2.1.

FLIm-OCT System Description

The dual-modality system is composed of an imaging catheter, an MDU, and a console as depicted in Fig. 1. It includes OCT and FLIm-specific components in the console and MDU, as well as shared components that enable bimodal imaging, distributed between catheter and MDU, as described in the following.

Fig. 1

(a) FLIm-OCT system schematic. The system is composed of an imaging catheter, an MDU, FLIm, and OCT modules. (b) 3D model of rotary collimator. The motor drives the rotation of shaft. The encoder read out of the angular position of the rotating shaft is used for closed-loop control of the shaft rotation as well as image reconstruction. (c) Top view of the MDU highlighting the OCT path, FLIm excitation and detection paths, and rotary collimator.

2.2.

FLIm Module

The FLIm capability relies on a pulse sampling implementation, derived from instrumentation previously developed for multimodal intravascular use,¹⁹ and that incorporates individual solid-state detectors, similar to a FLIm instrument designed for intraoperative use recently developed by our group.²⁰ The excitation light source consists of a pulsed 355-nm diode-pumped solid-state laser (4-kHz repetition rate, $2\text{-}\mu \mathrm{J}$ pulse energy, STV-02E-140, TEEM photonics, France). The excitation light is first transmitted to the MDU with a $50\text{-}\mu \mathrm{m}$ core multimode fiber (Polymicro FVP050055065, Arizona, United States) and then coupled into the catheter as described in Sec. 2.5. The FLIm signal detection is implemented by first resolving the fluorescence signal into three spectral bands (390/40, 450/45, and $540/50\mathrm{nm}$) by a set of dichroic mirrors and bandpass filters.¹⁹ The optical signal from each spectral band is then focused (C610TME-A, Thorlabs, New Jersey, United States) into the sensitive area of individual photodetector modules (APD430A2-SP1, Thorlabs, New Jersey, United States). Each module consists of a UV-enhanced avalanche photodiode (APD) ($200\text{-}\mu \mathrm{m}$ diameter sensitive area, S12053-02, Hamamatsu, Japan), adjustable biasing circuitry, and trans-impedance amplifier ($5\mathrm{kV}/\mathrm{A}$). The output from each APD module is sampled by a $6.25\mathrm{GS}/\mathrm{s}$ digitizer (NI PXIe-5185, National Instruments, Austin, Texas, United States).

2.3.

OCT Module

The intensity-based swept-source OCT (SS-OCT) is implemented as described in our previous work based on an OCT OEM engine from Axsun Technology.¹⁵ It includes a $1310\pm 55\mathrm{nm}$ swept-source laser and enables an A-scan rate of 50 kHz. The interferometry is realized with fiber couplers for splitting the laser output from the OEM engine to the sample arm and reference arm and mixing the reflected light from both arms. A single-mode fiber (SMF) is used in the reference arm to reduce the ghost artifacts caused by the light leakage to the multimode cladding, different from previous work.¹⁵ The material of the SMF core is the same as that of DCF to avoid introducing dispersion. The sample arm light is coupled from the OCT modality to the MDU by an SMF and finally coupled to the DCF based fiber probe. The interference signal is generated in the fiber coupler and a 180-deg phase shift is introduced between the two outputs of the fiber coupler.²¹ The differential signal of these two outputs is detected with the dual balanced detector and sampled by a 12 bit, $500\mathrm{MS}/\mathrm{s}$ FPGA DAQ board in the OCT engine.

2.4.

Motor Drive Unit

The MDU fulfills multiple roles of mechanical actuation of the catheter, multiplexing and demultiplexing of FLIm and OCT excitation and collection light, and detection of the FLIm signal. A UV collimator and an infrared (IR) collimator are used to generate collimated UV (FLIm) and IR (OCT) beams. The UV beam is reflected by a first dichroic mirror used to separate FLIm excitation and fluorescence emission light (Di01-R355, Semrock, Rochester, NY), and then merged with the OCT beam by a second dichroic mirror (FF735-Di02-25x36, Semrock, Rochester, New Jersey, United States). A periscope was integrated within the OCT optical path provide the four degrees of freedom required for proper beam alignment. The overlapping IR and UV beams are sent to a custom fiber optic rotary collimator (FORC) (Fig. 1).

The FORC is based on a stationary air bearing (OAV0500IB, Princeton, New Jersey, United States) and a rotating shaft. The shaft comports an FC/PC fiber receptacle (for connection with the catheter’s imaging core) as well as a focusing lens (11-mm focal length, A397TM, Thorlabs, New Jersey, United States) adjustable over three degrees of freedom (two orthogonal continuous lateral adjustments and discrete axial adjustment) that enables focusing of the IR and UV beams into the catheter’s core and inner cladding, respectively (see Sec. 2.5). The FORC shaft is driven by a brushless motor (EC-max 22, Maxon motors, Sachseln, Switzerland). Angular position information of the shaft used for FLIm and OCT image reconstruction is provided by the optical angular encoder (US Digital, Vancouver, Washington, United States) and used as the feedback to the shaft rotation speed controller (ESCON 24/2, Maxon Motors, Sachseln, Switzerland) [Fig. 1(b)].

Both fluorescence emission and OCT backscattered signal transmitted by the catheter are collimated back by the FORC lens. The IR OCT beam is transmitted to the IR collimator for OCT detection, whereas the fluorescence optical signal is spectrally resolved into three different spectral bands and directed to the respective APD modules (see methods Sec. 2.2).

2.5.

Rotational Dual-Modality Catheter

The catheter consists of a rotating imaging core housed in a saline solution-filled sheath. The 1.6-m imaging core consists of a double-clad fiber (SM-9/105/125-20A, Nufern, East Granby, Connecticut, United States) enclosed in a two-layer torque coil (Asahi Intecc, Aishi, Japan). The proximal end of the imaging core was terminated with a 150-mm stainless steel shaft and an FC/PC connector for connection to the FORC shaft. A broadband freeform reflective micro-optic was integrated at the distal end of the imaging core; beam reflection off the curved aluminum-coated reflective surface enables both beam redirection and accurate focusing over an extended wavelength range.¹⁸ The DCF and the freeform optic were actively aligned by coupling 1310-nm light into the fiber’s core and adjusting the fiber/s relative position in the vertical direction to obtain the nominal beam forward tilt. The optic was then attached to the fiber using UV-curing acrylate optical adhesive (OG603, Epotek, Billerica, Massachusetts, United States). The imaging sheath is composed of a 4-Fr braided PEBAX shaft (Merit Medical, Utah, United States), terminated with an optically transparent polymethylpentene 1.25-mm outer diameter imaging section (Vesta Inc., California, United States).

2.6.

Data Acquisition and Processing Methods

Data was acquired using a catheter rotation speed of 10 rps to enable sufficient FLIm angular sampling with the current laser (4 kHz, 400 points/rotation). The average fluorescence lifetime for each spectral band was processed for each point measurement using a Laguerre basis expansion technique.²² For OCT processing, the A-lines were obtained by Fourier transforming the measured spectra. To suppress the fixed pattern noise observed in the signal, the reference spectrum extraction was applied.²³ Cross-sectional OCT images (B-scans) were reconstructed based on the angular information obtained from the FORC encoder. Finally, lifetime color-coded B-scans for each FLIm spectral band were generated by interpolating the average lifetime values for each OCT A-line angular location and setting each A-line’s color based on the color-coded average lifetime value.

2.7.

Fiber Optic Rotary Collimator Alignment and Characterization

A key figure of merit of the FORC performance is the coupling efficiency (CE) of light between the FLIm OCT modality and the catheter (355 to 1365 nm). A high and stable CE shall be achieved during the rotation of the FORC shaft. Light coupling between the core of the OCT subsystem fiber and the core of the rotating imaging DCF is more challenging due to the small core sizes ($9\mu \mathrm{m}$) than the coupling of FLIm optical beams to and from the DCF’s inner cladding ($105\text{-}\mu \mathrm{m}$ diameter); therefore, we optimized the alignment of FORC by maximizing the OCT coupling performance.

The IR CE was maximized by first minimizing the eccentricity of the collimated single-mode beam exiting the FORC, measured from the image of the focused IR spot behind the OCT collimator lens [Fig. 2(a)]. The measurement was performed using a custom beam profiler based on a camera (½″ CMOS sensor, 3.1MPix, AmScope, Irvine, California, United States), $100\times$ Plan Apo objective, and a 150-mm focal length tube lens that provided a pixel size of 0.08 mm. Variation of the beam’s centroid location was minimized by adjusting the lateral position of the FORC lens during the shaft rotation [Fig. 2(b)]. In a second step, optical coupling of the rotary fiber to the stationary fiber was maximized by adjusting the vertical/horizontal position and yaw/pitch of the IR beam directed to the OCT collimator by adjusting the two flat mirrors (M1, M2), as well as the IR collimator focus. The effect of IR beam misalignment was investigated by calculating the CE of single-mode coupling versus the offset between beam centroid and fiber core center using physics optics propagation module in Zemax [Fig. 2(c)].

Fig. 2

(a) Beam profile of the focused IR beam after the OCT IR collimator lens. The large red circle has a diameter of $9\mu \mathrm{m}$, same as the fiber core. The small red circular spot is the calculated beam centroid. (b) Trace of beam centroid positions when rotating the FORC. (c) The simulated single-mode CE over the later shift of beam spot relative to the fiber core center. (d) The measured IR and UV CE during the rotation of FORC.

The UV excitation coupling was then maximized by adjusting the two dichroic mirrors on the UV path [Fig. 1(b): DIR, DL] as well as the focusing of the UV collimator.

Finally, the coupling efficiencies of OCT IR ($1310\mathrm{nm}\pm 55\mathrm{nm}$) and FLIm UV excitation light (355 nm) were acquired by measuring the power before and after the FORC [Fig. 2(d)].

2.8.

Rotational Catheter Performance Characterization

The FLIm lateral resolution and collection efficiency of the freeform reflective optic were compared using a ray-tracing simulation of the performance obtained from a side-viewing ball lens described in a recent multimodal catheter publication.¹⁷ A nonsequential model was built using the simulation model of the freeform reflective optic and the dimensions of the side-viewing ball lens probe [Fig. 3(a)]. A $105\text{-}\mu \mathrm{m}$ diameter circular source was used to simulate the light exiting the DCF cladding with the divergence computed from the NA of the inner cladding. For simplicity, the glass tube surrounding the ball-lens design was omitted. The axial and lateral resolutions were obtained by simulating the beam profile and computing the FWHM in both directions at multiple distances from the optical axis. The beam centroid positions were also computed and used to simulate each design’s collection efficiency with respect to target distance. For each distance from the fiber’s optical axis, a circular Lambertian source whose diameter equals the RMS diameter of the excitation beam obtained at the previous step was defined by illuminating a collimated beam to the Lambertian scattering plane. A circular detector with an acceptance angle defined from the fiber’s inner cladding NA and located in correspondence to the fiber’s distal end was used to compute the collection efficiency, defined as the ratio of detected photons and emitted photons.

Fig. 3

(a) Ray-tracing model for simulating the FLIm excitation and emission collection. (b) Ray tracing comparison between the freeform optics and ball lens probes. (c) Comparison of FLIm excitation FWHM and emission collection efficiency between the freeform optic and ball lens probes. (d) IR and UV beam profiles at different distances from the optical axis. (e) The UV and IR FWHM were calculated from the beam profiles versus the distance from the optical axis.

The FLIm excitation beam (355 nm) and OCT IR beam (1310 nm) were measured at different distances from the fiber’s optical axis using a second custom beam profiler based on a camera (½″ CMOS sensor, 3.1 MPix, AmScope, Irvine, California, United States), $20\times$ Plan Apo objective and a 40-mm focal length tube lens that provide a pixel size of $1.4\mu \mathrm{m}$.

2.9.

Human Coronary Artery Imaging

The artery specimen was provided by the University of Pennsylvania heart transplant program in accordance with current legal requirements and guidelines and was approved by the University of Pennsylvania Hospital Institutional Review Board as well as UC Davis Biological Use Authorization. This 50-mm long coronary artery section was immersed in phosphate-buffered saline, a guidewire was inserted in the segment, the FLIm-OCT catheter was inserted, and the guidewire was removed before imaging to obtain a full view of the vessel without guidewire shadow. After imaging, it was inked, fixed (10% neutral-buffer formalin), sectioned in $\sim 2.5\text{-}\mathrm{mm}$-long rings, and embedded in paraffin. Sections ($5\text{-}\mu \mathrm{m}$ thick) were stained with hematoxylin and eosin stain and modified Movat’s pentachrome. Anti-CD68 antibodies were used to identify macrophages. Registration of imaging data with histology was performed based on angular registration of the container bottom surface (visible at the beginning of the pullback) and the top surface inking of the artery sample. Axial registration was performed based on pullback location and ring number. Final adjustments of the registration were performed by matching morphological features between the OCT data and histological sections.

3.1.

FORC Performance Evaluation Results

The stability of free-space beam characteristics (position and dimensions) during rotation is key to ensure efficient single-mode fiber coupling. Figure 2(a) shows the focused IR beam after the OCT collimator lens and the red circle represents the fiber core with a diameter of $9\mu \mathrm{m}$. That demonstrates a good match between the sizes of fiber core and focused beam profile. The beam centroid was calculated from the beam profile [Fig. 2(a)] and traced when rotating the FORC. Figure 2(b) depicts the beam centroid positions when rotating the FORC, after the alignment was optimized. The residual eccentricity of the optical axis of the FORC during the rotation generates a shift of the beam centroid position with respect to the location of the OCT stationary collimator fiber core. This shift, characterized by three times the standard deviation of the centroid coordinates, was measured to be 0.40 and $0.47\mu \mathrm{m}$ in the lateral and vertical directions, respectively. Figure 2(c) shows the simulated CE when the beam centroid is laterally offset from the fiber core center. For a $0.5\text{-}\mu \mathrm{m}$ offset, an expected coupling of 78% was obtained. Figure 2(d) shows the measured IR and UV CE for the OCT and FLIm excitation during the rotation of FORC. The average IR CE was 75.7%, which is very close to the simulation result. The min-max difference was 0.8%, indicating an extremely stable coupling to the fiber core. The average UV CE is 45.7%, with a min-max difference of 0.7% demonstrating high stability.

3.2.

Freeform Reflective Optics Evaluation

The optical performance of freeform optic and ball lens was evaluated from raytracing simulation as described in Sec. 2.8 and compared as shown in Figs. 3(b) and 3(c). Figure 3(b) shows that the waist of the UV beam obtained from the ball lens catheter is located inside the catheter sheath (2.6-Fr diameter), thus all the FLIm measurements are performed with the vessel wall located past the beam waist, reducing both lateral resolution and collection efficiency. The freeform optic significantly improves the UV waist position, now located at $\sim 100\mu \mathrm{m}$ from the sheath’s outer surface, and the excitation beam divergence is also drastically reduced. This improvement resulted in an improved FLIm lateral resolution as shown by the FLIm excitation beam full width at half maximum [FWHM, Fig. 3(c), black curves]. The excitation beam FWHM using the freeform optics is $\sim 30\%$ smaller than that using the ball lens. The emission collection efficiency is also improved significantly by using the freeform optic for all distances above 0.6 mm from the catheter’s central axis [Fig. 3(c), blue curves]. Variations in signal collection due to variations of catheter to vessel wall distance at different angular positions of the core are detrimental to FLIm signal SNR. This variation, expressed as ratio of FLIm emission/collection efficiency at the catheter’s surface and 1.5-mm distance is much lower for the freeform optic ($R=5.6$) than the ball lens design ($R=29.1$), demonstrating another key benefit of this implementation.

The performance of freeform optic was validated by beam profiling. The beam profiles obtained for both IR (delivered through the fiber core) and UV (delivered through the inner cladding) are displayed in Figs. 3(d) and 3(e). For IR and UV, beam waist FWHM of 17 and $50\mu \mathrm{m}$, respectively, are obtained. The position of the beam waist is also in good agreement with ray-tracing simulations [Fig. 3(b)]. The OCT IR beam is focused at 1.15 to 1.35 mm from the probe axis with a small astigmatism (0.2-mm focus offset between the beam in tangential and axial directions). The FLIm UV beam is focused at 0.4 mm from the probe axis without astigmatism.

3.3.

Artery Results

Three sections of color-coded FLIm-OCT renderings are displayed in Fig. 4 with matching Movat’s pentachrome and CD68 stained sections. First, the penetration depth of OCT enables visualization of the whole vessel wall with well-resolved intima/media/adventitia layers as well as perivascular adipose tissue visible in normal and adaptive intimal thickening (AIT) regions [Figs. 4(a) and 4(b)].

Fig. 4

Imaging of human coronary specimen. FLIm/OCT B-scans where brightness corresponds to the OCT backscattered intensity whereas the Hue corresponds to the FLIm averaged lifetime for both (a) the 390-nm spectral band and (b) 540-nm spectral band. The OCT penetration depth enables the visualization of the full vessel wall in healthy regions, including some perivascular adipocytes. Movat’s pentachrome sections highlight that low 390-nm lifetime is associated with healthy regions (Normal, AIT), whereas lifetime increase is associated with PIT and early FA (eFA) regions (c). (d) CD68-stained sections highlight that increase in 540-nm lifetime corresponds to areas of superficial foam cells accumulation. 3D color-coded volumetric images enable the identification of high FLIm-contrast areas over the entire vessel segment. (e) The location of each B-scan (A/B/C) is represented by the cross-sectional planes. FLIm measurement variability was estimated by computing average lifetime mean and standard deviations for two homogeneous ROIs. ROI 1 (390-nm band): 3.82/0.15 ns. ROI 1 (540-nm band): 4.86/0.14 ns. ROI 2 (390-nm band): 5.21/0.13 ns. Mean FLIm waveforms (curve) and standard deviation (shaded area) from ROIs 1 and 2 demonstrate clear differences and high SNR (f) (Video 1, mp4, 6.15 MB [URL: https://doi.org/10.1117/1.JBO.27.7.076005.1]).

As expected, OCT attenuation is higher in pathological intimal thickening (PIT) and fibroatheroma (FA) regions. 390-nm lifetime contrast shows consistent low values in healthy and AIT regions, but increased lifetime values in PIT and FA regions, as expected based on earlier work¹² that associated increased 390-nm lifetime increase with presence of recently formed matrix (collagen, proteoglycans) often present in PIT and FA regions.

The vessel segment imaged here presents few locations of dense macrophage foam cells (mFC) accumulation in the vicinity of the lumen ($<150\mu \mathrm{m}$, expected penetration depth of FLIm), but locations of superficial mFC accumulation are well-matched with locations of increased 540-nm lifetime. Based on these observations, the FLIm spatial resolution is sufficient to obtain strong contrast even in presence of small mFC clusters (Section C, $<100\text{-}\mu \mathrm{m}$ wide). Conversely, no mFC could be identified in areas of low 540-nm lifetime, highlighting the specificity of the 540-nm contrast with respect to mFC detection. Finally, 3D renderings of FLIm-coded OCT [Fig. 4(e)] facilitate the identification of possible disease location over the whole artery segment. Clear separation between FLIm waveforms obtained from regions of interests (ROIs) 1 to 2 [Fig. 4(f)] illustrate the high SNR of the proposed direct-detection solid-state pulse-sampling approach, obtained without application of any spatial filtering. This high SNR translates into distinct lifetime values obtained in the 390-nm spectral bands in both normal regions (region 1, 3.82/0.15 ns mean/standard deviation) and lesion surfaces (region 2, 5.21/0.13 ns mean/standard deviation).