2.1.

3.2F NIRF–IVUS Imaging System

The hybrid NIRF–IVUS imaging system consists of a control and readout system, a helical pullback module, and a 3.2F NIRF-IVUS imaging catheter [Fig. 1(a)]. An ultrasound pulser-receiver (5073PR, Panametrics, Waltham) was used to excite and detect IVUS signals. A 750-nm fiber-coupled laser (B&W Tek, Newark, Delaware) was implemented to excite NIRF signals, which were detected by a photomultiplier tube (Model H7422-50, Hamamatsu, Japan). The pullback module includes a fiber-optic rotary joint combined with an electrical slip ring (Alpha Slip Rings, Austin, Texas), a rotary motor (3268G042BX4AES-4096, Faulhaber, Germany), and a linear stage (X-LRQ stage, Zaber, Vancouver, BC, Canada).

Fig. 1

Imaging setup and description of experiments. (a) Illustration of NIRF–IVUS imaging setup including an intravascular 3.2F NIRF–IVUS catheter. (b) Cross-sectional image visualizing the dimensions of the NIRF–IVUS catheter inserted into a size 3.2F sheath. (c) Image of a commercial intravascular guidewire coated with epoxy resin and ICG. (d) Schematic of the calibration phantom used to obtain NIRF–IVUS measurements of the fluorophore-coated guidewire at variable distances (sheath-guidewire distances, $d_{{\mathrm{GW}}_{\text{water}}}$) in water to provide a correction factor for NIRF signal loss in water [$C_{\mathrm{w}}$, Eq. (1)] and construct a look-up-table containing NIRF reference signals. (e) Schematic of an intravascular measurement of an ICG-injected artery perfused with blood, with the adjacent fluorophore-coated guidewire providing reference signals for the calculation of frame-by-frame correction factors for NIRF attenuation in blood [$C_{\mathrm{b}}$, Eq. (4)]. (f) Cross-sectional schematic of an intravascular measurement showing the distances between the catheter and the sheath ($d_{{\mathrm{s}}_{\mathrm{GW}}}$ and $d_{{\mathrm{s}}_{\mathrm{T}}}$) used to correct for signal loss in water within the sheath, see Eqs. (2) and (3), the sheath and the guidewire ($d_{{\mathrm{GW}}_{\text{blood}}}$), and the sheath and the tissue ($d_{{\mathrm{T}}_{\mathrm{blood}}}$) containing ICG. (g) Image of an ICG-injected porcine coronary artery indicating ROI for imaging and correction analysis. (h) Schematic of the capillary phantom. The phantom contains three different target capillaries (T1-3) and a reference (R) capillary that simulates the fluorophore-coated guidewire, each filled with solutions of ICG at different concentrations. The target capillaries are positioned at variable distances ($d_{\mathrm{T}1-3}$) resulting in a range ($\mathrm{\Delta }{d}_{T1-3}$), whereas the reference capillary was positioned at a fixed distance ($d_{\mathrm{R}}$) from the catheter sheath. All capillaries are exposed to signal loss in water ($C_{\mathrm{w}}$) inside the sheath ($d_{{\mathrm{S}}_{\mathrm{T}1-3}}$, $d_{{\mathrm{S}}_{\mathrm{R}}}$) and attenuation due to blood ($C_{\mathrm{b}}$) outside the sheath. In all panels: NIRF, near-infrared fluorescence; IVUS, intravascular ultrasound; ICG, indocyanine green; GW, guidewire; and ROI, region of interest.

The miniaturized NIRF-IVUS catheter previously developed by our group¹¹ was used to implement all the measurements described herein. In short, the catheter comprises a ball-lens attached to a $50/105\text{-}\mu \mathrm{m}$ fiber (Thorlabs, Newton, New Jersey) and a piezoelectric element (size: $0.4\times 0.6\mathrm{mm}$; center frequency: 40 MHz; Blatek, Boalsburg, Pennsylvania) for NIRF-IVUS signal excitation and detection. Both sensors are implemented into a ferrule in a serial design featuring a total outer diameter of 0.55 mm. The NIRF–IVUS catheter is introduced into size 3.2F catheter sheath [Fig. 1(b)] made from low-density polyethylene and combined with a custom-made flushing port that utilizes a clinical introducer sheath (RT-R50G10PQ, Terumo Europe) coupled to the catheter sheath via three-dimensional (3D) printed flexible connector. The sheath was filled with distilled water to ensure ultrasound coupling into the tissue and prevent blood contact with the NIRF sensor. The distance between the catheter and the sheath wall must be accounted for when correcting fluorescence intensities [see Fig. 1(f) and Eqs. (2) and (3)].

All data presented herein were acquired with 60-rpm rotation speeds and $0.25\text{-}\mathrm{mm}/\mathrm{s}$ pullback speeds for a distance of 20 mm (i.e., 80 cross-sectional NIRF/IVUS imaging frames), resulting into a helical scan of the intimal arterial wall. A custom-made LabVIEW (National Instruments Corp., Austin, Texas) software allows for system control and data acquisition, whereas all data analysis was implemented in MATLAB (Mathworks, Natick, Massachusetts).

2.2.

Guidewire Coating

A standard clinical guidewire with a 0.35-mm diameter (${\mathrm{PT}}^2$ Moderate Support, Boston Scientific Corporation, Marlborough, Massachusetts) was immersed in a $50\text{-}\mu \mathrm{M}$ solution of ICG in epoxy resin and cured at room temperature. The method has been optimized to afford a thin and homogeneous fluorescent coating with an average thickness of $45\mu \mathrm{m}$ [Fig. 1(c)] and a high NIRF signal-to-noise ratio (SNR) based on an ICG concentration, which was maximized until quenching occurs above $50\mu \mathrm{M}$.²²

2.3.

Calibration Measurements

The method consists of a calibration step [Fig. 1(d)], which is performed once for every guidewire prior to tissue measurements [Fig. 1(e)]. For this calibration step, the fluorophore-coated guidewire was measured through water in a phantom with varying distances (0.1 to 1. mm) between the guidewire and the catheter sheath wall to (i) calculate a correction factor for water [$C_{\mathrm{w}}$, Eq. (1)] and (ii) construct a look-up-table containing NIRF signals for variable distances. The water correction factor was used to account for NIRF signal loss between the NIRF sensor and the sheath wall [Fig. 1(f), Eqs. (2) and (3)] and between the sheath wall and tissue to calculate ground truth values [Secs. 2.4–2.5, Eq. (6)]. The data from the look-up-table served as the reference values to calculate a correction factor for blood attenuation [$C_b$, Eq. (4)]. The calibration phantom comprised a 3D-printed reservoir $26\times 21\times 18\mathrm{mm}$ ($L\times W\times H$) in size with designated placeholders that allowed the guidewire to be angled relative to the imaging catheter.

2.4.

Ex Vivo Tissue Measurements

A fresh pig heart was collected from a butcher shop and a 40-mm segment of the right coronary artery with an inner diameter of $\sim 2.8\mathrm{mm}$ was excised and positioned into a designated perfusion holder [Fig. 1(g)]. An aqueous solution of ICG ($38\mu \mathrm{M}$) was injected into the media layer of the artery to create a localized target fluorescence signal in the tissue. The fluorophore-coated guidewire (described in Sec. 2.2) was inserted into the vessel lumen first, followed by the imaging catheter. By inserting the guidewire into an opening at the distal end of the catheter sheath, the imaging catheter was guided to its starting position for imaging next to the guidewire, similar to the configuration in standard clinical intervention procedures [see Fig. 1(e)]. Figure 1(g) shows the region-of-interest (ROI) visualizing the location of fluorophore injection and the segment for final data analysis. We performed two NIRF–IVUS pullbacks of the artery in a water bath while perfusing it once with water (as a control) and once with a solution of 75% blood diluted with an isotonic solution of aqueous 0.9% NaCl [Fig. 2(a)]. Finally, the attenuated NIRF maps were corrected frame-by-frame using the NIRF reference signals acquired from the guidewire as described in Sec. 2.5. The ICG concentrations obtained after correction of blood attenuation were then compared with the ground truth values calculated from the control measurements through water [Sec. 2.5, Eq. (6)] for each angular position in all frames to evaluate the quantification accuracy of the method. We also performed a sensitivity analysis to investigate the influence of IVUS distance measurements on correction accuracy. For this, we altered the distances between the sheath wall and the guidewire ($d_{{\mathrm{GW}}_{\text{blood}}}$) and tissue $(d_{{\mathrm{T}}_{\text{blood}}})$ in $\pm 50\mu \mathrm{m}$ steps from the values acquired in a representative frame and angular location and calculated quantification accuracy for each deviation as described above.

Fig. 2

Validation of the correction method in an excised pig artery. (a) NIRF–IVUS measurements of a porcine coronary artery placed in a water bath and perfused with blood. (b) Acquired uncorrected intravascular NIRF maps of an ICG-injected porcine artery measured through water and a 75% blood solution, showing NIRF signals from both the guidewire (GW) and tissue. (c) Distance between the fluorophore-coated guidewire and the catheter sheath measured during blood perfused tissue measurements ($d_{{\mathrm{GW}}_{\text{blood}}}$) per pullback position (frame). (d) Exemplary IVUS imaging frame in polar coordinates showing distances from the sensor to the catheter sheath, the fluorophore-coated guidewire (GW), and the artery wall tissue. (e) The same intravascular NIRF maps as in (b) after correction showing quantified ICG concentrations. (f) Selected frames showing uncorrected intensity values and corresponding corrected ICG concentrations for the GW and tissue signals through blood and water. (g) Sensitivity analysis investigating the influence of simulated errors in sheath to guidewire ($d_{{\mathrm{GW}}_{\text{blood}}}$) and tissue ($d_{{\mathrm{T}}_{\text{blood}}}$) distance measurements on correction accuracy acquired during blood perfused tissue measurements; (h) NIRF measurements of the fluorophore-coated guidewire imaged through water and 50%, 65%, and 75% blood mixtures with corresponding one-term exponential model (averaged NIRF intensities shown as points and fits as solid lines; $R^2$ for all fits = 0.99). In all panels: ICG, indocyanine green; GW, guidewire; Acq, acquired; Cor, corrected.

2.5.

Fluorescence Signal Correction

The same fluorophore (ICG) was used to coat the guidewire and injected into the tissue. The fluorophore-coated guidewire was measured concurrently with the vessel wall to enable readouts that accurately depict the true distribution of ICG in the tissue.^23,24 Light attenuation by blood in intravascular fluorescence imaging was modeled as an exponential function of the distance between the outer wall of catheter sheath and the vessel wall or fluorophore-coated guidewire.^4,8–11 The NIRF signals of the guidewire were segmented from tissue signals using the coregistered IVUS data.

Correction factors for blood attenuation were calculated for each frame (see details below) by comparing NIRF signals of the guidewire acquired during the tissue measurements through blood to the reference values obtained during the calibration measurements through water. Fluorophore concentrations in the tissue were calculated using these blood correction factors, and the known distances between the catheter and the artery wall (manually delineated in coregistered IVUS frames). In all calculations, NIRF signal loss in the water-filled gap between the NIRF sensor and the catheter sheath is accounted for. The steps of our correction method for a typical measurement are described in detail below (S1-S5):

2.6.

Validation of the Method with a Capillary Phantom Experiment

The accuracy and robustness of the fluorescence signal correction for variable imaging conditions using a guidewire-like reference signal was assessed using two phantoms: a capillary-calibration (calibration measurements) and a target phantom [Fig. 1(h)]. The capillary-calibration phantom replicated the calibration measurements in Sec. 2.3 and contained only a single capillary filled with $27\mu \mathrm{M}$ ICG dissolved in DMSO, which was angled to create imaging distances ranging from 0.45 to 1.6 mm [similar like Fig. 1(d)]. The target phantom-simulated NIRF signals at different concentrations and distances using three glass capillaries ($25\mu \mathrm{l}$ per capillary), which were filled with 6.8 (T1), 13 (T3), and 27 (T2) $\mu \mathrm{M}$ ICG dissolved in DMSO and angled within the phantom to create variable distances ($d_{T1-3}$) between the wall of the catheter sheath and the targets, ranging from 0.55 to 1 mm [$\mathrm{\Delta }{d}_{\mathrm{T}1-3}$, Fig. 1(h)]. A fourth capillary was filled with the same ICG concentration as in the capillary-calibration phantom ($27\mu \mathrm{M}$), fixed at $\sim 0.45\mathrm{mm}$ distance ($d_R$), and was used to acquire a frame-by-frame reference ($R$) signal. NIRF intensities in both the target and capillary-calibration phantoms were recorded via helical pullbacks similar to the tissue measurements. Recorded NIRF intensities were dependent on the distance between the sheath wall and the capillaries, as well as the distance between the sheath wall and the NIRF detector [$d_{{\mathrm{S}}_{\mathrm{T}1-3}}$ and $d_{{\mathrm{S}}_{\mathrm{R}}}$, see Fig. 1(h)] as mentioned in Sec. 2.5. While the capillary-calibration phantom was imaged through water, the target phantom was imaged through water (as a control) and solutions of blood diluted to 75% and 50% with an isotonic solution of aqueous 0.9% NaCl. The data from the capillary-calibration measurement were fit with an exponential function to model NIRF signal loss in water. Finally, we corrected the blood-attenuated intensity of the target capillaries within the target phantom using the reference signal as described in Sec. 2.5 and compared the agreement of the corrected intensities with the ground truth values calculated from the control measurements through water [similar like in Sec. 2.5, Eq. (6)] to estimate the correction accuracy. Furthermore, we correlated corrected NIRF intensities to ICG concentrations to estimate the accuracy of fluorophore quantification.

3.1.

Correcting Signal Attenuation by Blood Using a Fluorophore-Coated Guidewire as a Reference

Figure 2 shows NIRF–IVUS measurements of a fluorophore-coated guidewire inside of an ICG-injected ex vivo pig coronary artery (see Sec. 2.4) that was perfused alternately with blood [Fig. 2(a)] and water (as a control), recorded to assess the feasibility of correcting NIRF signals for blood attenuation on a frame-by-frame basis using the guidewire signal as a reference. Figure 2(b) shows acquired uncorrected NIRF maps of imaging pullbacks through water and 75%-blood dilution. As expected, the recorded NIRF intensities were significantly lower in the presence of blood due to increased attenuation when compared with the measurements in water. Furthermore, the fluorescent signals from the guidewire showed variable intensities throughout the pullbacks due to changing distances from the sheath [Fig. 2(c)], as quantified by IVUS [Fig. 2(d)]. The intensity distribution of the signal from the ICG in the tissue is affected by both postinjection fluorophore diffusion and variations in sheath–tissue distance. Figure 2(e) shows the corresponding NIRF maps after correction of the intensities [see Secs. 2.3–2.5 and Figs. 1(d)–1(f)] to afford ICG concentrations throughout the pullback (see colorbar). Figure 2(f) shows acquired uncorrected and corrected concentration profiles of the guidewire and the tissue from four exemplary frames [Figs. 2(b), 2(e); F28, F33, F37, and F40], to more closely evaluate quantification accuracy. The quantified concentrations based on the corrected water measurements served as the ground truth [see Sec. 2.5, Eq. (6)]. The estimated ICG concentrations in blood corresponded well to these ground truth measurements in water for both the signals from the guidewire and the tissue. After correction, the calculated ICG concentrations in the tissue measured through blood were 89% (variance = 9.6%) accurate on average compared with the measurements through water, which yielded an improvement of 2.2-fold compared with uncorrected measurements. The correction accuracy of our method was dependent on the IVUS distance measurements [see Eqs. (4) and (5) and Fig. 2(g)]. Our sensitivity analysis indicated that our method achieved an average correction accuracy of 70% for simulated errors in IVUS distance measurements between the sheath wall and the guidewire ($d_{{\mathrm{GW}}_{\text{blood}}}$) and tissue ($d_{{\mathrm{T}}_{\text{blood}}}$) of up to $\pm 100\mu \mathrm{m}$.

3.2.

Evaluating Attenuation Correction Efficacy Versus Distance and Fluorophore Concentration in a Capillary Phantom

Pullback measurements of the capillary-calibration phantom afforded dozens of data points (highest five intensity values averaged per frame) at distances of 0.5 to 1.6 mm from the sheath wall. These data points were fit with an exponential function to retrieve a correction factor in water [Fig. 3(a); $R^2$ for fit, 0.99]. The fitted exponential functions indicate increasing attenuation with increasing distance. The average correction factor for water measurements ($C_{\mathrm{w}}$) was estimated to be $0.0009{\mathrm{mm}}^{-1}$.

Fig. 3

Validation of correction method for intravascular NIRF with capillary phantoms. (a) Measurements of the capillary-calibration phantom in water and both blood dilutions (for validation) showing NIRF intensities (averaged per frame) versus distance from the NIRF sensor to the reference capillary, with corresponding model fits and values for $C_{\mathrm{w}}$ and $C_{\mathrm{b}}$ (data points shown as dots and fits as solid lines). (b) NIRF intensities (averaged per frame) of the reference capillary recorded during target measurements in 50% blood (RB50) and 75% blood (RB75) at distances from the sheath wall to the reference capillary ranging between 0.43 and 0.53 mm. (c) Calculated correction factor ($C_{\mathrm{b}}$) for the 50% and 75% blood mixtures versus distance from the sheath wall to the reference capillary for each frame. Acquired uncorrected NIRF maps of reference and target capillaries measured through (d) water, (e) 50% blood, and (f) 75% blood. Corrected NIRF maps of reference and target capillaries measured through (g) water, (h) 50% blood, and (i) 75% blood. (j) Acquired uncorrected NIRF intensities (averaged per frame) from the target capillaries measured in the target phantom through water (T1-3W) and 75%-blood mixture (T1-3B75) plotted versus imaging distance between the sheath wall and capillary. (k) Corrected NIRF intensities ($I_{\mathrm{T}**}$, averaged per frame) from the target capillaries measured in the target phantom through water (T1-3W) and 75%-blood mixture (T1-3B75) plotted versus imaging distance between the sheath wall and capillary. (l) Quantification of ICG concentration via corrected NIRF intensities ($I_{\mathrm{T}**}$, averaged over pullback) of the three different target capillaries measured through water, 50% blood, and 75% blood. RW, Reference capillary measured in water; RB50, reference capillary measured in 50% blood mixture; RB75, reference capillary measured in 75% blood mixture; $C_{\mathrm{w}}$, correction factor for water; $C_{\mathrm{b}50\%}$, correction factor for the 50% blood mixture; ${\mathrm{C}}_{\mathrm{b}75\%}$, correction factor for the 75% blood mixture; Acq, acquired; Cor, corrected; T1-3W, target capillaries 1 to 3 measured in water; T1-3B50, target capillaries 1 to 3 measured in 50% blood mixture; T1-3B75, target capillaries 1 to 3 measured in 75% blood mixture; ICG, indocyanine green.

Figures 3(d)–3(f) show acquired uncorrected NIRF maps of the reference capillary (R-) and all three target capillaries (T1-3-) acquired throughout the pullback of the target phantom when measured through water (RW, T1-3W), the 50% blood dilution (RB50, T1-3B50), and the 75% blood dilution (RB75, T1-3B75). As expected, the fluorescence intensities decreased with increasing blood content and the concomitant increase in light attenuation.

Subsequently, averaged intensity values of the reference capillary (substituting for the guidewire) recorded during target measurements at distances ranging between 0.43 and 0.53 mm in blood [RB50 and RB75, Figs. 3(b), 3(d)–3(f)] were used to quantify $C_{\mathrm{b}}$ [see Sec. 2.5]. The average correction factors for the blood measurements were 0.0028 and $0.0034{\mathrm{mm}}^{-1}$ for the 50%-blood ($C_{\mathrm{b}50\%}$) and 75%-blood ($C_{\mathrm{b}75\%}$) dilutions, respectively [Fig. 3(c)]. These calculated values agree well (98% accuracy) with extracted correction factors ($C_{\mathrm{b}50\%}=0.0028{\mathrm{mm}}^{-1}$; $C_{\mathrm{b}75\%}=0.0033{\mathrm{mm}}^{-1}$) from a model fit ($R^2$ for fits, 0.99) applied to validation measurements of a capillary, containing the same ICG concentration as the reference, imaged in the capillary-calibration phantom at distances of 0.5 to 1.6 mm from the sheath wall through 50%-blood and 75%-blood dilutions [Fig. 3(a)]. The variance of the calculated blood correction factors over all imaging frames in the target phantom was found to be 4.8% and 2.8% for calculated $C_{\mathrm{b}50\%}$ and $C_{\mathrm{b}75\%}$, respectively.

Finally, we evaluated the correction accuracy of our method by applying the calculated blood correction factors ($C_{\mathrm{b}50\%}$, and $C_{\mathrm{b}75\%}$) to correct the NIRF signal intensities from the target capillaries, which were compared with the ground truth values measured in water. Figures 3(g)–3(i) show corrected NIRF maps of the reference capillary, and three target capillaries measured through water, 50%-blood, and 75%-blood. Because the target capillaries were angled within the phantom (see Sec. 2.6), their uncorrected NIRF intensities were acquired at distances ranging from 0.55 to 1 mm from the sheath wall to the capillary, resulting in average variances of 13% in water, 32% in 50%-blood, and 39% in 75%-blood. Furthermore, the acquired uncorrected NIRF intensities of target capillaries containing the same ICG concentration measured through 50%-blood [Fig. 3(e)] and 75%-blood [Figs. 3(f) and 3(j) did not agree with measurements in water [Figs. 3(d) and 3(j)] due to varying NIRF attenuation. In contrast, corrected NIRF intensities of all target capillaries through both 50%-blood [Fig. 3(h)] and 75%-blood [Figs. 3(i), 3(k)] show good agreement with the corrected water measurements [Figs. 3(g), 3(k)] over all acquired frames and imaging distances. In addition, we observed that the average variances of the corrected intensities over all imaging frames decreased by 10% for the water measurements, 23% for the 50%-blood measurements, and 31% for the 75%-blood measurements. Finally, Fig. 3(l) compares quantified ICG concentrations converted from corrected NIRF intensities of the blood measurements of the three target capillaries with the ground truth, which shows that our method achieved an average accuracy of 91% (average variance = 4.3%), which is a 4.5-fold improvement over uncorrected measurements.

Furthermore, we sought to quantify the impact of correcting for sensor-to-sheath distance by calculating the difference in intensity-quantification accuracy if signal loss inside of the sheath is neglected. For this analysis, we used the measurements through the 75% blood mixture. Due to natural bending of the probe, the distance between the sensor and the wall of the sheath varies at the angular locations of the different target capillaries ($d_{{\mathrm{S}}_{\mathrm{T}1-3}}$), as seen both in an exemplary IVUS frame [Fig. 4(a)] and quantified distances from an entire pullback [Fig. 4(b)]. Figure 4(c) shows corrected NIRF intensities of the target capillaries measured through blood with and without additional correction for sensor-to-sheath distance. Sheath-corrected intensities agree better with the ground truth values than the uncorrected intensities. Thus, the accuracy in ICG quantification improved by 10% for T1, 5% for T3, and 19% for T2, which provides evidence that the additional sheath correction becomes more relevant with increasing sensor-to-sheath distances [$d_{{\mathrm{S}}_{\mathrm{T}3}}<d_{{\mathrm{S}}_{\mathrm{T}1}}<d_{{\mathrm{S}}_{\mathrm{T}2}}$, see Fig. 4(b)]. Moreover, the variances of the blood-corrected intensities decreased or were unchanged when additional sheath-correction was applied ($-2\%$ for T1, 0% for T2, and $-5\%$ for T3).

Fig. 4

Considering catheter location inside of the sheath improves overall accuracy in attenuation correction as demonstrated with measurements through 75% blood mixture in capillary phantom. (a) Representative IVUS imaging frame showing the three target capillaries, the reference capillary, and the sheath. (b) Sensor–sheath distance ($d_{\mathrm{s}}$) recorded by IVUS at the angular location of the three target capillaries over pullback distance. (c) Corrected intensities of the blood measurements with and without additional sheath correction and corrected intensities measured through water as the ground truth. Averaged ($n=5$ samples) intensities are shown as lines and standard deviation is shown as shaded areas. T1-3, target capillary 1-3; R, reference capillary.