper placed in front of the bundle in order to fill out its numerical aperture.³¹ The PMTs were illuminated with the laser light at wavelength used for ICG excitation (760 nm). The IRF was evaluated sequentially in the individual emission channels by adjustment of the grating. In such a way the IRFs of all the detecting channels for excitation wavelength was obtained (see Fig. 2 in Ref. 29). The full width half maximum (FWHM) of the IRF was in range of 130 to 150 ps for all spectral channels. Collection time for acquisition of 16 DTAs of fluorescence photons was 300 ms.

Fig. 2

Geometry of the liquid phantom. Boli with higher concentration of indocyanine green (ICG) were passing through the tube positioned in the fish tank at two depths in respect to the front wall of the phantom. Source fiber (red square) and optical detection module (blue circle) were fixed on the surface of Mylar film which formed the front wall of the phantom.

The PML-SPEC multiwavelength detection assembly is a combination of the 16 channel detector head (PML-16C-1) and a small grating polychromator. The PML-16C-1 with the multialkali cathode has a different spectral sensitivity depending on the wavelength.³² Typical spectral sensitivity of the bialkali and the multialkali photocathode is presented in the Ref. 31. However, we obtained the sensitivity curve of our PMT in NIR spectral range by measuring the number of detected counts in corresponding channels (wavelength). We set the wavelength of the laser on the corresponding detection channel and measured the IRF, keeping the fixed power and other parameters of the experiment. Then, the sensitivity coefficient was estimated for all of the measured channels, and the measured optical signals were calibrated according to these sensitivity factors.

According to Ref. 33, for a multichannel PMT the constant fraction discriminator (CFD) threshold has a noticeable influence on the uniformity of the channels in terms of efficiency. Although all PMT channels use the same dynode system and the same operating voltage, the pulse amplitude distribution for the individual channels may differ noticeably. Therefore, in our measurements the CFD threshold was low enough for our detector assembly to obtain reasonable channel uniformity.

We also did measurements and determined the sensitivity curve which reflects the influence of the grating on the signal measured. We scanned the grating along the fixed wavelength (760 nm) from the laser source. Thus, the number of the detection channels changed with respect to the excitation wavelength. As a result, the difference between detected counts was approximately 8% for channels 2 to 15 (the first and last detection channels were not taking into operation during all time-resolved measurements). It should be noted, that the count discrepancy for the detection channels operating with spectral range of emission properties of ICG, was 2.5%. On the basis of the above we suggest that the imperfection of our grating has little influence on the differences between the detector channels. Our results for channel uniformity is in good agreement with the manufacturer’s report.³³

Dependence of the transit time on the location on the photocathode shows a systematic wobble depending on the channels. The effect is taken into account in the data analysis by determining individual IRFs for the channels.

The continuous wave measurements of absorption and emission spectra of ICG were performed using a USB2000 spectrometer (Ocean Optics Inc.) with grating set with a spectral range from 367 to 1108 nm and optical resolution of 0.35 nm (FWHM). Measured spectra contained 2048 data points. The spectrometer was powered and controlled by USB cable connected with the computer. In these measurements, fluorescence of the ICG was excited by femtosecond pulses generated with a Ti:Sa laser source Mai Tai (Spectra Physics) adjusted at 760 nm wavelength. Fluorescence spectrum measurement was performed using interference filter RLP793 (Thin FilmImaging Technologies, Inc.) used to block the excitation wavelength and pass the emission light in the wavelength range from 800 to 860 nm. To obtain absorption spectrum, the supercontinuum laser light source was used (Fianium, UK) emitting radiation in the range from 650 to 850 nm. The optical fibers were fixed on the forehead of a subject with help of a flexible rubber holder with a 1-cm source-detector separation. Measurements started 1 min. before injection to store the reference spectrum, and continued for averaging of the ICG spectrum over a 5-min. period after injection of 5 mg of the dye dissolved in 3 mL of aqua pro injectione.

2.2.

Phantom Experiments

The liquid phantom used consisted of a fish tank (dimensions: $20\times 20\times 15\mathrm{cm}$) and a silicon, optically turbid tube (inner diameter of 3 mm, outer diameter of 3.8 mm). The tube was positioned inside the phantom in such a way that two of its segments were located at different depths in the liquid as shown in Fig. 2. The front wall of the phantom was made of a thin (50 μm) optically turbid Mylar film (DuPont Teijin Films) through which the measurements were recorded. The fish tank was filled with a mixture of milk and water ($1:3$) with small amount of black ink-water solution ($1:100$, ca. $1\mathrm{mL}/\mathrm{L}$) added to obtain optical properties of the liquid (${\mu }_a=0.18{\mathrm{cm}}^{-1}$, ${\mu }_s^{\prime }=19{\mathrm{cm}}^{-1}$) close to those observed in living tissues.

Two series of experiments were performed. First, the measurement was carried out in a homogeneous medium to determine the relationship between concentration of the dye and the measured number of fluorescence photons. In this experiment, concentration of the ICG in the optically turbid mixture was changed gradually in 100 steps from 0 to $0.5\mathrm{mg}/\mathrm{L}$. Next, time-resolved measurements of fluorescence were carried out in the phantom with dynamic inflow of ICG trough the silicon tube positioned inside the medium. The outflow part of tube was fixed next to the Mylar film, whereas the inflow part of the tube was located deeper at the distance of $r=2\mathrm{cm}$ from the Mylar film (see Fig. 2). Locations of the segments of the tube allow simulation of the inflow of dye to superficial layers of the human head and deeper layers corresponding to the brain tissue. A small amount of ICG ($0.005\mathrm{mg}/\mathrm{L}$) was added to the milk-water-ink mixture in the fish tank which created a fluorescent background which is typically used for in vivo measurements. The same mixture was pumped through the tube at a rate of $0.6\mathrm{mL}/\mathrm{s}$ using a peristaltic pump. A bolus of 2 mL of milk-water-ink mixture with a higher concentration of ICG was rapidly ($<1\mathrm{s}$) injected into the tube in such a way that the bolus appeared first in the segment of the tube located deeper and then in the superficially positioned part of the tube. The dye was injected into the tube at a distance of about 15 cm from the phantom.

2.3.

In Vivo Measurements

The measurements were carried out on the heads of two healthy adults during intravenous injection of ICG. The volunteers were examined in supine position. A bolus of 5 mL of liquid containing 5 mg of ICG dissolved in aqua pro injectione was administered into forearm vein. The ICG injection was rapid (shorter than 1 s) and followed by a 10-mL injection of normal saline.

3.1.

Absorption and Emission Spectra of ICG In Vivo

The absorption and emission spectra of ICG measured in vivo with the use of a continuous wave NIR spectrometer on the surface of the human head during the injection of the dye are presented in Fig. 3. The normalized amplitude of absorption (light green line) and fluorescence light (dark green line) of ICG are shown. The measured spectra were normalized by their maximum after subtraction of the background, which allows for comparison of data obtained at different wavelengths. The dark green symbols represent data obtained from wavelength- and time-resolved measurement: only several points corresponding to selected wavelengths were obtained because of the limited spectral resolution of time-resolved device (dotted line is provided for guiding eyes only). The overlapping between the absorption and emission spectra can be seen in the range from 790 to 825 nm. It can be clearly noted that the reabsorption effect is very strong at shorter wavelengths (790 to 825 nm) and can be effectively avoided at wavelengths longer than 825 nm.

Fig. 3

Normalized absorption (light green line) and fluorescence (dark green line) spectra of indocyanine green (ICG) measured on the surface of the human head during injection of the dye. The measured spectra were normalized by their maximum after subtraction of the background. The dark green symbols represent data obtained from wavelength- and time-resolved measurement, dotted line was provided for guiding eyes only.

3.2.

Phantom Studies

Dependence of the fluorescence signal on concentration of the ICG dye was studied on the homogeneous liquid phantom for multiple fluorescence emission wavelengths. Changes in number of fluorescence photons $N_{\mathrm{tot}}$ and mean time of arrival of fluorescence photons $\langle t\rangle$ were obtained by analysis of DTAs successfully recorded for 6 emission wavelengths in range from 798 to 861 nm. Results are presented in Fig. 4. The ICG concentration at which the total number of fluorescence photons $N_{\mathrm{tot}}$ reaches the maximum strongly depends on the emission wavelength. For shorter emission wavelengths (798–822 nm), $N_{\mathrm{tot}}$ reached the maximum value at ICG concentration of about $0.1\mathrm{mg}/\mathrm{L}$. At these wavelengths, for higher ICG concentrations $N_{\mathrm{tot}}$ rapidly decreases and reaches about 30% of its maximal value at $C_{\mathrm{ICG}}=0.47\mathrm{mg}/\mathrm{L}$. For emission wavelengths of 835, 848, and 861 nm, the maximum value of $N_{\mathrm{tot}}$ was observed when $C_{\mathrm{ICG}}$ was about 0.17, 0.2, and $0.3\mathrm{mg}/\mathrm{L}$, respectively, and the decrease of $N_{\mathrm{tot}}$ after the maximum was smaller (less than 40% of the maximum).

Fig. 4

Normalized number of fluorescence photons $N_{\mathrm{tot}}$ and mean time of arrival of fluorescence photons $\langle t\rangle$ as a function of indocyanine green (ICG) concentration obtained for different emission wavelengths.

In terms of mean time of arrival of fluorescence photons, $\langle t\rangle$, differences in the pattern of changes depend strongly on the emission wavelength. It can be noted that $\langle t\rangle$ is longer for longer wavelengths. This effect reflects the fact that fluorescent light detected at longer wavelengths originates from the deep layer of the studied structure, and the fluorescence photons have a longer way to reach the detector. Moreover, for all spectral channels, $\langle t\rangle$ tends to decrease with increasing ICG concentration. This effect can be explained by the fact that at a higher concentration of dye, the excitation light could not penetrate the deep layers of the phantom. However, at very low ICG concentrations (up to $0.025\mathrm{mg}/\mathrm{L}$), $\langle t\rangle$ is larger at the shortest wavelengths.

Time-resolved measurements of fluorescence were also carried out on a phantom with dynamic inflow of ICG in a tube located at different depths. A large amplitude of changes in the number of photons, $N_{\mathrm{tot}}$, and mean times of arrival of fluorescence photons $\langle t\rangle$ calculated from DTAs, which correspond to the inflow into both superficially and deeper located segments of the tube, are obtained for nonhomogeneous phantom measurements at different emission wavelengths (Fig. 5). Two peaks at $T\approx 27\mathrm{s}$ and $T\approx 40\mathrm{s}$ in the $N_{\mathrm{tot}}$ and $\mathrm{\Delta }\langle t\rangle$ are related to the inflow of ICG into the part of tube located deeper and superficially, respectively. Amplitudes of these peaks depend on the emission wavelength, which is more distinctly visible in the number of photons, $N_{\mathrm{tot}}$. In terms of mean time of arrival of fluorescence photons $\langle t\rangle$, the flow of dye through the tube leads to increased amplitudes of change in this statistical moment with increased emission wavelength, both in a positive peak caused by ICG inflow into the deeper part of the tube, as well as a negative peak related to ICG inflow to the superficial tube segment. For inflow of the dye into the deeper segment, the increase in $\langle t\rangle$ is more pronounced than the change in $N_{\mathrm{tot}}$. Moreover, the fluorescence signal detected at longer wavelengths is a bit more sensitive to the inflow of dye into deeper compartments of the medium since the amplitude of increase in $\langle t\rangle$ is larger for longer wavelengths (the largest amplitude of change in $\langle t\rangle$ is observable at 836 nm—see Fig. 5, lower panel). For different emission wavelengths, the $\langle t\rangle$ signal reaches a maxima and minima at different times (Fig. 5, middle panel).

Fig. 5

Changes in number of fluorescence photons $\mathrm{\Delta }{N}_{\mathrm{tot}}$ and mean time of arrival of fluorescence photons $\mathrm{\Delta }\langle t\rangle$ for different emission wavelengths recorded after a $0.45\text{-}\mathrm{mg}/\mathrm{L}$ indocyanine green (ICG) bolus injection into the tube of the phantom.

Study of the dependence of the fluorescence signal on the ICG concentration in the boli injected into the tube showed that the amplitudes of change in $N_{\mathrm{tot}}$ and $\langle t\rangle$ depend on emission wavelengths (not published data).

3.3.

In Vivo Measurements

Results of the in vivo measurements carried out on two healthy volunteers are presented in Fig. 6. Total number of fluorescence photons, $N_{\mathrm{tot}}$, is shown for selected emission wavelengths as a function of the time of passage of an ICG bolus ($T$) and time in respect to laser pulses ($t$). The fluorescence signals were successfully detected in five (out of 16) spectral channels, covering about 52 nm in the range from 798 to 850 nm. Because of limited spectral resolution of the setup, the wavelength of the maximum fluorescence light intensity cannot be estimated precisely; however, it can be observed that the maximum fluorescence light intensity is between 811 and 823 nm. This result corresponds with our previous studies of ICG dissolved in solvents frequently used in multiple scattering media, in which the maximum fluorescence peak was found at $\lambda \approx 822\mathrm{nm}$.²⁹ A highly dynamic increase in the number of fluorescence photons starts about 20 s after injection of the dye, and the maximum is reached about 35 s after injection.

Fig. 6

Distributions of times of arrival of fluorescence photons following injection of ICG bolus at $T=0\mathrm{s}$. Panels refer to different emission wavelengths. Number of fluorescence photons are shown as a function of time of experiment $T$ and time of the photon with respect to the laser pulse.

Figure 7 illustrates changes in statistical moments of the distribution of times of arrival of fluorescence photons ($N_{\mathrm{tot}}$, $\langle t\rangle$) following the injection of the ICG bolus. The distribution of times of arrival of fluorescence photonsis connected with a rapid increase in the total number of fluorescence photons $N_{\mathrm{tot}}$ because of the increase of the concentration of the fluorophore in the volume of tissue which is penetrated by the photons. In signals of $\langle t\rangle$, the initial increase can be observed during the inflow of the ICG bolus followed by a pronounced decrease and a final return to the initial level. This pattern of change corresponds to the theory describing the earlier appearance of ICG in the brain, which is connected with the increase of $\langle t\rangle$, followed by the inflow of dye to the extracerebral layers of the head, and reflected by a decrease of $\langle t\rangle$.^23,34 For different emission wavelengths, $\langle t\rangle$ reaches its maximum at different times. Consequently, $\langle t\rangle$ reaches its minimum at different times $T$ for different spectral channels. This effect may be explained by different penetration depths of photons detected at different emission wavelengths, and hence, by different contributions of intracerebral tissue in the studied volume. Differences in amplitudes of change in $N_{\mathrm{tot}}$ and $\langle t\rangle$ have been observed for different emission wavelengths. The highest amplitudes for $N_{\mathrm{tot}}$ can be observed at wavelengths of 811–823 nm while distinctly lower amplitudes were noted at shorter and longer emission wavelengths. For both subjects studied, maxima in $\langle t\rangle$ appear at different emission wavelengths, but a low signal-to-noise ratio does not allow for any quantitative evaluation of this effect.

Fig. 7

Changes in number of fluorescence photons $\mathrm{\Delta }{N}_{\mathrm{tot}}$ and mean time of arrival of fluorescence photons $\mathrm{\Delta }\langle t\rangle$ following the injection of indocyanine green (ICG) bolus in healthy volunteers, measured for different emission wavelengths.

Despite high noise in $\langle t\rangle$ signals, it can be noted from the start of the experiment, that the time at which $\langle t\rangle$ reaches its maxima and minima (due to the inflow of ICG) is different for different emission wavelengths (Fig. 7, middle panel). A similar effect was observed for the signals detected during measurements on a phantom with dynamic inflow of ICG (Fig. 5, middle panel). This may suggest that the pattern of signals depends strongly on the spatial heterogeneity of optical properties in the medium at different wavelengths.