2.1.

Near-Infrared Functional Lymphatic Imaging System Setup

The NIR lymphatic imaging device, which is depicted in Fig. 1, was developed using a 150 mW 808 nm laser diode (Thorlabs part no. M9-808-0150) powered by accompanying diode driver and temperature control boxes to provide excitation light. A 20 deg beam diffuser (Thorlabs part no. ED1-C20) was mounted in front of the diode to achieve a uniform excitation field of approximately $75{\mathrm{cm}}^2$ with less than $1.9\mathrm{mW}/{\mathrm{cm}}^2$. Fluorescence emission centered at 840 nm was captured using a PIXIS 1024B back-illuminated CCD camera (Princeton Instruments) with an attached Infinity K2/SC video microscope lens (Edmund Optics) and a bandpass filter ($\mathrm{CW}:840\mathrm{nm}$, $\mathrm{FWHM}:15\mathrm{nm}$, Omega Optical). NIR images were recorded via a custom LabView (National Instruments) image acquisition code.

Fig. 1

NIR lymphatic imaging system schematic. Excitation light is provided by 150 mW 808 nm laser diode powered by accompanying diode driver and temperature control boxes. Emission light centered at 840 nm is captured by a CCD Camera with an Infinity K2/SC video microscope lens and a bandpass filter ($\mathrm{CW}:840\mathrm{nm}$, $\mathrm{FWHM}:15\mathrm{nm}$).

2.2.

ICG Solution Preparation

To determine the optimal excitation and emission wavelength of ICG for use in NIR lymphatic imaging, we created an albumin-physiological salt solution (APSS) (in mM: 145.00 NaCl, 4.7 KCl, 2.0 ${\mathrm{CaCl}}_2$, $1.17{\mathrm{MgSO}}_4$, $1.2{\mathrm{NaH}}_2{\mathrm{PO}}_4$, 5.0 dextrose, 2.0 sodium pyruvate, 0.02 EDTA, 3.0 MOPS, and $10\mathrm{g}/\mathrm{L}$ bovine serum albumin)³¹ designed to mimic interstitial fluid, and we added a commonly used concentration of $250\mu \text{g}/\mathrm{mL}$ of ICG (Across Organics)²² to simulate an injection of ICG into the interstitial space. For comparison purposes, the same concentration of ICG was also dissolved in 0.9% saline water. Peak excitation of both solutions was recorded using a spectrophotometer (Hitachi U-2900), and peak emission at the previously recorded peak excitation was recorded using a fluorometer (Shimazu RF-1501).

The optimal ICG solution for maximizing fluorescence yield within the dermal layer was determined by dissolving various concentrations of ICG ranging from 0.01 to $1,000\mu \text{g}/\mathrm{mL}$ in 0.9% saline and in APSS solutions with albumin concentrations ranging from 0 to $100\mathrm{g}/\mathrm{L}$. The various solutions were flowed through the tissue phantom at a depth of 2 mm to simulate flow through a dermal lymphatic vessel. The vessel was imaged using the NIR system, and fluorescence intensity was recorded for each sample to determine the optimal ICG and albumin concentrations. In order to quantify the enhancement of premixing ICG with albumin, follow-up testing was performed in vivo to compare the signal to noise ratio (SNR) of the optimal ICG/albumin solution ($150\mu \text{g}/\mathrm{mL}\mathrm{ICG}+60\mathrm{g}/\mathrm{L}\text{albumin}$) and ICG alone ($150\mu \text{g}/\mathrm{mL}$ ICG) both at the injection site and 10 cm downstream in the collecting vessel, where SNR was calculated as $\mathrm{SNR}=20\times \log (\frac{\text{Fluorescence}}{\text{Background}})$. Functional lymphatic testing (detailed below) was also performed to verify that premixing ICG with albumin does not alter lymphatic function, as measured by transport time, packet frequency, and packet velocity.

2.3.

Tissue Phantom Preparation

In order to characterize the parameters of NIR lymphatic imaging in the dermis, a tissue phantom was created with the same optical properties as the dermal layer. Mock lymphatic vessels of known diameters were created in the tissue phantom at known depths, thus affording complete control over the phantom and allowing idealized characterization of NIR imaging capabilities regarding spatial resolution and signal penetration depth. As can be seen in Fig. 2, the tissue phantom was molded in a standard petri dish using a mixture of 97.52% silicone elastomer base (Sylgard 184, Dow Corning), 2.22% Aluminum Oxide (Sigma Aldrich), and 0.26% cosmetic powder (Max Factor Crème Puff Deep Beige 42) according to previously published methods.³² Channels were created in the tissue phantom molds by suspending standard copper electrical wire of known diameters at known depths in the mold prior to curing and removing the wires after curing. The tissue phantom was also connected to a syringe pump (PHD 2000, Harvard Apparatus) to flow various ICG solutions through the mock vessels for imaging, a schematic of which is depicted in Fig. 2(d).

Fig. 2

Tissue phantom schematic and operation. (a) Tissue phantoms were molded in standard petri dishes using a mixture of 97.52% silicone elastomer base, 2.22% Aluminum Oxide, and 0.26% cosmetic powder. Channels were created in the tissue phantom molds by suspending standard electrical wire of known dimensions at known depths. (b) Image of the resulting channels after the molds are cured and the wires are removed. This is an example image created using 100% silicone to allow visualization of the channels. (c) Image of the final tissue phantom construct in which the channel outlets can be seen protruding from the side of the phantom. (d) The tissue phantom was connected to tubing containing preloaded “packets” of ICG to test the spatial and temporal resolution of the NIR imaging system. The flow rate through the tissue phantom was precisely controlled with a syringe pump, and the ICG packets were imaged as they passed through the phantom.

2.4.

Sensitivity Analysis of NIR System

Mock vessels of 1 mm diameter were created in a phantom at depths ranging from 1 to 10 mm in 1 mm increments, the optimal ICG solution ($150\mu \text{g}/\mathrm{mL}\mathrm{ICG}$, $60\mathrm{g}/\mathrm{L}\text{albumin}$) was loaded into each of the mock vessels, and fluorescence intensity was recorded at each depth in order to characterize the change in signal sensitivity of the NIR system as a function of depth. Vessel diameter calculations were performed at all depths to characterize the scattering effect on apparent vessel diameter. Fluorescence intensity measurements were recorded for four conditions to quantify excitation light leakage: (1) CCD shutter closed (background), (2) excitation light source on without ICG in the phantom, (3) low concentration of ICG ($1\mu \text{g}/\mathrm{mL}+60\mathrm{g}/\mathrm{L}\text{albumin}$), and (4) ideal concentration of ICG ($150\mu \text{g}/\mathrm{mL}+60\mathrm{g}/\mathrm{L}\text{albumin}$).

ICG typically flows through lymphatic vessels in the form of discrete packets,^15,20 presumably due to valve closure that is known to occur during periods of short flow reversal, when a favorable pressure gradient exists to close the valve.^33,34 To mimic the pulsatile packet flow of ICG in lymphatic vessels, mock ICG packets were created by preloading a length of tubing with drops of ICG solution separated by olive oil (to prevent mixing of the ICG packets through diffusion). A syringe pump was then used to flow the packets of ICG through the tissue phantom at known velocities to test the accuracy of a custom lymph velocity quantification algorithm we developed. The algorithm, a similar version of which was first reported by Sharma et al. in 2007,²⁰ utilizes line intensity profiles placed sequentially along a lymphatic vessel in the direction of flow at known distances from each other [Fig. 3(a)]. The line intensity profiles record spikes when packets pass over that particular area [Fig. 3(b)], and by measuring the time between spikes in the three sequential line profiles, average velocity of packets can be calculated.

Fig. 3

Quantification of ICG packet travel through tissue phantom. Packets of ICG were created by separating small amounts of ICG with olive oil (to prevent mixing of separate packets) in a segment of tubing connected to the tissue phantom. A syringe pump was used to precisely control the flow rate of the fluid through the tubing/tissue phantom construct. A custom code was used to process the data by calculating intensity values over three line profiles placed sequentially along the channel. (a) Example image of ICG packets flowing through the tissue phantom at a depth of 1 mm. The three lines depicted show the placement of the three line integrals used in the processing algorithm to detect fluorescence intensity. (b) Example of the fluorescence intensity plots at the three line intensity profiles over time. Peaks in fluorescence intensity correspond to packets traveling over the lines. Fluid velocity can be calculated using the known dimensions of the channel and the time intervals between packets reaching sequential line profiles.

2.5.

In Vivo Imaging

Lymphatic function was quantified in vivo in the tail of six-week-old female hairless rats (Charles River Laboratories, Wilmington, MA) that were divided into a treatment group and a control group ($n=3$). The treatment group received a topical application of a glyceryl trinitrate ointment (GTNO) ($0.2\%\mathrm{wt}/\mathrm{wt}$, Rectogesic, Care Pharmaceuticals, commercially available), which is an ointment with an NO donor group that has previously been reported to slow lymphatic transport time.³⁵ The control group did not receive any topical treatment. Both groups were anesthetized with an intramuscular injection of Fentanyl ($0.12\mathrm{mg}/\mathrm{kg}$), Droperidol ($6\mathrm{mg}/\mathrm{kg}$), and Diazepam ($2.5\mathrm{mg}/\mathrm{kg}$ given 10 minutes after Fentanyl/Droperidol). The treatment and control groups were then both given 10 μL intradermal injections of ICG ($150\mu \text{g}/\mathrm{mL}$ ICG, $60\mathrm{g}/\mathrm{L}$ albumin) in the tip of the tail (given one minute after the GTNO application for the treatment group).

The NIR lymphatic imaging system was positioned such that the excitation source and the field of view of the CCD emission detector were centered on the rats’ tail 10 cm downstream (towards the base of the tail) from the injection site at the tip of the tail. The animals were imaged continuously from the time of injection until 20 min post-injection with a camera exposure time of 0.05 s. To evaluate lymphatic function in each of the rat subjects, three parameters were measured: the time necessary for the bolus injection of ICG to travel the 10 cm distance from injection site to emission recording site (transport time), the average velocity of the packets traveling through the field of view of the recording site, and the average frequency of packets passing through the field of view.

The transport time was calculated as the time between ICG injection and the arrival of fluorescence in the field of view 10 cm downstream from the injection site. The arrival of fluorescence was defined as a 20% increase in fluorescence intensity in the collecting vessel. An example of fluorescence arrival in the collecting vessel can be seen in Video 1, and a plot of fluorescence intensity over time during fluorescence arrival can be seen in Fig. 4.

Fig. 4

Fluorescence intensity over time during fluorescence arrival. (a) Image showing location of line profile for fluorescence arrival example. (b) Example plot of intensity versus time during arrival of fluorescence. (Video 1) Example video of ICG fluorescence arrival in the collecting vessels of a normal rat 10 cm downstream from the intradermal injection. The timestamp represents time since injection. (Video 1, QuickTime, 4.2 MB). [URL: http://dx.doi.org/10.1117/1.JBO.17.6.066019.1]

.

Packet frequency and velocity were measured using plots of fluorescence intensity over time generated from two regions of interest (ROIs) in a collecting vessel. ROIs were placed approximately 3 to 6 mm apart in regions of the vessel exhibiting large fluctuations in fluorescence intensity over time, where packet movement could easily be visualized and quantified. Packet frequency and velocity measurements began 10 frames after the arrival of fluorescence (to allow fluorescence values to stabilize) and measured for a duration of 10 packets. Of the two vessels in the tail, measurements were taken only on the vessel first producing fluorescence. Average packet frequency was calculated as 10 packets divided by the time necessary for 10 packets to occur (in minutes). Average packet velocity was calculated as the distance between the two ROIs divided by the average time necessary for packets to travel between the two ROIs (as indicated by peaks in the intensity plots). Figure 5 shows control and GTNO treatment examples of ROI selection and intensity versus time plots of the 10 packets used for frequency and velocity measurement. Videos 2 and 3 show the 10 packet segment of ICG flowing through the collecting vessels associated with the intensity plots in Fig. 5 for normal and GTNO conditions, respectively.

Fig. 5

Example intensity plots over time for normal and GTNO treatment conditions. (a) Image showing location of line profiles for normal condition example. (b) Image showing location of line profiles for GTNO treatment condition example. (c) Example plot of intensity versus time for normal condition. (d) Example plot of intensity versus time for GTNO treatment condition. (Video 2) An example video of ICG flowing through the two collecting vessels in the tail of a normal rat. The video lasts for the duration of 10 packets and is used to calculate packet frequency and velocity. The timestamp represents time after fluorescence arrival. (Video 3) An example video of ICG flowing through the two collecting vessels in the tail of a rat after GTNO treatment. The video lasts for the duration of 10 packets and is used to calculate packet frequency and velocity. The timestamp represents time after fluorescence arrival (Video 2, QuickTime, 3.1 MB); [URL: http://dx.doi.org/10.1117/1.JBO.17.6.066019.2] (Video 3, QuickTime, 12.6 MB) [URL: http://dx.doi.org/10.1117/1.JBO.17.6.066019.3].

To calculate the average delay time between contractions, we wrote a Matlab script that analyzes a given video sequence to find the region of highest fluctuation within the vessel. The fluorescence in this region was then quantified as a function of time, and that signal was processed by the code to calculate the average number of frames for each interval in which there was no fluorescence fluctuation. This value was multiplied by the time interval between frames and reported as the average delay time, $t_d$, for that vessel.

The data for each of the three functional imaging parameters were averaged for the treatment and control group, subsequently checked for normality using the Anderson-Darling test, and analyzed for statistical significance using a two-sample $t$-test.

3.1.

ICG Spectrum and Fluorescence

The excitation peaks for ICG dissolved in saline and APSS were approximately 785 and 805 nm, respectively, while the emission peaks of ICG in saline and APSS were approximately 815 and 840 nm, respectively (Fig. 6).

Fig. 6

ICG dissolved in albumin salt solution (APSS) exhibits a shift in the excitation/emission spectrum. Excitation and emission curves were generated for ICG ($250\mu \text{g}/\mathrm{mL}$) dissolved in 0.9% saline (dotted lines) and APSS (solid lines). A shift of approximately 20 nm was observed in the excitation spectrum of ICG dissolved in APSS versus saline, with peak fluorescence occurring at 805 and 785 nm, respectively. A shift of approximately 25 nm was observed in the emission spectrum of ICG dissolved in APSS versus saline, with peak fluorescence occurring at 840 and 815 nm, respectively. These spectra guided the design of excitation and emission detection wavelengths of the NIR imaging system.

ICG fluorescence is highly dependent on albumin concentration, and the intensity reached a maximum at an albumin concentration of $60\mathrm{g}/\mathrm{L}$ (902.8 μM) [Fig. 7(a)], and the maximum fluorescence yield at this albumin concentration was produced at an ICG concentration of $150\mu \text{g}/\mathrm{mL}$ (193.5 μM) [Fig. 7(b)]. Thus the solution producing maximal fluorescence was $150\mu \text{g}/\mathrm{mL}$ ICG and $60\mathrm{g}/\mathrm{L}$ albumin. When injected into a rat tail, premixing $150\mu \text{g}/\mathrm{mL}$ ICG with $60\mathrm{g}/\mathrm{L}$ albumin produced a greater SNR as compared to $150\mu \text{g}/\mathrm{mL}$ ICG alone with more than a four-fold increase in SNR observed in the collecting vessels [Fig. 7(c)] ($p<0.05$). Additional functional lymphatic testing was performed in response to ICG and ICG/albumin injections, and no significant differences were observed in transport time, packet frequency, or packet velocity (Fig. 8).

Fig. 7

Prebinding ICG with albumin enhances fluorescence and the resulting SNR upon in vivo intradermal injection. Fluorescence intensity through the phantom at a depth of 2 mm was measured for various concentrations of ICG and albumin to optimize the two concentrations to produce maximum fluorescence intensity of the ICG solution. (a) Peak ICG fluorescence intensity was measured as a function of albumin concentration in APSS ranging from $0\mathrm{g}/\mathrm{L}$ albumin to $100\mathrm{g}/\mathrm{L}$ albumin. Peak ICG fluorescence was produced at $60\mathrm{g}/\mathrm{L}$ albumin. (b) Fluorescence intensity of ICG dissolved in $60\mathrm{g}/\mathrm{L}$ was measured as a function of ICG concentration to determine the final concentrations of the optimal ICG solution for producing maximum fluorescence intensity. Maximum fluorescence intensity was measured at an ICG concentration of $150\mu \text{g}/\mathrm{mL}$. (c) $150\mu \text{g}/\mathrm{mL}$ ICG solution and $150\mu \text{g}/\mathrm{mL}$ ICG premixed with $60\mathrm{g}/\mathrm{L}$ albumin were injected into rat tails, and the signal to noise ratio (SNR) was calculated for each sample at the injection site and 10 cm downstream in the collecting lymphatic vessel. ICG premixed with albumin produced more than a four-fold increase in SNR compared to ICG alone from 1.8 to 7.8 dB in the collecting vessels and an increase from 10.9 to 14.2 dB at the injection site. Error bars represent standard deviation. $*=p<0.05$.

Fig. 8

Premixing ICG with albumin does not alter lymphatic function compared to ICG alone. Results of functional lymphatic testing reveal no significant differences between injection of ICG alone and ICG plus albumin in the tails of rats ($n=3$). (a) No significant difference in transport time to travel 10 cm. (b) No significant difference in packet frequency. (c) No significant difference in packet velocity.

3.2.

Tissue Phantom Sensitivity Analysis

The minimum detectable concentration of ICG at 2 mm depth was $0.1\mu \text{g}/\mathrm{mL}$, and ICG emission was detectable as deep as 6 mm with the signal at depths below 7 mm being indistinguishable from background (Fig. 9). Quantifying excitation light leakage showed a four-fold increase of signal over the thermal noise background. However, even low ICG concentrations produced a signal much larger than that due to leakage, and values of fluorescence typically seen in the vessel in vivo have fluorescence intensity values 14-fold greater than the excitation light source (Fig. 10). Vessel diameter calculations were very accurate at a depth of 1 mm with a 0.74% error, but error increased with depth to over 1,000% at 5 mm and was incalculable beyond 5 mm due to excessive scattering. The results also show that the calculated velocities were within 1% of the true velocities over a range from 0.15 to $1.5\mathrm{mm}/\mathrm{s}$ (Fig. 11).

Fig. 9

ICG can be detected up to a depth of 6 mm with minimal loss of spatial resolution at a depth of up to 3 mm. The optimal concentration of ICG solution ($150\mu \text{g}/\mathrm{mL}$ ICG, $60\mathrm{g}/\mathrm{L}$ albumin) was flowed through the tissue phantom at depths between 1 and 10 mm in 1 mm increments to determine how signal sensitivity changes with depth. (a) Example images of ICG flowing through the tissue phantom from 1 to 6 mm, which was the depth limit of detection. (b) Plot of minimum detectable ICG concentration at 2 mm depth in the tissue phantom. Minimum detectable ICG concentration was $0.1\mu \text{g}/\mathrm{mL}$. (c) Plot of ICG fluorescence intensity as a function of depth showing fluorescence intensity decreased successively with depth until 7 mm, which was indistinguishable from background. The depth limit of signal detection was 6 mm. (d) The apparent diameter of the channels at each depth was measured and compared to the true diameter of the channel to determine the accuracy of vessel diameter detection as a function of depth. At 1 mm, there was a 0.74% error between the true diameter and the measured diameter. Percent error increased with depth to a maximum of 1,095.06% error at 5 mm.

Fig. 10

Characterization of excitation light leakage. Intensity values were quantified for four conditions: (1) CCD shutter closed (background), (2) excitation light source on and phantom present without ICG, (3) low concentration of ICG in tissue phantom ($1\mu \text{g}/\mathrm{mL}+60\mathrm{g}/\mathrm{L}$ albumin), and (4) ideal concentration of ICG ($150\mu \text{g}/\mathrm{mL}+60\mathrm{g}/\mathrm{L}$ albumin). The results show that the excitation light produces a four-fold increase in intensity over background, but the ideal concentration of ICG produces a 14-fold increase in intensity over the excitation light source, which corresponds to a SNR of 23.1.

Fig. 11

Calculated packet velocity predicts true velocity with less than 1% error. Packets were flowed through the tissue phantom at a depth of 3 mm at velocities ranging from 0.15 to $1.5\mathrm{mm}/\mathrm{s}$. Velocities were calculated using a custom algorithm and compared to the known true velocities. Calculated velocities were accurate to within less than 1% error.

3.3.

Quantifying Functional Effects of NO on Lymphatics In Vivo

Application of GTNO significantly reduced lymphatic function (Fig. 12). Video 2 shows normal lymph propulsion through the tail of a control rat, while Video 3 shows significantly reduced lymphatic pump function in a rat after GTNO application. Transport time increased from 60 s under normal conditions to 414 s after GTNO application ($p<0.01$). Packet frequency decreased from 5.92 packets per minute and under normal conditions to 3.1 packets per minute after GTNO application ($p<0.05$). Packet velocity decreased from $1.5\mathrm{mm}/\mathrm{s}$ under normal conditions to $0.48\mathrm{mm}/\mathrm{s}$ after GTNO application ($p<0.05$). GTNO application decreased effective contraction length from 17.6 to 5.1 mm ($p<0.0005$). Contraction duration after GTNO application was significantly increased from 4.1 to 5.4 s ($p<0.05$), and systolic pumping power per unit mass was drastically reduced after GTNO application from 1.25 to $0.024{\mathrm{mm}}^2/{\mathrm{s}}^3$.

Fig. 12

Dermal nitric oxide delivery significantly reduces lymphatic pump function. 10 μL of ICG ($150\mu \text{g}/\mathrm{mL}$ ICG, $60\mathrm{g}/\mathrm{L}$ albumin) was injected intradermally into the tip of the tail of hairless rats divided into a treatment group that received a topical application of glyceryl trinitrate ointment (GTNO) prior to ICG injection ($n=4$) and a control group that did not receive any treatment prior to ICG injection ($n=4$). The NIR lymphatic imaging system was positioned to view the tail 10 cm downstream (towards the base of the tail) from the injection site. (a) The time required for the initial bolus injection of ICG to travel 10 cm downstream (transport time) significantly increased after GTNO application. (b) Packet frequency was significantly reduced after GTNO application. (c) Packet velocity was significantly reduced after GTNO application. (d) Effective contraction length was significantly decreased after GTNO application. (e) Contraction duration was significantly increased after GTNO application. (f) Contraction power per unit mass was decreased after GTNO treatment. $*=p<0.005$.

4.1.

Effects of Protein Binding on ICG Fluorescence

The NIR lymphatic imaging system that we developed in this study represents a departure from the setup of many of the NIR lymphatic imaging systems previously reported in that we premixed ICG with albumin, and our system used an excitation wavelength of 808 nm and emission wavelength centered at 840 nm.^18,21 Previous systems have used excitation sources of 785 nm, presumably because of the large availability of diodes at this wavelength. Our results indicate that ICG produces more than a three-fold increase in fluorescence when it binds to albumin, and the peak excitation and emission wavelengths are 805 and 840 nm, respectively. The same effect is observed when ICG is introduced in APSS, thus suggesting that ICG binds to albumin in the interstitial space. Therefore, ICG-based NIR lymphatic imaging systems that excite at 808 nm and capture emission centered at 840 nm will achieve higher SNR.

ICG has previously been shown to rapidly and completely bind to albumin in plasma.³⁶ Given that albumin concentration in the interstitium is approximately half of its concentration in plasma,³⁷ and the albumin concentration in lymph has been measured to be about 40% of its value in plasma,³⁸ it is reasonable to assume that all of the ICG present in lymph is bound to albumin as well. This assumption is further justified by the fact that the molecular weight of ICG (775 daltons) does not preclude it to lymphatic partitioning. Thus, the preferential uptake of ICG into lymphatics that is observed following dermal injections suggests that it must be bound to something of a large enough size to require lymphatic transport. Since albumin is the most prevalent soluble protein in the interstitium, is preferentially taken up into lymphatics after a dermal injection, and binds readily to ICG, it follows that even after dermal injection of ICG alone, the dye in the lymph is bound to albumin. Premixing ICG with albumin prior to injection thus not only increases the fluorescence of the dye, but it also eliminates interstitial albumin availability as a limiting factor in ICG uptake into lymphatics.

This approach to ICG delivery could be of particular importance when using this imaging technique in pathologies such as lymphedema, as the disease often results in accumulation of macromolecular proteins in the interstitium³⁹ that could significantly influence the uptake of injected ICG, confounding the interpretation of the experimental data. It is important to note that the injection of 10 μl of $60\mathrm{mg}/\mathrm{ml}$ albumin solution, while a very small volume, will disrupt the local gradients governing plasma filtration, temporarily increasing fluid extravasation from the blood and thus lymph formation. However, these values are well within the range of what the lymphatics would be expected to resolve during a mild inflammatory event, as average flow rates in a collecting lymphatic of fasted rats have been reported to range from $40\mathrm{nl}/\min$ to $200\mu \text{l}/\min$ depending on the vessel size and state of hydration.^33,40

It should be noted that Ashitate et al. recently reported that ICG alone was a better fluorophore for lymphatic visualization in the thoracic duct than ICG prebound to albumin,⁴¹ but there are several differences in experimental setup and technique worth exploring. Firstly, the NIR imaging system they employ excites at 760 nm, while our system is optimized to excite ICG bound to albumin, which is maximally excited at 805 nm. Their experimental setup also does not require imaging through the dermis and thus does not have to account for scattering and absorption effects, since most scattering and absorption occurs in the dermis. Interestingly, Ashitate and colleagues report a SNR for ICG of about 2, which is very similar to our results for ICG in collecting vessels. Given that we also report a SNR of nearly 8 for ICG bound to albumin in collecting vessels, we are confident prebinding ICG to albumin provides a more fluorescent tracer. Translating this technique into the clinic will produce additional regulatory challenges, but premixing the dye with autologous serum prior to dermal injection could provide one route of protein-bound ICG delivery.

4.2.

Tissue Phantom Sensitivity Analysis

The tissue phantom was constructed to recapitulate characteristics of lymphatic vessels in vivo that are essential to parameters historically quantified in NIR imaging, such as vessel morphology and propulsion frequency and velocity. Specifically, we constructed channels of similar size to lymphatics and imbedded them in a phantom with effective absorption and scattering coefficients of skin at depths characteristic of dermal lymphatics in vivo. According to our tissue phantom sensitivity analysis, the NIR lymphatic imaging system was capable of detecting ICG fluorescence as deep as 6 mm. However, scattering effects resulted in a deterioration of spatial resolution with increasing depth, and geometric vessel features became difficult to accurately identify below a depth of 3 mm. These results suggest that the NIR lymphatic imaging system is best used to detect vessel geometry and architecture above a 3 mm depth, but an assessment of gross ICG accumulation and transport in vivo can be obtained as deep as 6 mm (or perhaps deeper if features being resolved are greater than 1 mm, such as lymph nodes). Given that the average human skin layer is between 1 and 3 mm thick,⁴² these imaging characteristics are well suited for imaging dermal lymphatic function. Clinically, however, lymphatic diseases often result in a severe remodeling of the dermis, and fibrosis and lipid deposition can increase the thickness of the dermis well beyond this 3 mm limit.³⁹

In addition to chronic lymphedema resulting in a thickening of the dermis, it is likely that the optical properties of the tissue itself would change as the angiogenesis, adipogenesis, and fibrosis often associated with lymphedema would change the absorption and scattering coefficients of the dermal layer. Therefore, care should be taken in interpreting clinical data from ICG injections in patients with lymphatic disease, as the appearance of hyperplastic or dilated lymphatics could be due in part to changes in the thickness and the optical properties of the diseased limb, thus increasing the apparent diameter of vessels in these patients. Future studies are warranted to determine how exactly these changes would affect the ability of NIR imaging to assess lymphatic function in diseased patients.

The primary tool in functional ICG imaging is the ability to quantify the kinetics of packet flow in lymphatic vessels in vivo that presumably occur due to the fluctuating pressure gradients in coordination with lymphatic valves creating segmented flow of the dye. While our phantom does not contain these valves or the intrinsic mechanics that promote lymph transport, we have artificially reproduced this packet flow at a physiologically relevant depth in the phantom to quantify our system’s accuracy for measuring packet velocity in the presence of a scattering dermal layer, and we have demonstrated excellent accuracy in measuring velocity. Most NIR lymphatic imaging is performed giving an intradermal ICG injection and monitoring transport through dermal collecting vessels, which we have validated can be achieved with our device with a high degree of accuracy. Future work to enhance the device should focus on the implementation of diffusion theory (e.g., using a Monte Carlo approach to predict light propagation through a tissue of known optical properties) to predict scattering effects and recreate a more accurate image of vessel geometry at various depths.⁴³

4.3.

Quantifying Functional Effects of NO on Lymphatics In Vivo

In this study, we showed for the first time that immediate changes in lymphatic function resulting from the introduction of NO can be detected using non-invasive NIR lymphatic imaging. Our findings, that GTNO significantly reduces lymphatic transport, corroborates existing knowledge that NO has an inhibitory effect on lymphatic pump function.^29,31,35 We have shown that NIR lymphatic imaging can provide real-time in vivo measurements of lymphatic pump function in response to NO, which has never previously been available, and may help to further elucidate the relationship between NO and lymphatic contractile regulatory mechanisms. The ability to measure this response non-invasively would be particularly useful given recent findings that certain immune cells migrate to the lymphatics and release NO as a means of regulating local lymphatic draining.²⁹

The gold standard for quantifying lymphatic pump function has been to utilize diameter tracking of contracting vessels to calculate parameters such as stroke volume and ejection fraction. These temporal traces of diameter changes have been achieved in isolated vessel preparations,^31,44 invasive in vivo intravital brightfield microscopy,^33,45 and more recently through invasive intravital fluorescence microscopy using vessels filled with FITC labeled dextran.²⁹ All of these approaches require invasive surgery to access and visualize the lymphatics, thus allowing for accurate diameter tracings. While the approach reported here has the advantage of being non-invasive, the scattering effects of the dermal layer and the lower frame rates do not currently provide the necessary spatial and temporal resolution to achieve accurate diameter tracings, which explains why this and other NIR lymphatic imaging systems have been unable to quantify these more traditional metrics of pump function. Thus we sought to define quantitative metrics of pump function similar to these parameters that could be calculated from our system, namely effective contraction length and systolic pumping power.

Effective contraction length describes, on average, how far a packet of fluid would travel down the lymphatic vessels before another contraction event is needed. Stronger contractions would propel fluid further (assuming that the immediate downstream valves are open), when compared to weaker contractions with lower ejection fractions. Systolic pumping power provides an estimation of the average power generated per unit mass by lymphatic pumping. A calculation of the actual power would require knowing the mass of the fluid packet, but this is difficult to estimate, since accurate diameter measurements are difficult to achieve given the limitations of NIR imaging discussed above. It is likely that this mass would be different between treatment groups, since it is known that NO increases the vessel diameter.³¹ However, any changes that would occur in packet mass due to vessel dilation would be small ($\sim 2$-fold increase) when compared to the changes seen in the power per unit mass parameter ($\sim 50$-fold decrease). Both of the new parameters developed here demonstrate the potential to describe remarkable differences in lymphatic pump function that could be difficult to capture when tracking packet frequency or velocity alone.

Our findings also have the potential to establish NIR lymphatic imaging as an early-stage lymphatic disease diagnostic. To date, NIR imaging has been reported in the literature to be capable of identifying differences in lymphatic pump function between healthy states and several late-stage disease states.^19,22,24 However, given that most lymphatic disorders are characterized by a progressive deterioration of lymphatic pump function prior to the presentation of clinical manifestations, NIR imaging may be capable of detecting changes in lymphatic pump function in the very early stages of the disease before visible symptoms begin to present. Our findings suggest that NIR imaging is very sensitive to detecting differences in lymphatic transport function and could be used as a screening mechanism for patients at a high risk for developing lymphatic disorders, such as post-mastectomy breast cancer patients. In this way, corrective measures could be taken before irreversible tissue damage would occur, thus improving patient outcomes with lymphatic diseases.