High-Speed Intravascular Near-infrared Fluorescence-Ultrasound Imaging In Vivo

Abstract

Objective:

The combination of near-infrared fluorescence (NIRF) with intravascular ultrasound (IVUS) has shown promising applications for imaging atherosclerosis in ex vivo human arteries and in vivo animal models. However, long acquisition times, rotational distortion causing inconsistent image quality and poor catheter durability have hampered clinical translation. Technical limitations have included motor drive unit (MDU) instability, and catheter designs with a single-layer drive shaft and long rigid length of the distal tip.

Methods:

Herein, we present an improved, next-generation NIRF-IVUS system by integrating an aluminum v-block-based high-speed MDU and 3.0 French (Ø=1.0mm) catheter with a dual-layer drive shaft and reduced rigid tip length. We show a sixfold increase in imaging speed (30 FPS, 6 mm/s pullback) mirroring speed capabilities of standalone, commercial IVUS imaging.

Results:

In phantoms, we find that key NIRF-IVUS specifications like co-registration, rotational stability and NIRF resolution/sensitivity are preserved. Furthermore, we demonstrate that NIRF-IVUS molecular imaging of cathepsin protease activity can highlight stent-induced arterial inflammation independent of imaging speed, in rabbits in vivo. We calculate similar (not significant, p=0.65) NIRF target-to-background ratios (TBRs) in stented tissue areas at low-speed (1.88) and high-speed (2.00) imaging. Finally, in vivo NIRF-IVUS imaging of FDA-approved indocyanine green detects early-stage plaques in rabbit aorta not visible on standalone IVUS. Similar NIRF TBRs are calculated in low-speed (4.13) and high-speed (4.08) pullbacks.

Conclusion and Significance:

Our study demonstrates that NIRF-IVUS can highlight key pathobiological markers of atherosclerosis beyond standalone IVUS at clinical imaging speeds, further supporting the clinical translation of the NIRF-IVUS technology.

I. Introduction

INTRAVASCULAR imaging has been established to assess coronary plaque burden and composition [1, 2] and to guide percutaneous coronary intervention (PCI) [3–7]. Clinically available intravascular imaging technologies include intravascular ultrasound (IVUS) or optical coherence tomography (IVOCT) which can detect structural features of coronary artery disease (CAD) like plaque burden, with IVUS, or cap thickness, with IVOCT, and they are regularly used to guide stent placement and optimize stent expansion. However, IVUS and IVOCT have limited capabilities to identify biological aspects of vulnerable plaques or PCI complications because they lack molecular contrast to reliably identify key pathobiological characteristics of atherosclerosis like accumulated lipids, inflammatory burden or vascular injury [8–10]. Intravascular near-infrared fluorescence (NIRF) molecular imaging can detect injectable NIRF agents which can target pathobiological markers of atherosclerosis not visible on standalone IVUS or IVOCT and has shown promising results for imaging atherosclerosis in ex vivo human specimens [11, 12] and living animals [13–20]. The combination of NIRF with IVUS into a hybrid imaging catheter enables integration of standard structural imaging (IVUS) with molecular (NIRF) information of the vascular wall, allowing improved in vivo assessment of CAD without the need for blood clearance [16, 19, 21]. However, previous designs of hybrid catheters included single-layer drive shafts and long rigid lengths of housings (>3 mm) which resulted in slower imaging speeds (≤5 FPS rotational and 1 mm/s pullback) and non-uniform rotation distortion (NURD) [12, 16, 19, 21]. As a result, longer acquisition times, lack of real-time image display and the presence of imaging artifacts limited in vivo image quality in coronary arteries of a beating heart in large animal models [16, 19]. In addition, mechanical instability and an over-exposed optic can lead to increased wear during pullbacks and hence, breakage of fragile components like fiber-ball lens assemblies [16, 21]. Hybrid intravascular photoacoustic-ultrasound (IVPA-US) and near-infrared spectroscopy (NIRS)-IVUS catheters share a similar sensor architecture with NIRF-IVUS, using a serial arrangement of a multimode fiber optic and IVUS transducer. IVPA-US catheters have achieved high imaging speeds (up to 100 FPS, 5 mm/s) in preclinical models [22, 23], but their use of large drive shaft diameters (≥0.9 mm) and rigid lengths (≥5 mm) may limit their translatability for imaging human coronary arteries in patients. Commercial NIRS-IVUS catheters (InfraRedx, Bedford, MA, USA) operate at real-time rotational speed (30 FPS) with a modest size increase (Ø_DS=0.69 mm; rigid length=2 mm) but operate at a maximum of 2 mm/s pullback speed for automated pullbacks. To overcome limitations of previous NIRF-IVUS imaging systems [12, 16, 19, 21], we developed a v-block-based motor drive unit (MDU) optimized for high-speed operation and translated established design features of standalone, commercially available, mechanical IVUS catheters – dual-layer drive shaft and reduced rigid length – into our updated NIRF-IVUS catheter for improved imaging. We integrated a 3D-printed alignment ring into our catheter housing to implement design changes while matching the size of a standalone, commercial IVUS catheter. Our high-speed NIRF-IVUS imaging system achieved robust operation at commercial standalone-IVUS rotational and pullback imaging speeds, enabling a sixfold increase over previous designs. The NIRF-IVUS system was benchmarked at low rotational and pullback imaging speeds (5 FPS, 1 mm/s) and high rotational and pullback imaging speeds (30 FPS, 6 mm/s) in phantoms and in vivo rabbit models of atherosclerosis to provide simultaneous molecular-structural imaging.

II. Material and methods

A. NIRF-IVUS Imaging System and Catheter

The optical/electrical system is similar to one previously described [16, 21]. In brief, it comprises a 750 nm NIRF excitation fiber-coupled laser (B&W Tek, Newark, DE, USA), a photomultiplier tube (Model H7422–50, Hamamatsu, Japan) for NIRF detection, an IVUS Pulser/Receiver (Model 5073PR, Panametrics, Waltham, MA, USA), a PC for data acquisition and system control. The MDU is optimized for high-speed scans with the imaging catheter. All hardware components were integrated on a cart to mobilize the system (Fig. 1a). The MDU consists of a rotational motor (Model 2657W012C-R, Faulhaber, Germany); a linear stage (X-LRQ stage, Zaber, Vancouver, BC, Canada) for catheter-based helical pullback scans; and a hybrid fiber-optical rotary joint (FORJ, Princetel, Hamilton Township, NJ, USA) and slip ring (Senring, Shenzhen, China). A custom-made aluminum v-block was integrated into the MDU for an optimized design with reduced interfaces and precise alignment of the hybrid FORJ-Slip ring which increased stability and enabled rotational and longitudinal pullback speeds up to 100 Hz and 10 mm/s, respectively (Fig. 1b). In addition, the design of the MDU was compact and mobile for improved usability in the surgery room. For the first time, the drive shaft of our new NIRF-IVUS imaging catheter consists of two layered coils for increased stability and durability during rotation, instead of a single layer featured in previous prototypes (Fig. 1c) [12, 16, 19, 21]. In addition, to increase catheter flexibility, the rigid length of our catheter distal tip is reduced by 33% (from 3 to 2 mm) in comparison to previous designs [16, 21]. The NIRF-IVUS catheter comprises a ball lens-fiber (100/110/125um) assembly and piezoelectric transducer (center frequency: 50 MHz; bandwidth: 48%) which are integrated into the housing-drive shaft assembly (Ø=550 μm) for NIRF and IVUS imaging. Switching to a dual-layer drive shaft and reducing the rigid length of the housing, minimized available space for integration of both sensors. Therefore, we designed, and 3D printed (Nanoscribe, Karlsruhe, Germany) an alignment ring that spatially separated the IVUS wire and NIRF fiber for successful dual-sensor sensor integration (Fig. 1c). The catheter was inserted into a clinical 3.0F diameter imaging sheath (Opticross HD, Boston Scientific, Natick, MA) with a transparent imaging window and a monorail tip compatible with standard 0.014” coronary guidewire. NIRF excitation power at the tip of the catheter was set to 14 mW for all measurements.

Fig. 1:

Intravascular NIRF-IVUS system. (a) Imaging system with hardware components integrated onto a mobile cart. (b) High-speed MDU. (c) Previous catheter design [16, 21] and new, optimized NIRF-IVUS catheter design with a dual-layer DS, reduced rigid length and 3D-printed alignment ring (red box). Scale bar: 1 mm. NIRF: Near-infrared fluorescence; IVUS: Intravascular ultrasound; MDU: Motor drive unit; PMT: Photomultiplier tube; FORJ: Fiber optic rotary joint; DS: Drive shaft.

B. Software for System Control, Data Acquisition and Post-processing

Custom Labview software (Version 2023, National Instruments, Austin, TX) was developed to control the MDU, acquire and process data for real-time NIRF-IVUS image display during high-speed operation. All data acquired in the scope of his work was post-processed in Matlab (Version 2023a, MathWorks, Natick, MA) for image presentation. Distance-corrected NIRF data is generated using the procedure previously described [16]. Throughout this manuscript we report variability of quantified performance metrics as the mean standard deviation (SD) to show that, similar to the mean of performance metrics, there is minimal change in variability. All statistical analysis were performed in Prism 10 (Version 10.3.0, GraphPad Software, Boston, MA, USA). P<0.05 was considered statistically significant.

C. Phantom Experiments

A variety of phantoms were developed to characterize the enhanced NIRF-IVUS system and compare its performance between low- and high-speed imaging.

NIRF phantom.

A conical NIRF phantom with tissue optical properties (similar to [12, 18]) was made from 3.5% intralipid, 0.05% India Ink and 3% of noble agar which was mixed with distilled water at 90°C. The FDA-approved NIRF agent indocyanine green (ICG; Diagnostic Green LLC, Farmington Hills, MI; 500 nM) was added into the mixture which was poured into a mold to cool down and create a solid phantom with a hollow, conical lumen. The inside surface of the phantom was imaged with the NIRF-IVUS system at various distances from the catheter through either distilled water or fresh, whole rabbit blood to assess NIRF sensitivity at high imaging speed. NIRF signal-to-background ratios (SBRs) was calculated by dividing NIRF intensities through a NIRF background value averaged from intensities recorded outside the phantom. A Wilcoxon signed-rank test was performed to compare SBRs at five matched, spaced distances (0, 0.5, 1, 1.5 and 2 mm) for water-based and four matched, spaced distances (0, 0.5, 1, 1.5 mm) for blood-based speed evaluation.

Tortuosity phantom.

To test the new NIRF-IVUS catheter in a torturous pathway that simulates bending in coronary arteries, flexible tubing was 3D printed to model a human-sized aortic arch and coronary ostium. Additionally, a hollow octagon was 3D printed to evaluate the presence of NURD artifacts in IVUS images at different imaging speeds and with and without torturous catheter bending. A clinical, standalone IVUS imaging system (AVVIGO+, Boston Scientific, Marlborough, MA, USA) was operated at 30 FPS and 6 mm/s as a gold standard to compare performance to our hybrid NIRF-IVUS system. We quantified the presence of NURD artifacts by 1) measuring the misalignment angles (α_Mis; n=32) between each corner of the octagon-shaped phantom in our CAD model and cross-sectional images recorded by IVUS at four matched pullback locations (2, 9, 16 and 23 mm) and 2) calculating the mean of α_Mis±SD throughout the pullbacks. A one-way repeated measures ANOVA was performed to test for significant differences in α_Mis.

Stent phantom ex vivo.

Prior to in vivo imaging experiments, a prototypical clinical coronary stent (2.25 mm diameter x 23 length mm; Model Xience Sierra, Abbott Vascular, Santa Clara, CA, USA) was implanted into the iliac artery of a rabbit. Next, ICG (0.4 mg/kg) was injected intravenously to coat the microthrombus formed on the stent with a NIR fluorophore. After sacrifice of the animal, the stent was explanted for ex vivo NIRF-IVUS imaging and positioned in a 3D printed holder filled with distilled water. Pullback scans of the luminal surface of the stent were recorded to assess NIRF-IVUS co-registration and NIRF lateral resolution at low and high imaging speeds. Finally, ten spatially matched locations were identified within the stent scan, and lateral resolution was quantified at each location for both imaging speeds measuring full width at half maximum (FWHM) in the NIRF intensity profile of the stent strut width. A paired, two-tailed t-test was performed to compare lateral resolution between low- and high-speed imaging.

D. In Vivo Experiments

In vivo studies were performed on an institutional (Massachusetts General Hospital) animal care and use committee (IACUC) protocol (2023N000097). All procedures complied with institutional guidelines and the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. Prior to the in vivo experiments presented in this study, we tested our new NIRF-IVUS system in the aorta of New Zealand White (NZW) rabbits in vivo (n=10) to assess expected reduced breakage rates of our improved imaging catheter. Thereafter, stent-induced inflammation and atherosclerosis rabbit models were employed to test the new NIRF-IVUS system for high-speed imaging of arterial and stent pathobiology in coronary artery sized arteries in vivo.

Stent inflammation.

A clinical-grade coronary stent (2.5–2.75 mm x 12 mm; Xience Sierra, Abbott Vascular) was implanted into the right and left iliac arteries of healthy NZW rabbits (n=2) to develop stent-induced vascular inflammation [17]. Six days later, 400 nmol/kg of the NIRF molecular imaging agent ProSense VM110 (Perkin Elmer, Waltham, MA) was i.v injected to image cathepsin proteinase activity [13, 17]. Twenty-four hours later (day 7 post-stent implantation) standard angiography, standalone IVUS imaging and then hybrid NIRF-IVUS molecular-structural imaging of stented areas through blood, without blood displacement, was performed. Finally, stented iliac arteries were dissected after sacrifice of rabbits and cut longitudinally to acquire en face images with a custom-built fluorescence reflectance imaging (FRI) system and then underwent fluorescence microscopy (FM) with an epi fluorescence microscope (Model NiE Eclipse, Nikon Instruments, Melville, NY). Distance-corrected intravascular NIRF intensities (see section II-B) were averaged over the entire stent areas and in 5 mm region of interests (ROIs) centered at i) stent edges (proximal and distal) and ii) stent centers. NIRF intensities of the non-stented tissue areas were averaged as a third ROI excluding the 5 mm area from both stent edges which may be subject to vascular injury during stent implantation. Mean NIRF intensities of ROIs were divided by the mean NIRF background value (Minimum value of NIRF intensities averaged in 5 mm non-stented tissue areas) to calculate mean NIRF target-to-background ratios (TBRs)±SD. A Mann-Whitney U test was performed to test for significant differences in TBRs detected at low- and high-speed imaging.

Atherosclerosis.

The atherosclerotic rabbit model was obtained by feeding 1% cholesterol diet to healthy NZW rabbits (n=2) [23]. At the 2-week timepoint, vascular injury over 40 mm segments in the aorta weas induced with an embolectomy balloon (Model 3FR Fogarty catheter, Edwards Lifesciences, Irvine, CA) followed by installation of an angiotensin II osmotic mini pump (2ML4, Alzet, Cupertino, CA, USA) to accelerate atherosclerosis development. The final imaging experiments were performed 4 weeks later (6-week timepoint). First, in vivo angiography and standalone IVUS imaging of injured aortas were performed. Next, 0.4 mg/kg ICG was i.v. injected (equivalent to 0.13 mg/kg human dose, an FDA-approved dose, based on allometric scaling [24]) and in vivo NIRF-IVUS imaging of aortas at low and high imaging speed was performed 15–20 minutes after ICG injection. After sacrifice, in situ NIRF-IVOCT imaging [13] was performed and aortas were dissected for further ex vivo imaging and processing. The first aorta was imaged with FRI, fresh frozen and embedded in optimal cutting temperature compound to cut cross-sections for FM imaging and histological staining. The second aorta was cut longitudinally and imaged en face with FRI and FM. Mean intravascular NIRF TBRs±SD were calculated by dividing distance-corrected NIRF intensities of entire in vivo pullbacks by the NIRF background value (minimum value of NIRF intensities averaged in 5 mm tissue areas).

III. Experiments and Results

A. NIRF Phantom

The NIRF-IVUS catheter was inserted into the conical NIRF phantom (see II.C) to record NIRF-IVUS signals at variable distances between the catheter and the inside surface of the phantom (Fig. 2a). The NIRF phantom was 20 mm long with an inner diameter ranging from 1 mm to 6 mm (Fig. 2b). We recorded NIRF pullback maps at low and high imaging speeds through water (Fig. 2c) and rabbit blood (Fig, 2d). Fig. 2e shows NIRF SBRs plotted over distance (quantified by IVUS) for imaging pullbacks recorded at low and high imaging speeds through water or blood. NIRF sensitivity was similar for imaging at low and high speed with higher SBRs detected through water (13.4 at 1 mm distance) than in blood (2.3 at 1 mm distance) due to increased light attenuation. NIRF SBRs were compared between low- and high-speed imaging at matched spatial distances using a Wilcoxon signed-rank test. No significant differences in NIRF SBRs were observed between imaging speeds in either water (p = 0.63) or blood (p = 0.88).

Fig. 2:

NIRF-IVUS imaging of a NIRF phantom to evaluate NIRF sensitivity at high speeds. (a) Illustration of NIRF-IVUS imaging of the conical NIRF phantom containing an ICG concentration of 500 nM. (b) Dimensions of the NIRF phantom; Units are in mm. NIRF pullback maps of the phantom recorded at high speed (30 FPS, 6 mm/s) through (c) water and (d) whole rabbit blood. (e) NIRF SBR as a function of distance measured from the catheter to the phantom surface. Measurements shown for low and high imaging speed and through water and blood. Data shown as points, exponential fits as solid lines. NIRF: Near-infrared fluorescence; IVUS: Intravascular ultrasound; BG: Background; SBR: Signal-to-background ratio.

B. Tortuosity Phantom

Utilizing the 3D printed octagon-shaped phantom placed in a water bath, we performed NIRF-IVUS and clinical standalone IVUS imaging to assess the presence of NURD artifacts in IVUS images (Fig. 3a). The NURD phantom was imaged twice: 1) without any catheter bending and 2) when passing the catheter through a tortuosity phantom simulating a human-sized aortic arch (Fig. 3a, radius: 25 mm, ID: 20 mm) and a coronary ostium (ID: 4 mm) before inserting it into the NURD phantom (Length: 25 mm). Areas of misalignment (NURD artifacts) were highlighted in orange in IVUS images. Fig. 3b-d show representative IVUS cross-sections of the NURD phantom recorded without catheter bending. Good agreement between the octagon shape (outlined in cyan) and IVUS data (No NURD artifacts) was found for images recorded with the NIRF-IVUS catheter at both low (Fig. 3b) and high imaging speeds (Fig. 3c) and with the clinical IVUS system at high imaging speed (Fig. 3d). Fig. 3e-f show representative IVUS cross-sections of the NURD phantom recorded when guiding the catheters through the tortuosity phantom introducing catheter bending. NURD artifacts were found for images recorded with the NIRF-IVUS catheter at low imaging speed (NURD artifacts highlighted in orange, Fig. 3e) with a mean α_Mis of 7.0±8.3°. NURD artifacts were found for images recorded with the NIRF-IVUS catheter at low imaging speed (NURD artifacts highlighted in orange, Fig. 3e) with a mean α_Mis of 7.0±8.3° (95% CI: 4.0–10.0). NURD was reduced in images recorded with the NIRF-IVUS catheter at high imaging speed (Fig. 3f) and was similar to the clinical IVUS system at high imaging speed (Fig. 3g) with mean α_Mis of 4.4±5.7° (95% CI: 2.4–6.5) and 3.7±4.3° (95% CI: 2.2–5.3), respectively. We found no significant differences in α_Mis among groups (F(1.56, 48.48) = 3.07, p = 0.07, ε = 0.78; one-way repeated measures ANOVA).

Fig. 3:

NIRF-IVUS imaging of a tortuosity phantom to evaluate the presence of NURD artifacts. (a) Picture of the tortuosity phantom with 3D printed tubing simulating a human aortic arch and a coronary ostium; Illustration of the octagon-shaped NURD phantom. Representative IVUS cross-sections of the NURD phantom recorded without catheter bending with the NIRF-IVUS system at (b) low (5 FPS, 1 mm/s) and (c) high (30 FPS, 6 mm/s) imaging speed and (d) with a clinical IVUS system at high imaging speed (30 FPS, 6 mm/s). Representative IVUS cross-sections of the NURD phantom recorded in the tortuosity path with the NIRF-IVUS system at (e) low and (f) high imaging speed and (g) with a clinical IVUS system at high imaging speed. Shape of octagon-phantom highlighted in cyan. Areas with NURD artifacts and misalignment angles (α_Mis) highlighted in orange. NURD: Non-uniform rotational distortion; NIRF: Near-infrared fluorescence; IVUS: Intravascular ultrasound; All scale bars: 1 mm.

C. Ex Vivo Stent Phantom Imaging

After ICG i.v. injection and sacrifice, the explanted stent was positioned in the phantom holder (Fig. 4a). We inserted the NIRF-IVUS catheter and recorded NIRF-IVUS imaging pullbacks at low (5 FPS, 1 mm/s) and high (30 FPS, 6 mm/s) rotational and pullback imaging speeds, respectively. The stent structure can be seen in high detail on NIRF-IVUS pullback maps recorded at low (Fig. 4b) and high imaging speed (Fig. 4c). After intravascular NIRF-IVUS catheter imaging was completed, the stent was cut longitudinally and imaged en face with FRI. We detected homogenous NIRF intensities throughout the stent validating ICG uptake on the luminal stent surface (Fig. 4d). We found a high accuracy in co-registration between the NIRF and IVUS modality when looking at cross-sectional images recorded at low (Fig. 4e) and high (Fig. 4f) imaging speed. Finally, we quantified lateral resolution as the FWHM in the NIRF intensity profile of the stent strut width (Fig. 4g) at 10 matched locations (Blue arrows; Fig. 4 b,c). We calculated a mean lateral resolution of 82.7 ± 5.3 μm (95% CI: 78.9–86.5) for low-speed imaging and 83.4 ± 5.3 μm (95% CI: 79.6–87.2) for high-speed imaging. A paired t-test revealed no statistically significant difference between imaging speeds (t(9) = 0.32, p = 0.76), with a mean difference of 0.7 μm (95% CI: – 4.3 to 5.7 μm).

Fig. 4:

NIRF-IVUS imaging of an NIR fluorescent coronary stent phantom to evaluate NIRF-IVUS co-registration and NIRF lateral resolution at high imaging speed. (a) Photograph of a stent dissected from a rabbit iliac artery and mounted in a holder for imaging. NIRF and IVUS pullback maps of the stent recorded at (b) low (5 FPS, 1 mm/s) and (c) high (30 FPS, 6 mm/s) imaging speed (Black arrows indicate location of cross-sections shown in e and f; blue arrows indicate matched locations for quantification of NIRF lateral resolution. (d) En face WL and FRI reference image of stent. Exemplary NIRF-IVUS cross-sectional images of the stent recorded at (e) low and (f) high imaging speed. (g) Exemplary NIRF intensity profile of stent strut width to quantify FWHM indicative of NIRF lateral resolution. Data shown as points, gaussian fit as solid lines. NIRF: Near-infrared fluorescence; IVUS: Intravascular ultrasound; WL: White light; FRI: Fluorescence Reflectance Imaging. Scale bars: 1 mm.

D. In Vivo NIRF-IVUS Imaging of Stent Inflammation

In vivo NIRF-IUVS imaging of stent-induced arterial inflammation was conducted in rabbit iliac arteries (Fig. 5; Described in II-D). Seven days after stent implantation was completed, angiography and standalone IVUS scans were recorded. The implanted stent is visible on angiography and standalone IVUS pullback map (Fig. 5a). Next, hybrid NIRF-IVUS imaging pullbacks at low (Fig. 5b) and high (Fig. 5c) imaging speed were recorded to image stent inflammation in vivo. The NIRF pullback maps revealed enhanced ProSense VM110 signal, indicating inflammatory cathepsin protease activity, in stented tissue areas expected to harbor arterial inflammation as a vascular injury response [17]. In accordance with these findings, representative NIRF-IVUS cross-sections recorded at either low or high imaging speed display low ProSense VM110 NIRF signal in non-stented arteries (Fig. 5d) and high NIRF signal in stented arteries (Fig 5e). Enhanced NIRF signal in stented tissue areas was confirmed by FRI (Fig. 5f) and epi-FM (Fig. 5g). The highest NIRF signals were detected at stent edges (Areas 1 and 3, Fig. 5g) in comparison to the stent center (Area 2, Fig. 5g) and non-stented tissue area (Area 4, Fig. 5g). We found no significant differences (p=0.65, Mann-Whitney U test) between NIRF TBRs averaged over entire stent areas detected at low (1.88±0.54, 95% CI: 0.95–2.68) and high (2.00±0.65, 95% CI: 0.95–3.13) imaging speed (Fig. 5h). Finally, we calculated TBRs from different areas in NIRF pullbacks recorded at low and high imaging speed (Fig. 5i). We found mean NIRF TBRs of 1.97±0.64 (low-speed; 95% CI: 1.44–2.51) and 2.24±0.72 (high-speed; 95% CI: 1.65–2.85) at stent edges, 1.57±0.43 (low-speed; 95% CI: 0.88–2.26) and 1.64±0.41 (high-speed; 95% CI: 0.99–2.30) at stent centers and 1.03±0.05 (low-speed; 95% CI: 0.96–1.11) and 1.10±0.11 (high-speed; 95% CI: 0.92–1.27) in non-stented tissue areas.

Fig. 5:

High-speed imaging of stent inflammation in vivo. (a) Angiography and standalone IVUS imaging of implanted metallic stent in rabbit right iliac artery. Hybrid in vivo NIRF-IVUS pullback maps of stented tissue area recorded at (b) low and (c) high imaging speed. Representative NIRF-IVUS cross-sections recorded in (d) healthy and (e) stented tissue area (Locations indicated in c). (f) En Face FRI and WL image of explanted stent and artery. (g) Epi-FM image of different tissue and stent areas (Locations 1–4 indicated in f). NIRF TBRs averaged (h) over entire stent (n=4) areas and (i) at stent edges (n=8), stent centers (n=4) and non-stented (n=4) areas in NIRF pullback maps which were recorded at low and high imaging speed. TBRs plotted as black points; Boxes, bars and whiskers represent interquartile range, mean values and minimum and maximum values. n.s.: not significant (p > 0.05). IVUS: Intravascular ultrasound; NIRF: Near-infrared fluorescence; FRI: Fluorescence Reflectance Imaging; WL: White Light; FM: Fluorescence Microscopy; GFP: Green Fluorescence Protein; TBR: Target-to-background ratio; All scale bars unless otherwise specified: 1 mm.

E. In Vivo Detection of Atherosclerosis by ICG-enhanced NIRF-IVUS

Fig. 6 and 7 present ICG-enhanced NIRF-IVUS imaging of atherosclerosis acquired at low and high imaging speed in rabbit aortas in vivo. ICG, an FDA-approved NIR fluorophore, is known to target lipid-rich inflamed atheroma and vascular injury and enables in vivo intravascular NIRF imaging [11, 14, 19]. Angiography and standalone IVUS pullback images of the injured zone in the rabbit aorta were recorded (Fig. 6a). Following i.v. injection of ICG at a dose of 0.4 mg/kg, the aorta was scanned in vivo with NIRF-IVUS at different imaging speeds. NIRF-IVUS detected focal ICG uptake in atherosclerotic plaque areas recorded at low (Fig. 6b) and high (Fig. 6c) imaging speeds 15–20 min after ICG injection. We confirmed ICG-uptake patterns with ex vivo FRI (Fig. 6d) after the animal was sacrificed and the aorta was dissected (35 min post ICG injection). FM and histological staining revealed minimal ICG uptake in non-plaque tissue areas (Fig. 6e) that showed minor accumulations of lipids (Fig. 6e-ORO) and macrophages (Fig. 6e-RAM-11). In contrast, increased ICG uptake was confirmed in more evident plaque areas (Fig. 6f) showing intimal thickening with accumulated lipids (Fig. 6f-ORO) and macrophages (Fig. 6f-RAM-11).

Fig. 6:

High-speed, ICG-enhanced NIRF-IVUS imaging of atherosclerosis plaque permeability, lipid deposition and inflammation in rabbit aorta in vivo. (a) Angiography and standalone IVUS (Boston Scientific) pullback map of rabbit aorta; NIRF-IVUS pullback maps of rabbit aorta recorded at, (b) low and (c) high imaging speed. (d) WL and FRI image of excised aorta. (e-f) FM images and histological staining of exemplary sections (locations of sections marked in d as upper and lower orange dotted lines). NIRF: Near-infrared fluorescence, IVUS: Intravascular ultrasound, FRI: Fluorescence reflectance imaging, WL: White light, FM: Fluorescence microscopy, TRI: Trichrome, ORO: Oil Red O. Scale bars: 1mm.

Figure 7:

Comparison of NIRF TBRs between low- and high-speed NIRF-IVUS imaging of ICG in atherosclerotic rabbit aorta in vivo. NIRF-IVUS pullback maps of a rabbit aorta recorded in vivo at, (a) low and (b) high imaging speed. (c) Ex vivo NIRF-IVOCT imaging of rabbit aorta performed in the absence of blood. (d-e) Exemplary in vivo NIRF-IVUS and ex vivo NIRF-IVOCT cross-sections (locations of co-registered cross-sections marked in b and c). (f) En face WL and epi FM image of excised aorta. (g) NIRF TBRs calculated from low and high-speed NIRF-IVUS imaging pullbacks. Bars and whiskers represent mean and SD values. NIRF: Near-infrared fluorescence, IVUS: Intravascular ultrasound, IVOCT: Intravascular optical coherence tomography, WL: White light, FM: Fluorescence microscopy, TBR: Target-background-ratio. Scale bars: 1 mm unless otherwise specified.

We repeated in vivo intravascular NIRF-IVUS imaging of ICG targeting of atherosclerosis in a second rabbit and recorded NIRF-IVUS pullback scans at low (Fig. 7a) and high imaging speeds (Fig. 7b) of the plaque-rich rabbit aorta. Both scans detected a focal ICG uptake at the 15–20 mm pullback position localizing to atheroma by IVUS. Focal ICG uptake was further confirmed by ex vivo NIRF-IVOCT imaging (Fig. 7c). Exemplary NIRF-IVUS and NIRF-IVOCT cross-sectional images showed relatively low ICG concentration in a minimal plaque area (Fig. 7d) and increased ICG concertation in a more evident plaque region (Fig. 7e). En face epi-FM confirmed ICG-uptake patterns of minimal ICG deposition in healthy aorta tissue (Areas 1 and 3, Fig. 7f) and increased ICG in the center of the plaque area (Area 2, Fig. 7f). Finally, we calculated NIRF TBRs and averaged them over the entire NIRF-IVUS pullback length in the aorta for each rabbit (Fig. 7g). We found similar NIRF TBRs for low and high-speed imaging with a mean of 5.15±3.63 (low-speed) and 5.17±3.83 (high-speed) for the first rabbit (Results summarized Fig. 6) and 3.10±0.99 (low-speed) and 2.99±0.83 (high-speed) for the second rabbit (Results summarized Fig. 7a-f).

IV. DISCUSSION

In this study, we present a next-generation, clinical-type high-speed NIRF-IVUS imaging system and compared results between low-speed (5 FPS, 1 mm/s) and high-speed imaging (30 FPS and 6 mm/s) in phantoms and in vivo rabbit models of stent-induced vascular injury and atherosclerosis. We demonstrate that the re-engineered system can provide hybrid NIRF-IVUS molecular-structural images at increased rotational and pullback speeds without degradation of image quality. To the best of our knowledge, this is the first time that NIRF-IVUS imaging is demonstrated with a clinical size coronary catheter (3.0F) at a frame rate that is identical and a pullback speed that exceeds many conventional, clinical standalone IVUS systems (0.5–1 mm/s) and approaches the maximum speed of the state-of-the-art systems (e.g., maximum pullback speed: 8 mm/s, Boston Scientific). Complementing IVUS with NIRF provides the opportunity to enhance image quality at increased pullback speeds by offsetting reduced anatomical resolution (IVUS) with complementary molecular information (NIRF), representing a translational advancement for intravascular imaging. As part of this current work, the following technical improvements were integrated to enhance catheter helical pullback motion: (1) replacement of the single with a dual-layer drive shaft for improved torque delivery from the MDU to the catheter distal tip and protection of fragile components while maintaining a clinical drive shaft and sheath diameter of 0.55 mm and 3.0F (1.0 mm), (2) decrease of the rigid length of the catheter tip by 33% from 3 mm to 2 mm for improved catheter flexibility, (3) integration of a 3D-printed alignment ring for improved sensor arrangement in a miniaturized housing-drive shaft assembly, and (4) development of a hybrid MDU with a custom-made v-block design for increased stability that enables maximum rotational and longitudinal pullback speeds of 6000 rpm and 10 mm/s, respectively. Phantom experiments validated that important technical specifications of our NIRF-IVUS system were not comprised when increasing imaging speed. High-speed imaging of a NIRF sensitivity phantom with tissue-like optical properties (Fig. 2) showed that the re-engineered NIRF-IVUS system could detect a small concentration of ICG (500 nM) at a distance up to 1.2 mm from the catheter through ex vivo whole blood (SBR=1.5), without flushing, identical to clinical IVUS workflows. We expect NIRF sensitivity to increase for in vivo imaging due to decreased NIRF attenuation in in vivo vs. ex vivo blood [19]. Furthermore, we found that NURD artifacts in IVUS images was reduced (Mean α_Mis=7° to 4.4°) when imaging an octagon-shaped phantom at increased imaging speed in a torturous aorta and ostium model (Fig. 3f). We hypothesize that the increase in rotational speed may have improved torque delivery from the MDU to the catheter tip causing a more stable catheter rotation and a reduction in NURD. Most importantly, NURD artifacts in IVUS images recorded with our hybrid catheter were comparable to images recorded with a commercial IVUS system (Mean α_Mis=4.4° vs 3.7°; Fig. 3f,g) which indicates that our NIRF-IVUS system offers rotational stability similar to clinical standards for in vivo imaging of coronary-sized arteries. The quantified NURD of 4.4° is not expected to impede intravascular NIRF imaging performance, as it corresponds to an arc length of approximately 77μm in a typical 3.0 mm diameter coronary artery - below the measured optimal lateral resolution in water (83μm). Furthermore, the effective resolution is anticipated to degrade further in vivo due to increased scattering in blood, rendering the impact of this angular deviation negligible. Finally, our hybrid NIRF-IVUS system achieved accurate co-registration between both modalities (Fig. 4e-f) and a lateral NIRF resolution of 84 μm (Fig. 4g) independent to imaging speed when scanning an ICG-coated stent phantom (see III-C).

Previous studies have demonstrated the value of standalone NIRF or hybrid NIRF-IVUS imaging in in vivo preclinical animal models of atherosclerosis. Intravascular NIRF has been demonstrated to image injected NIRF molecular imaging agents like ProSense VM110 to detect inflammatory cathepsin protease activity [16, 17] or ICG to highlight vascular injury and lipid- and macrophage-rich plaques [14, 19]. However, previous NIRF-IVUS studies all demonstrated limited imaging speed (maximum 5FPS and 1 mm/s), hindering real-time image display and causing long image acquisition times (e.g. 40 seconds for a 40 mm pullback), a feature less desirable compared to contemporary clinical IVUS, or to IVOCT (20–40 mm/s pullback speeds). To address this limitation, we tested our high-speed system in two different rabbit models of atherosclerosis and reported real-time rotational imaging speeds and sixfold reduction in acquisition time compared to previous NIRF-IVUS studies. First, in vivo NIRF-IVUS imaging of implanted stents in rabbit iliac arteries demonstrated improved assessment of arterial inflammation by detecting a high cathepsin proteinase activity signal in stented areas (Fig. 5b,c,e) not visible on angiography or standalone IVUS (Fig, 5a). In general, we did not observe a change in imaging results when increasing imaging speed and calculated TBRs that were not significantly different from NIRF maps recorded at low and high speed (Fig. 5h,i). Similar to a previous study [17], we detected higher in vivo proteinase activity signal at stent edges (Fig. 5i, mean TBR=2.24) than in stent centers (mean TBR=1.64) or non-stented tissue areas (mean TBR=1.10) when imaging at high speed. These findings were confirmed by ex vivo FRI (Fig. 5f) and epi FM (Fig. 5g). The observed patterns of inflammatory activity may be the result of increased mechanical injury of the vascular wall at the edges in comparison to the center of the stent. The importance of detecting inflammation induced by stent- or angioplasty-induced vascular injury and atherosclerosis has been well established. For example, the proximal and distal stent edge vascular response has been clinically recognized as a hotspot for complications like stent edge restenosis [25, 26]. Our study underlines that NIRF-IVUS imaging may be an effective tool to highlight stent-related inflammatory response during CAD assessment and PCI which may help to avoid clinical events, like vulnerable plaque progression and stent-edge restenosis and improve long-term patient outcomes. Second, we tested our high-speed NIRF-IVUS system for imaging ICG, an FDA-approved plaque-targeted NIRF agent [11, 14], in atherosclerotic rabbit aortas. NIRF-IVUS imaging highlighted early-stage plaque regions by detecting increased ICG uptake (Fig. 6b,c) co-localized to accumulated lipids and macrophages (Fig. 6d,f) not visible on standalone IVUS maps and angiography (Fig. 6a). In contrast, healthy tissue regions without relevant lipid and macrophage presence showed only minor ICG uptake (Fig. 6d,e). Furthermore, we demonstrated that ICG-enhanced NIRF-IVUS can provide hybrid molecular-structural assessment of atherosclerosis at clinical imaging speed without sacrificing NIRF sensitivity (Fig. 7g). In addition, we confirmed similar ICG uptake patterns in a plaque region with high-speed NIRF-IVUS and flushed, ex vivo NIRF-IVOCT (Fig. 7d,e). Finally, our previous 3F catheter design [16,21] suffered from significant fragility resulting in routine fiber breakage during in vivo imaging which often limited the ability to acquire just one reliable pullback even at low speed. Our new dual-layer DS design improved catheter durability which enabled us to record pullbacks with a single catheter, including acquisitions in a torturous phantom and 14 different animals in vivo. In this work, we report SD of performance metrics as a measure for variability which is not equivalent to noise. It is instead caused by variation in features and can act alone as a standalone metric of quality or a diagnostic parameter especially when calculating SDs of TBRs over entire in vivo pullback maps including healthy and pathological tissue areas in our atherosclerosis imaging model (II.D, III.E and Fig. 7g). An important requirement for the clinical translation of novel intravascular imaging technologies is catheter size and flexibility. In this work, we improved the design of our NIRF-IVUS catheter to enable high-speed image acquisition. Importantly, we achieved high-speed imaging capabilities without violating clinical requirements for catheter dimension. In particular, the outer diameter (OD_coil=550 μm; OD_sheath=3.0F=1.0 mm) and rigid length (2 mm) of our hybrid NIRF-IVUS catheter are identical to state-of-the-art, standalone IVUS catheters (Opticross HD, Boston Scientific) routinely used in clinical intravascular imaging procedures. We used the FDA-approved NIR agent, ICG, for most of our experiments to highlight the translational potential of ICG-enhanced NIRF-IVUS. Finally, we limited NIRF excitation power to 14 mW to comply with the ANSI safety standards for maximum permissible exposure limits [27].

V. LIMITATIONS

While this study describes a major technical advancement for intravascular NIRF-IVUS imaging, additional improvements are possible. To further reduce the presence of NURD artifacts, we plan to increase space between the inner coil of the drive shaft and the inserted fiber and wire, to increase flexibility during extensive catheter bending. In addition, we noticed that contrast in IVUS images was slightly decreased due to the reduction of acquired A-lines when acquiring data at high imaging speed. This observation also limited us to further increase pullback speed to the maximum potential of our MDU (10 mm/s). To account for the reduced time available for data acquisition during high-speed operation, we plan to implement an increased trigger frequency for IVUS A-line acquisition to maintain IVUS image quality. Similar as in previous studies [14, 16, 17, 19], we did not detect full circumferential NIRF signals in stented coronary-sized arteries (Fig. 5b,c) due to increasing NIRF attenuation of ProSense VM110 signals from tissue and stent areas located at a distance >1.4 mm through whole blood from the NIRF-IVUS catheter. While we expect this NIRF sensitivity to be sufficient for intravascular NIRF imaging through blood in human coronary arteries – particularly in culprit plaque areas with an average lumen diameter of 0.62 mm [28] – we plan to investigate additional methods to further enhance NIRF sensitivity. These include the use of a dual-clad fiber [29] and increasing the NIRF excitation wavelength from 750 nm to 780 nm to improve imaging with ICG or IR800CW. Finally, as a proof-of-concept study with a limited number of animals (n=2 per model), the findings should be interpreted with caution. The small sample size prevented an in-depth statistical analysis to detect subtler differences in image quality. However, image quality was sufficient to demonstrate similar TBRs recorded at low and high imaging speeds in the stent inflammation model. Given the successful demonstration of our high-speed NIRF-IVUS system in a torturous phantom and in vivo rabbit models, the next step is to test our system in vivo coronary arteries of swine as part of a feasibility study to demonstrate reduced presence of imaging artifacts at high rotational imaging speeds and further facilitate clinical translation.

VI. CONCLUSION

In summary, we developed a high-speed NIRF-IVUS system with a clinical-sized catheter and validated it in phantoms and in vivo rabbit models at clinical imaging speed. The technical achievements of this work demonstrate that NIRF-IVUS imaging should be seamlessly integrated into clinical intravascular imaging of coronary arteries and during PCI where acquisition times need to be minimized and, thus, provide a major step for the clinical translation of the NIRF-IVUS technology. We were able to accomplish a sixfold increase in imaging speed in comparison to our previously reported NIRF-IVUS system which achieved the clinical size requirements for catheters for the first time [16, 21]. NIRF-IVUS imaging at clinical imaging speed may improve the comprehensive molecular-structural assessment of CAD and guidance of PCI in living subjects.