2.1.

Animals

Adult female and male $\mathrm{IL}2\mathrm{R}\gamma$ salt-sensitive Sprague–Dawley (${\mathrm{SS}}^{\mathrm{IL}2\mathrm{R}\gamma -}$) rats were bred and maintained at our institution. Animals were anesthetized during imaging with 2% of isoflurane maintenance in an oxygen flow of $\sim 1\mathrm{L}/\min$, and temperature was maintained throughout the experiment with a heating pad. All animals were imaged in a supinated position. All animal protocols were reviewed and approved by the Medical College of Wisconsin, Institutional Animal Care and Use Committee, where these experiments were performed. All rats recovered fully after fluorescence imaging.

2.2.

Multispectral Imaging System

A combination of 785 nm (Thorlabs, L785P200, Newton, New Jersey, United States) and 808 nm (Diomed, D15 Plus, Cambridge, United Kingdom) laser light to excite the ICG and QDs, respectively, illuminated the entire animal body [Fig. 1(a)]. The imaging system comprised a 16-bit deep cooled intensifier electron-multiplying charge-coupled device (emICCD) camera (Princeton Instruments, PI-MAX4, Trenton, New Jersey, United States, $512\times 512\text{pixels}$, 400 to 900 nm sensitivity) and a 16-bit deep cooled InGaAs sensor focal plane array camera (Princeton Instruments, NIRvana 640ST, $512\times 640\text{pixels}$, 900 to 1700 nm sensitivity) in combination with a fused silica broadband beam splitter (Thorlabs, BSW30) to direct signal from the same field of view to both cameras. A combination of $785\pm 5\mathrm{nm}$ (SuperNotch-Plus™, HSPF-785.0-2.0, Kaiser Optical System, Ann Arbor, Michigan, United States) and $808\pm 20\mathrm{nm}$ (Semrock, StopLine^®, NF03-808E-50) notch filters were used sequentially, pre-beam splitter; an $830\pm 10\mathrm{nm}$ bandpass filter (830FS10-25, Andover, Salem, New Hampshire, United States) and 980 nm long pass filter (Semrock, EdgeBasic™, BLP01-980R-50, Rochester, New York, United States) were used post-beam splitter on the NIR and SWIR beam paths, respectively. For SWIR-only imaging, the NIRvana was used with the 808 nm laser and appropriate filters with no beam splitter.

Fig. 1

(a) Schematic of multispectral imaging system. (b) Capillary tube with $14\mu \mathrm{M}$ ICG excited with 785 nm with $830\pm 10\mathrm{nm}$ emission in 1% liposyn at respective depths. (c) Cross section of ICG-filled capillary tube normalized to 10 mm depth in liposyn. (d) Capillary tube with $14\mu \mathrm{M}$ ICG excited at 808 nm with 980 nm long pass emission. (e) Cross section of ICG-filled capillary tube normalized to 10 mm depth in liposyn.

2.3.

Imaging Agents

Imaging agents were administered in a single bolus of $20\mu \mathrm{L}$ intradermally into the interdigital space of the rat hindfoot. ICG was administered at a concentration of 50 to $100\mu \mathrm{M}$. QDs were synthesized via viscosity-mediated synthesis as described by Tang et al.²⁰ QD emission spectra are governed by the size of the particles and can be modulated by increasing or decreasing the synthesis reaction time. Eight-nm QDs fluoresce around 1100 nm upon 808 nm excitation. PEGylated quantum dots (QDs) are grossly different than ICG in physical characteristics and thus exhibit different pharmacokinetics. After PEGylation ($500\mathrm{g}/\mathrm{mol}$), the hydrodynamic radius of QDs was $\sim 60\mathrm{nm}$ with a zeta potential in the interval $[-5,-10]\mathrm{mV}$. QDs were administered with a concentration of $\sim 37.5\mu \mathrm{g}/\mathrm{mL}$ in equal volumes to ICG and imaged using a custom dual-camera multispectral imaging system in both NIR and SWIR. Animals imaged with ICG were also imaged with PEGylated QDs after the ICG had cleared $\sim 2$ weeks later. As a true measure of the lymphatic vessel diameter, Evan’s blue (EB) dye (5% in $25\mu \mathrm{L}$ Hank’s balanced salt solution) was injected in the interdigital foot pad of anesthetized rats following fluorescence imaging.²¹ Following euthanasia, $\sim 15\min$ after injection, the skin was dissected to reveal the dermal lymphatics, and RGB images of the EB-stained vessels on the underside of the skin were taken. To mimic the scattering properties of tissue, a 1% liposyn emulsion in PBS was used as a tissue phantom with a capillary tube serving as the vessel. The absorption spectra of tissue are relatively similar in both the NIR-I and SWIR, so no absorber was added to the liposyn emulsion.²²

2.4.

Image Processing and Derivation of Functional Parameters

Image streams were acquired using LightField software (Princeton Instruments), and image processing and data analysis were performed using MATLAB (Mathworks, 2023b) software. The MATLAB function “findpeaks” was used to identify and measure peaks and peak prominences in the vessel cross sections and identify peaks in the time series with empirically tuned parameters. For the purposes of comparison of the NIR versus SWIR imaging, vessel distinguishability was defined as the sum of the peak prominences in normalized, background-subtracted images from the same vessels in each camera frame divided by the number of vessels per centimeter of the line drawn, e.g.,

To simplify the task of finding peaks corresponding to lymph transport in the time series, images were “$Z$-normalized.” The $Z$-normalization process removes the bias in identifying peaks and calculating transport velocity. Image processing for “$Z$-normalization” in MATLAB includes (1) segment image stack into pixels of interest i.e., mask based on hand-drawn regions of interest (ROI) or pixel intensity. (2) Convert 3D stack to 2D array in pixel X time format and $Z$-normalize each row by subtracting the row mean and dividing by standard deviation. (3) Wavelet denoise each row (pixel time series) using the “wdenoise” function with eighth-level symlet wavelets, empirical Bayesian denoising method with median thresholding, and a level independent noise estimation. (4) Subtract a wavelet reconstruction (using “waverec” with the wavelet scaling factors from the denoising output) from the denoised row and square the difference to remove negative values. (5) Convert 2D matrix back to 3D image stack. To compute lymph bolus transport velocity after $Z$-normalization, the time difference between a bolus (as indicated by peaks in a time series from an ROI on a vessel) at two points on the same vessel was calculated, and the distance the bolus traveled along the vessel between those points was measured. The EB vessel width was measured using ImageJ.

3.1.

Validation of Multispectral Imaging System

Comparison of NIR-I and SWIR imaging of ICG phantoms in a tissue-like scattering medium (1% liposyn emulsion) showed a relative increase in sensitivity and spatial resolution owing to lower scattering in the SWIR imaging regime [Figs. 1(b)–1(e)]. These differences in the visualization of the capillary fluorescence phantom were particularly evident at the 10 mm depth.

3.2.

Determination of Lymphatic Vessel Cross Section Using NIR-I and SWIR Imaging

We observed quantitative differences between NIR and SWIR ICG fluorescence images in the lymphatics following intradermal footpad injection (Table 1). As was expected from the phantom experiments, SWIR imaging showed superior spatial resolution in vivo of both small and large vessels using ICG as contrast [Figs. 2(a)–2(c)]. SWIR imaging with ICG provided up to 1.7 times the vessel distinguishability and up to 1.7 times the resolution. In addition, there was less excitation contamination in SWIR emission images due to the wide spectral separation in excitation and emission spectra. As the InGaAs sensor is not sensitive to 785 or 808 nm wavelengths, it provides an additional layer of long-pass filtering.

Table 1

Data from Fig. 2(c) showing sensitivity and resolution. Averages with standard deviation are reported for plots with more than one peak.

Fig. 2

(a) Hindlimb and lower abdominal lymphatics in a supinated rat after intradermal ICG injection in the footpad with $830\pm 10\mathrm{nm}$ emission with magnified inset. (b) Hindlimb lymphatics in the same rat at 808 nm with 980 nm long pass emission with magnified inset. (c) Maximum normalized vessel cross sections in respective spectral bands (Video 1, mp4, 2.09 MB [URL: https://doi.org/10.1117/1.JBO.29.10.106001.s1]).

3.3.

Determination of Transport Velocity and Estimating Flow

Distinct peaks indicating each bolus of lymph (Video 1) transiting a region of interest drawn on a lymphatic vessel can be observed following $Z$-normalization and wavelet denoising. With two ROIs drawn on the length of a vessel, the peaks in the time series—corresponding to boluses of transiting lymph—at each vessel segment can be used to determine the time it took for the bolus to transit the length of the vessel and subsequently the transport velocity [Figs. 3(a)–3(c)]. QDs and ICG both showed similar transport velocities of $6.39\pm 2.40$ and $7.06\pm 1.58\mathrm{mm}/\mathrm{s}$, respectively ($n=4$ rats each), in the vessels visualized on the abdominal region. EB injections in the footpad allowed for hindlimb/abdominal subdermal lymphatics to be visualized upon dissection of the skin. The true physical size could then be measured and compared with the diameter of the vessels measured from fluorescence data using the same techniques as Fig. 2. The average diameter of vessels measured via EB was $0.22\pm 0.04\mathrm{mm}$ ($n=3$ rats), whereas the same vessel measured via SWIR fluorescence was $1.10\pm 0.38\mathrm{mm}$ ($n=3$ rats). Thus, the SWIR vessel cross section is five times greater due to multiple scattering of light through the skin than the true physical vessel measured by EB after dissection.

Fig. 3

(a) Unnormalized and $Z$-normalized time series from a circular ROI drawn on a lymphatic vessel. (b) ROIs (left) and the line trace on the vessel (right) for computing lymph bolus transport velocity. (c) $Z$-normalized peaks for ROIS 1 (blue) and 2 (orange) in panel (b) used to measure time in transport velocity. (d) Subdermal lymphatics visualization following Evan’s blue injection used for true vessel diameter.

3.4.

Comparison of QD and ICG

ICG and QDs were injected in the same vessels of the same animals 13 days apart to allow for time for clearance of previous contrast administration. These images indicated that ICG-tagged vessels had a higher absolute intensity than the QDs but had an inferior signal-to-background ratio (Table 2) due to a higher background even after both images underwent pre-contrast background subtraction. The emission peak for ${\mathrm{Ag}}_2\mathrm{S}$ QDs is red-shifted compared with ICG and has a tighter emission window (Fig. S1 in the Supplementary Material). This should result in sharper resolution of QD-SWIR imaging compared with SWIR imaging with ICG which was borne out by the results in Figs. 4(a)–4(c) for which both ICG and QD emissions were imaged with a 980 nm long-pass filter. Although the purely SWIR-emitting QDs show sharper spatial resolution due to reduced scattering, they are not as bright as ICG in the lymphatic vessels. Conjugation of thiol-terminated PEG to the carboxylic acid groups on the surface of the QDs created a more neutral zeta potential of approximately $-7\mathrm{mV}$ from the $-20\mathrm{mV}$ of the acid coating, which likely resulted in a faster clearance from lymphatics compared with ICG; consequently smaller lymphatics can show up as brighter for ICG imaging, as shown in Figs. 4(a) and 4(b).²³ SWIR imaging with QDs provided up to 2.5 times the vessel distinguishability, dependent on the vessel spacing, and up to 1.7 times the resolution (Table 2). This could be due to the shorter SWIR emission wavelength of ICG—leading to higher scattering. Co-registration of NIR (830 nm BP) and SWIR (980 nm LP) images with concurrent ICG and ${\mathrm{Ag}}_2\mathrm{S}$ QDs injections in opposite limbs shows that higher spatial resolution and lower background are possible with ${\mathrm{Ag}}_2\mathrm{S}$ QDs [Fig. 4(d)]. A peak between 1100 and 1200 nm is expected with ${\mathrm{Ag}}_2\mathrm{S}$ QDs of this size [Fig. 4(e)]. The rats injected with QDs did not suffer any apparent toxicity and cleared the residual QDs from the injection sites after approximately a week’s time.

Table 2

Data from Fig. 4(c) showing sensitivity and resolution. Averages with standard deviation are reported for plots with more than one peak. The resolution ratio was not calculated for dissimilar numbers of peaks.

Fig. 4

(a) Hindlimb and lower abdominal lymphatics after intradermal ICG injection in footpad (white arrow) with 980 nm long pass emission with magnified inset. (b) Hindlimb and lower abdominal lymphatics after intradermal QD injection in the footpad with 980 nm long pass emission with magnified inset. (c) Normalized vessel cross sections with respective SWIR contrast agents. (d) Co-registered images of simultaneous SWIR and NIR lymphatic images with NIR in purple and SWIR in green with overlapping signals appearing in white. QDs were administered to the supinated rat’s right hindlimb and ICG to the left. (e) Emission spectra of QDs.