1. Introduction

Optical coherence tomography (OCT) is a rapidly developing optical imaging modality which exploits the intrinsic reflectivity of tissue samples to provide 2-D or 3-D tomographic images. OCT is based upon low-coherence interferometry where the axial depth sectioning is accomplished by coherence gating which ensures that only photons which traverse a similar (within the coherence length of the light source) optical path in both the reference and sample arm of the interferometer produce interferometric signal. Typical light source coherence lengths are ~10 µm with ultra-broad band sources extending to the region of 1 µm providing cellular and subcellular resolution, respectively. The imaging depth for light sources in the near infrared region of the spectrum is ~1-2 mm. The scattering cross-section which gives rise to the intrinsic tissue reflectivity does not vary widely among different molecular species, hence OCT does not inherently have the ability to measure molecular signatures for molecular imaging. Molecular imaging is an important functionality, as it potentially illuminates the biochemical makeup and biochemical processes occurring in the sample which are not apparent from simple reflectivity measurements.

A number of molecular spectroscopic techniques, including pump-probe [1, 2], second harmonic generation [3–5], coherent anti-Stokes Raman scattering [6], and linear absorption [7], have been adapted to OCT to provide molecular contrast, with varying degrees of success. We have discussed the shot-noise limited signal to noise ratios of the spectroscopic techniques in a previous publication [8]. Additional techniques based upon modification of the tissue scattering properties by molecular aggregates [9], engineered microspheres [10], and gold nanoparticles [11] have been developed with similar goals as the molecular spectroscopic techniques.

We have revisited pump-probe spectroscopy and developed a technique shown in reference [8] to have favorable signal to noise properties. This technique is more efficient than previous implementations because it measures the transient absorption on very short time scales before parasitic spontaneous processes (such as spontaneous emission) are able to depopulate the molecular excited state. Previous implementations relied upon steady-state kinetics to drive molecular population into either a triplet state [1] or isomeric state [2] over extended periods of time, during which competing spontaneous processes depopulate the excited state. Moreover, this pump-probe technique is more generally applicable since it does not require unique molecular species with efficient spin-forbidden or isomeric transitions.

 

Fig. 1. Energy level diagram for a ground-state recovery pump-probe experiment. The pump radiation transfers ground state population into the excited state. The probe radiation then measures the population transfer induced by the pump, which is manifested as a reduction in ground state absorption and an increase in excited state stimulated emission.

Download Full Size | PDF

The simplest implementation of this technique is one in which the pump and probe wavelengths are degenerate, which we call ground state recovery Pump-Probe Optical Coherence Tomography (gsrPPOCT). The essence of the technique is shown graphically in Fig. 1. Two laser pulses, the pump and probe, temporally separated by a delay time t_d , connect the ground and excited state of a molecular species. The pump pulse drives ground state population into the excited state. The probe pulse probes the population change induced by the pump beam by measuring the net difference in probe attenuation with and without the pump pulse present. The attenuation of the probe pulse is proportional to the net change in ground state population induced by the pump.

There are two distinct molecular properties which may be garnered from a ground state recovery pump-probe experiment, both of which contribute to the specificity of the technique, i.e., the ability to distinguish signals from multiple chromophores. The magnitude of the transient absorption may be measured as a function of the probe wavelength. These measurements map the absorption spectrum of the chromophore. Since only molecular states connected by both the pump and probe radiation lead to signal, the absorption spectrum recorded in this fashion is typically more simple than that recorded using an absorption spectrometer. Any measured shifts in the peaks of the absorption spectrum may also provide clues to the local environment of the chromophore.

The ground state recovery time is the time required for the molecules excited by the pump to spontaneously relax back to the ground state, which is indicative of a particular molecular chromophore and its local chemical environment. The ground state recovery time is acquired by measuring the magnitude of the transient absorption at various pump-probe delay times. Taken together, the ground state recovery time and transient absorption spectrum may provide a powerful tool to differentiate among multiple chromophores.

Currently, high resolution 3-D optical molecular imaging is largely dominated by fluorescence imaging, either confocal fluorescence or its nonlinear analog, two-photon fluorescence. Both typically require modification of the sample either by chemically binding fluorescent markers, or by genetic alteration of tissue to express one of the fluorescent proteins. These steps are required since most endogenous biochemically significant molecules fluoresce poorly, if at all. In addition to the technical difficulties involved in impregnating the sample with the fluorescent marker, there is also the potential that the fluorescent markers will interfere with the process under study [12] or even prove toxic to the sample [13].

The measurement of transient absorption avoids these issues. Any molecular species with an allowed transition to a bound excited state will exhibit transient absorption, hence it is not necessary to label molecules of interest. In particular, ground state recovery only requires knowledge of a single excited state, while pump-probe schemes which utilize additional excited states require knowledge of and the existence of additional excited states.

2. Materials and methods

The optical design which embodies this prototype implementation of time-domain gsrPPOCT is shown schematically in Fig. 2. The pump and interferometer light was generated by multipassing a KdP crystal with the output of a Nd:Glass laser with 150 fs pulse length and 72 MHz repetition rate to produce ~35 mW of 530 nm light with the first pass and ~8 mW with the second pass (33 µm coherence length in air). The 530 nm light generated in the first pass was split off with a dichroic mirror and used as the pump radiation. The pump was amplitude modulated at 30 kHz with an electro-optic modulator. The pump delay was accomplished using an optical delay line with a translation stage, which provided a variable delay of ~0.1 ns to 5.3 ns. For lifetime measurements, a corner cube retroreflector was mounted to the translation stage to mitigate systematic errors arising in the measured gsrPPOCT signal due to misalignment of the pump and probe beams over the travel of the translation stage. This configuration resulted in significant loses in pump power. For imaging, a mirror was placed on the translation stage held at a fixed delay of ~180 ps, which made ~10 mW of pump power available.

While the pulse duration at the exit of the laser cavity was 150 fs, no effort was made to compensate for the dispersion of the optical system, hence the pulses interacting with the sample were likely broadened and chirped. The broadened and chirped pulse would have no adverse effect on the measured signal, since to a good approximation the signal is independent of pulse duration and chirp. As long as the pulse width remains much smaller than the excited state lifetime, the model for the gsrPPOCT signal discussed below remains valid. Since the chromophores used in this study have excited state lifetimes on the order of 1 ns we may safely assume that the pulse width is much shorter than the chromophore excited state lifetime.

The interferometer was a conventional free-space time-domain OCT interferometer with a scanning galvanometer mounted retroreflector in the reference arm and alternatively a 20X, 10X, or 5X objective in the sample arm to focus the pump and probe radiation onto the sample. The sample arm beam of the interferometer also serves as the probe. The pump beam was coupled into the sample arm by using a polarizing beam splitter. The interferometric signal was measured on a silicon photo diode and demodulated with two lock-in amplifiers. The first amplifier was set at the Doppler frequency of 96 kHz corresponding to the scanning reference delay and used to record the OCT signal. The second amplifier was set at the difference frequency of the Doppler and pump modulation frequencies, 66 kHz, and used to record the gsrPPOCT signal.

Since polarization is used to separate the pump and probe radiation, any retardance induced by the sample will send pump light into the detection channel and probe light into the pump optical pathway. The addition of pump light in the detection channel will have no adverse effect on the signal to noise ratio as long as the noise floor is still dominated by the reference arm light. The loss of probe light to the pump optical pathway will reduce the signal to noise ratio (SNR) by an amount equal to the fraction of the probe light which is lost. In the extreme case where half of the probe light is lost to the pump optical pathway, there will be a 3 dB reduction in the SNR.

 

Fig. 2. Schematic diagram of the prototype time-domain gsrPPOCT system. Abbreviations are as follows; DM dichroic mirror, BS 50/50 beamsplitter, EOM electro-optic modulator, PBS polarizing beamsplitter, GT Glan-Thomson polarizer.

Download Full Size | PDF

3. Theory

The equations describing the signal to noise ratio of a gsrPPOCT system have been derived previously in Ref. [8]. A summary of the results of the derivation is given below, as a prelude to comparing the results predicted by theory to those measured on our gsrPPOCT system. For details of the derivation see Ref. [8]. The predicted gsrPPOCT SNR for a model two level molecular system is given by:

where R is the reflectivity, ρ is the detector responsivity, P_pr is the sample arm power in the OCT interferometer, B is the detection bandwidth, e is the electron charge, f₀ is the laser pulse repetition rate, N₁ (t) and N₂ (t) are the time dependent ground and excited state populations integrated over the pump pulse duration, $N_{\mathit{1}}^{\mathit{0}}$ is the ground state population before interacting with the pump laser, σ₁ is the absorption cross section, z is the absorption path length, λ_pu is the pump wavelength, P_pu is the average pump power, ω is the pump modulation frequency, h is Planck’s constant, c is the speed of light, and r is the focal spot radius. The final approximation is a result of assuming weak absorption of the pump and probe beams where the exponential function may be approximated by the first two terms of its Taylor series expansion.

Several additional assumptions were made in the derivation of Eqs. (1) and (2). The excited state population at t=0 was assumed to be zero, i.e. $N_{\mathit{2}}^{\mathit{0}}$ =0. Under this assumption each set of pump-probe pulses are independent of the sets coming before and after. With the laser repetition rate of our system (72 MHz) there are ~14 ns between pulses, hence for a chromophore with a ground state recovery time of 4 ns, ~3% of the excited chromophores will not have relaxed back to the ground state before the next pump pulse. If we assume a fairly strong pump radiation/chromophore interaction where ~10% of the chromophores are excited by the pump radiation, then ~0.3% of the total population will be inaccessible on the following pump pulse due to fact that they have not relaxed back to the ground state, i.e. $N_{\mathit{1}}^{\mathit{0}}$ is reduced by ~0.3%. The corresponding reduction in SNR (from Eq. (2)) would be ~0.6%. The pump pulse duration was also assumed to be much shorter than the excited state lifetime which simply allowed for the neglect of excited state population decay during the pump pulse.

The gsrPPOCT signal is inherently pathlength (z) and reflectivity (R) dependent (see Eq. 2). In order to remove these dependencies and more accurately reproduce the chromophore concentration map we have processed the recorded images according to the left hand side of the following equation;

The rhs of Eq. (3) is dependent upon quantities which remain constant or approximately constant throughout the sample with the exception of the initial ground state population ($N_{\mathit{1}}^{\mathit{0}}$ ), hence the resulting image effectively maps the chromophore concentration. Prior to the application of Eq. (3) both the gsrPPOCT and OCT images were spatially averaged in order to reduce the deleterious affects speckle noise would produce in the derivative. The derivative was computed by differencing adjacent points after thresholding the data to avoid computing the derivative of the noise floor. The axial resolution of the derivate image is determined both by the dz used in calculating the derivative and the axial point spread function. In this case the derivative was determined with a dz of 2µm, hence the axial resolution of the derivative image is dominated by the filtered axial point spread function.

After the pump radiation is turned off, the ground and excited states are no longer coupled via an electromagnetic field, hence the excited state population decays via spontaneous processes, repopulating the ground state. By measuring the gsrPPOCT signal at various pump-probe delay times we can map the rate at which the excited state repopulates the ground state. For a simple two-level system such as the one used to derive Eqs. (1–2), the gsrPPOCT signal recorded at different delay times takes the form of an exponential decay in pump-probe delay time. If multiple chromophores are present the total signal is the sum of contributions from each to give:

where t_d is the pump-probe delay time, a is a product of constants such as pump power and chromophore concentration, τ is the ground state recovery time, and only spontaneous processes depopulate the excited state [16]. From Eq. (4) we see that it should be possible to separate the contribution to the signal from multiple chromophores by measuring the gsrPPOCT signal as a function of time and fitting it to Eq. (4).

4. Results

In order to demonstrate that the gsrPPOCT signal was dependent on the modulation of the pump radiation and was free of artifacts from the frequency overlap of the gsrPPOCT and OCT signals we conducted the following series of measurements. Figure 3 reproduces a set of

A-lines (axial depth scan) recorded using the sample vessel in panel A with a concentrated suspension of 4T1 mouse mammary carcinoma cells expressing the protein dsRed. Trace B is the OCT signal recorded at the Doppler frequency, 96 kHz. The three interfaces of the sample vessel, corresponding to the top and bottom surface of the cover slip and the mirror surface, were clearly visible. Scattering from the cells in suspension was negligible. Trace C is the gsrPPOCT signal recorded at the difference frequency, 66 kHz. Only the mirror interface was visible. This observation was consistent with the fact that absorption is inherently cumulative, hence absorption in the cell suspension was reported at reflections distal to the absorber. Trace D is the gsrPPOCT signal at the difference frequency when the pump modulation is turned off. As expected, if the excited state population is not modulated by the modulation of the pump beam, there is no signal at the difference frequency. In contrast, the OCT signal remains as depicted in trace B. The results in trace D suggested that the lock-in amplifier, when tuned to 66 kHz with a time constant of 100 µs was effective in filtering the OCT signal with a suppression of at least 87 dB, hence all of the signal recorded at the difference frequency (66 kHz) was attributed to gsrPPOCT.

 

Fig. 3. A) Cartoon of the sample vessel used for characterizing the gsrPPOCT system. The vertical dimension of the cartoon is on the same scale as the A-lines in B, C, and D. B) OCT A-line recorded at 96 kHz. C) gsrPPOCT A-line recorded simultaneously with the OCT A-line in panel A at 66 kHz. D) gsrPPOCT A-line with amplitude modulation of the pump beam turned off.

Download Full Size | PDF

The measured SNR as a function of probe power, pump power, and concentration are show graphically in Fig. 4. In panels A and B, the vertical axis is the measured SNR on a linear scale and the horizontal axis is the power of the pump or probe laser. In both experiments the probe power was attenuated by a neutral density filter with an optical density of 2 resulting in an effective mirror reflectivity of-40 dB. The results in Fig. 4 have been corrected to perfect reflectivity. Each point on the plot represents the average of 20 measurements. The pump power was fixed at 2.8 mW for the measurements of probe power dependence and the probe power was fixed at 1.7 mW for the measurements of pump power dependence. The concentration of R6G was 56 µM for the power dependence measurements. The signal to noise ratio in a given A-line was taken to be the square of the ratio of the peak of the measured interferometric signal at the mirror interface to the standard deviation of the noise floor as determined at an axial position where no reflector was present.

The circles in Fig. 4(A) are the experimental measurements of SNR as a function of probe power. The theoretical prediction from Eq. (1) after compensation for 6 dB of system inefficiencies is shown as a gray line. The 6 dB of inefficiencies was arrived at by measuring the SNR of the OCT system which was 102 dB, approximately 6 dB less than the ideal shot-noise limit. The significant disagreement between the theoretical prediction and experiment is likely related both to the propagation of measurement errors and the presence of physical phenomena not explicitly included in the model. The former is due to the fact that several experimental parameters, such as beam spot radius, must be measured and inserted prior to calculating the experimental curve. The model was derived for a two level system, hence phenomena such as photobleaching are not accommodated. During the acquisition of the data presented in Fig. 4 we saw clear evidence for photobleaching both by visually observing a reduction of fluorescence and observing the reduction of the gsrPPOCT signal over time. While there was clear evidence for photobleaching, the gsrPPOCT signal remained steady for long periods of time (>30 min), after the initial losses. In order to bring the theoretical results into agreement with the measured results we had to modify the theoretical model to account for photobleaching. The most rigorous approach would have been to solve the system of differential equations which describe the time dependent population of each state populated in a ground state recovery experiment on rhodamine 6G (R6G), however the analytical solution (if it exists) would only be applicable to aqueous solutions of R6G. We chose a more simple approach which is more broadly applicable. We have treated the effects of the photobleached states as a perturbation of the ground state population. In this model the initial ground state population ($N_{\mathit{1}}^{\mathit{0}}$ ) is no longer equal to the chromophore concentration, but rather the chromophore concentration (C) times one minus the probability (P) that a chromophore is in a photobleached state, i.e., $N_{\mathit{1}}^{\mathit{0}}$ =C(1-P). We have found that setting P=0.33 brings the theoretical curve and experiment into good quantitative agreement. The determined value of P is likely also compensating for any systematic errors incurred in measuring the experimental conditions.

 

Fig. 4. A) Measured SNR as a function of probe power using a 56 µM sample of rhodamine 6G. B) Measured SNR as a function of pump power using a 56 µM sample of rhodamine 6G. C) Measured SNR as a function of rhodamine 6G concentration. The theoretical curves are based upon predictions using eq. 1.

Download Full Size | PDF

Panel B of Fig. 4 is the measured SNR as a function of pump power. As above, setting P=0.33 resulted in good quantitative agreement between the theory and experiment. In summary, both the probe and pump power dependence of the SNR show qualitative agreement with the theoretical curves and may be brought into quantitative agreement by setting P=0.33.

The results of the measurements of SNR as a function of chromophore concentration are provided in Fig. 4(C). At the concentration used for the power dependence measurements (56 µM), there was very good agreement between experiment and theory, which degraded as we moved to higher concentrations. This degradation is likely due to increased photobleaching with increasing concentration. While recording the data in Fig. 4(C), we noted an increase in the severity of the photobleaching, especially at the highest concentrations. Equation 2 indicates a quadratic dependence of the SNR on concentration, which held for the first few experimental data points of Fig. 4(C). However, at higher concentrations the assumption of weak absorption, which leads to the approximation in Eq. (2), breaks down. In fact, at sufficiently high concentrations the SNR curve turns over and the SNR gets smaller with higher concentration.

After optimizing and characterizing the gsrPPOCT system we moved on to testing the prototype gsrPPOCT system on biological samples. The initial target chromophore was dsRed, a genetically expressible protein commonly used for fluorescence imaging in cell biology, developmental biology, and models for disease. We have observed signal using the sample vessel depicted in Fig. 3 from an aqueous suspension of 4T1 mouse mammary carcinoma cells expressing dsRed. We have also observe gsrPPOCT signal from an adult transgenic zebra danio strongly expressing dsRed in its skeletal muscle [14]. The transgenic zebrafish, sold commercially as glofish®, were engineered in part to demonstrate their potential use as biosensors.

Figure 5 are representative images recorded along different sections of the glofish. The total power on sample was ~8 mW focused to a ~6 µm diameter spot with a 20X objective. At this irradiance, there was no photodamage observable either visually or in B-scans (cross-sectional image comprised of multiple A-lines) repeated over the same region. Each A-line in the image is the result of averaging 10 data acquisitions, resulting in an A-line acquisition rate of ~1 Hz. The lateral scan step size was 10 µm. Panel B is an OCT B-scan recorded along the back of the animal as indicated by the white box on the photo panel A. The OCT image in is fairly devoid of structure as would be expected for a cross-section through the fairly homogenous epaxial muscles running along the spine. Panel C is the overlay of the gsrPPOCT image onto the OCT image in panel B. The gsrPPOCT image shows signal essentially across the entire region shown in the OCT image which we attribute to the dsRed being expressed in the skeletal muscle.

The second set of images (panels E, F, and G) in Fig. 5 is a cross-section which bisects the pectoral fin and continues toward the lateral line as indicated by the lower white box in panel A. The pectoral fin and the surrounding tissue are readily visible in the OCT image (Fig. 5(E)). In contrast there is only very weak signal in the gsrPPOCT image (Fig. 5(E)) in the region of the pectoral fin. The region medial to the pectoral fin exhibits significant signal in the gsrPPOCT image. Note that the strongest signal in the OCT image exhibits negligible gsrPPOCT signal while the comparatively weaker portion of the OCT image medial to the pectoral fin exhibits significant gsrPPOCT signal. Figure 5(D) is the spatially filtered OCT image overlaid with the derivative image derived from Eq. (3).

 

Fig. 5. A) Photo of transgenic zebra danio fish expressing dsRed in its skeletal muscle. B) OCT cross-section recorded through the back of the fish, anterior to the dorsal fin as indicated by the top white box in the photo. C) Overlay of the corresponding gsrPPOCT image onto the OCT image in B. The color bar indicates the SNR of the gsrPPOCT signal (max 47 dB). D) Derivative image derived from panel C. E, F, G) Same as B, C, and D except recorded along a cross-section (bottom white box) bisecting the pectoral fin (PF) and continuing into the lateral line (LL). The maximum recorded SNR_gsrPPOCT in panel E was 37 dB. The scale box is 200 µm×200 µm.

Download Full Size | PDF

Hemoglobin, the major vehicle for oxygen transportation in animals, is found in large concentrations in red blood cells. The ability to measure a gsrPPOCT signal from hemoglobin, which has a strong absorption band near 530 nm, could potentially provide a method for 3-D angiography which does not require an exogenous contrast agent. Since hemoglobin has extremely weak fluorescence[15], fluorescence angiography is typically accomplished with exogenous contrast agents such as indocyanine green and fluorescien.

Figure 6 is a set of images recorded of the gills in a euthanized adult wild-type zebra danio fish. The photo in panel A, recorded on a light microscope, shows the gill filaments which have been exposed by surgical removal of the operculum. A 3-D image set was recorded with a 5 µm step size in both lateral directions. The total power on sample was ~4 mW focused to a ~17 µm focal spot diameter with a 5X objective. At this irradiance no photodamage was detectable either visually or in repeated B-scans over the same region. Each A-line in the image is the result of averaging 10 consecutive data acquisitions resulting in a line rate of ~1 Hz.

Panels B, C, and D are a set of summed voxel projections (SVP), produced by summing the pixels of each A-line in the 3-D set. The efferent filament arteries (eFA) appear as dark lines (low signal) in the OCT SVP image because of the shadow cast by the strong absorption of the hemoglobin. In contrast, the eFA appear as bright lines (strong signal) in the gsrPPOCT SVP. The overlay SVP confirms the correlation of the shadows in the OCT SVP with the bright lines in the gsrPPOCT SVP.

Panels E, F, and G are representative B-scans which bisect one of the two eFA observed in the SVP. The shadow induced by the strong hemoglobin absorption noted above is clearly visible in the OCT image. While the shadow provides circumstantial evidence that there is a blood vessel directly above the shadow, it is not possible in the OCT image to make out the borders of the blood vessel or to determine its exact location. In contrast, the gsrPPOCT only shows signal in the blood vessel. The derivative image (Fig, 6(G)) maps the borders of the blood vessel, removing contributions to the gsrPPOCT signal from sample reflectors below the blood vessel.

 

Fig. 6. A) Photo of the gills of a euthanized adult wild-type zebra danio fish. The operculum was removed in order to expose the gills. The three images on the right (B, C, and D) were produced by summing the pixels along the axial direction of a 300 µm×500 µm 3-D volume set. The filament arteries appear as dark stripes, due to shadows created by the strong absorption by hemoglobin. E) OCT B-scan taken from the 3-D volume. F) Overlay of the corresponding gsrPPOCT image onto the OCT image in E. The color bar indicates the SNR of the gsrPPOCT signal (max 19 dB). The scale box is 50 µm x 50 µm. G) Derivative image derived from panel F.

Download Full Size | PDF

Figure 7 includes the results of several ground state recovery (gsr) time measurements made using the sample vessel depicted in Fig. 3. Panel A contains the t_d dependent gsrPPOCT signal measured for a 156 µM solution of R6G. The experimental points were taken from the peak of the signal in the A-line from the mirror surface. Each A-line was generated by taking the average of 100 data acquisitions. The pump and probe power on sample was 650 µW and 65 µW, respectively. The light was focused onto the sample with a 10X objective, resulting in a focal spot diameter of ~11 µm. Systematic errors associated with the maintenance of pump beam alignment as the delay was varied resulted in a practical limit of approximately+/-1 ns in determining the ground state recovery times. The systematic errors were manifest as patterned oscillations in the signal amplitude as a function of pump-probe delay time. The experimental points were processed before fitting by subtracting a small DC offset determined by finding the mean signal over a portion of the A-line where no reflector was present. This eliminated the need to add a constant term to Eq. (4). In order to correct for small variations in the signal with respect to time, the gsrPPOCT signal was normalized to the OCT signal. Since we were imaging a static sample any variations in the OCT signal represented system variations. Before plotting the experimental points and fit, they were normalized such that α=1. The black line in the plot is the fit, which had an R ²=0.96 with τ=5 ns for R6G. Attempts at fitting the experimental data to two exponentials resulted in essentially identical lifetimes for both exponentials, indicating that the data is well represented by a single exponential. To our knowledge, the ground state recovery time of R6G has not been previously measured on this time scale. We would expect that since the fluorescence quantum yield of R6G is ~90%[17], the ground state recovery time should be on the order of the fluorescence lifetime which has been measured to be 4.08 ns [18] in aqueous solution. In an isolated two-level system the ground state recovery time and the fluorescence lifetime would be equivalent, however in a more complex system like R6G, multiple intermediate states will cause the two quantities to diverge.

 

Fig. 7. gsrPPOCT signal plotted as a function of pump-probe delay time using the sample vessel depicted in Fig. 3. The experimental points and fit are for the maximum of the recorded peak from the mirror interface. A) 156 µM solution of rhodamine 6G and human whole blood diluted with distilled water to ~20% by volume. B) Mixture of rhodamine 6G and human whole blood.

Download Full Size | PDF

The second set of data points in Fig. 7(A) are the gsrPPOCT signal as a function of pump-probe delay time for a sample of human whole blood, diluted with distilled water at a ratio of approximately 1 part whole blood to 4 parts distilled water. The experimental conditions were the same as noted above, with the exception that 500 data acquisitions were averaged to produce each A-line in order to compensate for the lower signal levels in the whole blood sample. The fit yielded a ground state recovery time for whole blood of 9 ns with an R ² of 0.97. As before, attempts at fitting the data to two exponentials resulted in essentially identical lifetimes for both exponentials. To the best of our knowledge, the ground state recovery time of whole blood or hemoglobin has never been reported in the literature for this wavelength.

Panel B of Fig. 7 is the signal as a function of pump-probe delay time for a sample of human whole blood mixed with R6G. The dilution ratio was similar to above with the exception that 156 µM R6G was substituted for distilled water. The experimental conditions were the same as for the experiment with distilled water diluted whole blood. The two measured lifetimes were 1 ns and 9 ns with an R² of 0.98. Since the 1 ns lifetime is largely determined by the first three data points depicted in Fig. 7(B), we place less confidence in the absolute magnitude of the this measured lifetime although we can be assured that it is much smaller than the 9 ns lifetime. The relative contributions of the two chromophores based upon the fit, were 80% for the 9 ns lifetime and 20% for the 1 ns lifetime.

The 9 ns lifetime is easily assigned to the signal due to whole blood, since it is in good agreement with the lifetime measured above. The 1 ns lifetime must then be assigned to the R6G signal, which implies that the ground state recovery time of R6G is quenched by interactions with the constituents of whole blood. Similar interactions have been noted for related dyes such as rhodamine 800 [19]. Likewise, R6G is well known to be readily quenched by interaction with iodide in solution [20]. Hence it is not surprising that R6G interacts with its environment, resulting in the quenching of the ground state recovery time.

A quantitative assessment of the ratio of whole blood to R6G is not possible without knowing the absorption cross-sections in this environment and a more precise measure of the dilution, however the 80:20 ratio is consistent with what we would expect based on our observations of the signal strengths from the two chromophores.

5. Discussion

The OCT SNR is attenuated at a rate of exp(-2µ_Tz), where µT is the total attenuation coefficient and the factor of two accounts for the pass in and out of the sample. Similarly the gsrPPOCT SNR is attenuated at a rate of exp(-2µ_Tz-2µT,p_uz), where µ_T,pu is the total attenuation of the pump radiation and the factor of two is a result of the fact that SNR_gsrPPOCT is quadratic in pump power. In the limit that we require the pump radiation to be minimally scattered, µ_T,pu=µ_T and SNR_gsrPPOCT is attenuated a factor of two faster than SNR_OCT. In the other extreme we may assume that µ_T,pu=µ_eff<<µ_T, where µ_eff is the effective extinction coefficient from diffusion theory and SNR_gsrPPOCT is attenuated at the same rate as SNR_OCT. The correct limit for µ_T,pu lies somewhere between these two extremes, since while we are not concerned with the degree of scattering the pump radiation undergoes, we do require it to remain within the focal volume of the probe radiation in order to be effective. Based on these two limits and the data provided in Fig. 6(E) and 6(F) we predict the maximum depth (SNR=1) to which we would be able to measure signal from a blood vessel in zebra danio gill filaments to be between 570–1000 µm. For this prediction we have estimated the value of µT from an A-line in the OCT image just to the left of the eFA in Fig. 6(F).

The specificity with which gsrPPOCT may isolate signal attributable to a specific chromophore is limited by the “background” of chromophores also present in the sample which have absorption spectra which overlap with the spectrum of the light source. Both hemoglobin and dsRed are present at significantly high concentrations in the glofish, hence we cannot positively say that all of the signal shown in the images of Fig. 5 is attributable to dsRed. The background signal from chromophores outside the blood vessels in the cross-section of Fig. 6(F) is at the level of the detector noise. However, there are two areas exhibiting signal in Fig 6(C), which are not easily attributable to hemoglobin in the blood vessels. The unidentified chromophore is labeled U in the image. Positive identification of tissue chromophores will be an ongoing issue for gsrPPOCT.

The collection of additional physical properties like the pump-probe absorption spectrum as outlined earlier in this article and the ground state recovery time as demonstrated above will help establish greater specificity for gsrPPOCT. Additional properties of the sample, beyond those measureable with gsrPPOCT, may also be used for specificity. For instance, another source of specificity applicable to hemoglobin imaging in living organisms is to use the fact that the blood is flowing and combine gsrPPOCT with Doppler OCT(DOCT). DOCT provides complimentary information for living organisms by mapping out the vasculature based upon its sensitivity to blood flow. The Doppler shift is dependent on the projection of the velocity vector of the scatterer onto the optical axis, hence the amplitude of the DOCT signal is angular dependent with its maximum when the blood flow is parallel to the optical axis and zero when the blood flow is perpendicular to the optical axis. Since all the specimens in this study where euthanized before imaging, the DOCT signal was not measured simultaneously. Future studies in living animals will likely utilize this source of specificity.

While this work demonstrates the effectiveness of gsrPPOCT as a molecular imaging extension of OCT, the development of a robust imaging system based on this technique which is capable of in vivo imaging will require further, albeit straightforward engineering development. This will require line rates fast enough to mitigate motion artifact inherent to imaging living organisms as well as reduction of the laser fluence below the tissue damage threshold for animal imaging and the more stringent ANSI [21] limits for human subjects. For hemoglobin imaging in zebra danio, the gsrPPOCT system described here had a line rate of ~1Hz with laser fluence approximately a factor of 4 above the ANSI limit for human skin exposure based on a 1s exposure time. Future implementations of gsrPPOCT will harness the recent technological advances in spectral domain OCT, which provides between 100–1000 fold increase in sensitivity [22–24] over the time-domain technique used here. The increased sensitivity will be used to increase the line rate to permit imaging at ~1 image/sec, and will have the additional benefit of reducing the laser exposure to well below the ANSI limit due to the lower exposure time required.

6. Conclusions

We have developed a novel molecular imaging technique which is sensitive to a wide range of molecular chromophores, including many important biomolecules such as hemoglobin which are not directly accessible with standard fluorescence techniques because of their extremely small fluorescence quantum yield. The sensitivity of the prototype gsrPPOCT system has been shown to be in reasonable agreement with theoretical predictions. We have demonstrated the potential for gsrPPOCT imaging with tissue samples including transgenic and wild-type zebra danio to image the in situ spatial distribution of the genetically expressible protein dsRed and the oxygen transporting protein hemoglobin, respectively. The ground state recovery time of R6G and hemoglobin has been measured in separate samples and as a mixture. The amplitudes determined from fitting the ground state recovery of the mixture to eq. 4 were used to deduce the relative contribution of each chromophore to the total measured gsrPPOCT signal, thus illustrating the potential to use ground state recovery times to separate the gsrPPOCT signal from multiple chromophores. The continued development of gsrPPOCT holds the potential for the realization of a robust highly sensitive and specific molecular imaging technique which is capable of simultaneously imaging the spatial distribution of multiple chromophores. The simultaneous measurement of important chromophores like oxy and deoxy hemoglobin may pave the way for important applications including 3-D spatially resolved oximetry.