2.1.

Beta Detection System

The beta detection system is a catheter-based single-channel positron imaging unit prototyped as shown in Fig. 2 . It consists of a plastic scintillating fiber, clear light guide, photomultiplier tube (PMT), and PC-based multichannel analyzer (MCA). The plastic scintillating fiber is a $1\text{‐}\mathrm{mm}$ -diam Bicron BCF-12 blue-emitting fiber. The BCF-12 blue scintillator has an emission peak at $435\mathrm{nm}$ , decay time of $3.2\mathrm{ns}$ , 1∕e attenuation length greater than $2.2\mathrm{m}$ , and photon generation of $\sim 8000∕\mathrm{MeV}$ . The scintillating fiber was polished and glued to a clear light guide of the same diameter and $0.75\mathrm{m}$ lengths. The scintillating fiber and the clear light guide were connected using regular glass cement. The light attenuation of this fiber junction was measured to be $0.25\mathrm{dB}$ , which was minimal. The scintillating fiber was then cleaved to make a $2\text{‐}\mathrm{mm}$ tip, which compromises resolution and signal-to-noise ratio (SNR), as discussed in our earlier work.³² In the next prototype that we are planning to develop, the scintillating fiber will be fused with the regular optical fiber to reduce the coupling lose at the junction, and the distal surface of the scintillating fiber will be coated with reflective paint to further improve the scintillation collection. A light-tight rubber sleeve encloses the entire light guide, and the distal part of the light guide has an extra sleeve made with $120\text{‐}\mathrm{mm}$ -long 16-gauge stainless steel needle body. The stainless steel sleeve is mounted to a homemade three-way translation stage. The proximal end of the clear light guide is attached against the PMT window through a light-tight aluminum box, which encloses the PMT and a negative high-voltage power supply. The Hamamatsu R2949 PMT was selected for its high sensitivity, low dark current, and convenient side-on window. The wavelength of maximum response of R2949 is at $400\mathrm{nm}$ , well fitted for detecting the blue light emitted from the BCF-12 scintillating fiber. The PMT anode output is directly connected to the preamplifier input of a UCS-20 Spectrometer (Spectrum Techniques, Inc., Oak Ridge, TN). UCS-20 is a PC-based multichannel analyzer (MCA) featuring a maximum of 2048 channels. The communication between the UCS-20 and the computer is via the standard USB cable. UCS-20 includes an internal charge-sensitive preamplifier, computer-controlled linear amplifier for pulse shaping, lower level discriminator (LLD), upper level discriminator (ULD), analog-to-digital (A∕D) converter, and positive $0\text{to}2500\mathrm{V}$ high voltage power supply. The software provides three modes of photon counting: pulse height analysis (PHA), multichannel scaling (MCS), and Mossbauer.

Fig. 2

Schematic of the beta detector.

2.2.

OCT Imager

The details of the OCT imager [shown in Fig. 3(a) ] can be found elsewhere.²⁶ In brief, a superluminescent diode (SLD) is used as the low-coherence source. The SLD has a peak wavelength at $1300\mathrm{nm}$ , and spectrum bandwidth of $40\mathrm{nm}$ . The coherence length of this SLD is $18.6\mu \mathrm{m}$ . The reference arm consists of a grating-based time-domain scanning optical delay line³³ [not shown in Fig. 3(a)]. In the sample arm, we have designed a prototype rotating catheter probe $2\mathrm{mm}$ in diameter to perform intravascular circumferential OCT imaging [see Fig. 3(b) for the size of the probe]. In the detection arm, balanced configuration is used to suppress the source intensity noise.³⁴

Fig. 3

(a) Schematic of the OCT imager. The OCT is configured with balanced detection. (b) Picture of the prototype rotating OCT catheter probe. The structure of the probe is sketched beside the actual probe. A U.S. dime is placed nearby for size comparison.

We have designed and had OZ Optics fabricate a $1.6\text{‐}\mathrm{mm}$ -diam OCT catheter as shown in Fig. 3(b). The catheter body is composed of a flexible, rotary inner sleeve that fits loosely inside a homemade stationary outer sheath. The catheter body is made flexible to allow for bending during passage through the contours of blood vessels. The outer sheath allows the inner sleeve to rotate freely and provides a smooth exterior for passage through internal organ systems. The inner sleeve is connected to a FC∕PC connector at the proximal end and to the distal optics at the other end. The optical fiber is fixed within the center of the inner sleeve. Rotational torque exerted at the proximal end of the inner sleeve by the drive mechanism is transported to the distal end to rotate the beam focusing∕directing optical components.

The distal end of the catheter is composed of miniature beam focusing and directing optics. The single-mode fiber is attached to a gradient-index (GRIN) lens. A microprism is mounted at the distal surface of the GRIN lens to direct the beam perpendicular to the axis of the catheter. A homemade transparent cap is added on to enclose the distal end of the catheter for waterproof purposes. Due to the cap, the size of the catheter tip increases to $2.0\mathrm{mm}$ . The microprism, the GRIN lens, the inner sleeve, the optical fiber, and the proximal rotating optical connector are all attached with optical cement to form a single unit, the rotating shaft portion of the catheter. The GRIN lens is $1.0\mathrm{mm}$ in diameter, and the microprism has a $1\text{‐}\mathrm{mm}$ edge length. The predicted spot size is $30\text{to}35\mu \mathrm{m}$ at a working distance of $2\mathrm{mm}$ from the prism output surface. During image acquisition, the catheter is inserted into the blood vessel being imaged, and as the drive motor turns, the shaft of the catheter and the distal optics circumferentially scan the focused beam perpendicular to the axis of the catheter. The speed of imaging depends on the speed of the rotation and the OCT unit acquisition speed.

The light coupling mechanism from the interferometer to the rotary catheter is similar to that reported in Ref. 35. Incident light from a fixed single-mode optical fiber is coupled using an optical fiber FC∕PC adaptor into the single-mode fiber of the catheter. FC connectors mounting to the FC∕PC adaptor ensures precise alignment and optical coupling between the fixed and rotating fibers. The input fiber connector is fixed to the FC∕PC adaptor, while the locking notch of the catheter fiber connector is removed for free rotation. The drive assembly of the catheter is mounted with the FC∕PC adaptor enclosed. The FC∕PC adaptor is attached to a hollow rotary stage through a shaft assembly, and the rotary stage is connected to a dc motor through a timing belt pulley.

2.3.

Combined Beta-OCT Probe

We have constructed a coregistered beta∕OCT imaging probe. The OCT catheter passes into a light-tight box through a small hole, and the distal end of the catheter is confined side by side with the beta probe. The OCT catheter-driving assembly is mounted to a motorized linear stage, which renders image localization by pulling the combined beta∕OCT probe rather than moving the sample.³² Using this setup, we have acquired coregistered OCT and beta detection data from freshly excised pieces of bovine coronary arteries. Results are reported in Sec. 3(B).

For experiments using atherosclerosis rabbit models, we did not perform an intravascular study but modified the hybrid probe to scan the open vessel to preserve the artery for beta autoradiography. This ensured that the probe did not touch the endothelial layer. The combined beta∕OCT probe was positioned above the vessel and was aligned parallel to the long axis of the sample (see Fig. 4 ). The OCT probe, or the GRIN lens described in Sec. 2(B), was deployed vertically beside the beta probe. The OCT probe scanned transversely across the vessel, and the stationary beta probe detected the local beta activity. The sample was translated parallel to the scintillating fiber axis after each coregistered beta-OCT scan.

Fig. 4

The combined beta-OCT probe enclosed in a light-tight aluminum box.

Beta particle detection requires single-photon counting, where the ambient light level should be minimized. Therefore all of the light-sensitive components, including the scintillating fiber, light guide, and PMT, must be enclosed in a light-tight box. Because of the coregistered beta-OCT imaging configuration, the light-tight box must also accommodate the OCT probe. We have constructed a light-tight box, and repeated testing with the sealed source has shown that in using this light-tight box, the maximum counting and the background level show no significant difference for operation in a dark or light room.

2.4.

Rabbit Preparation

To validate the proposed hybrid intravascular OCT∕scintillator catheter, we have conducted experiments with an atherosclerotic rabbit model. New Zealand White (NZW) rabbits weighing $3\text{to}3.5\mathrm{kg}$ (Charles-River Laboratories, Wilmington, Massachusetts) were maintained on a high cholesterol (1% cholesterol) diet. One week following initiation of the high-fat diet, rabbits were anesthetized with ketamine $(30\mathrm{mg}∕\mathrm{Kg})$ and xylazine $(3\mathrm{mg}∕\mathrm{Kg})$ i.p., and a 5F sheath was inserted into the left carotid artery under fluoroscopy. An angioplasty balloon catheter ( $3.0\times 20\mathrm{mm}$ , Cordis, Johnson and Johnson Company) was passed over wire to the distal abdominal aorta, and was inflated to 10 atmosphere for $1\min$ . The abdominal artery was injured three times. Subsequently, the catheter was removed, the external carotid artery ligated, and the wound closed. Rabbits were continued on the hyperlipidemic diet for $6\text{months}$ before imaging experiments. Experiments were performed according to the regulations of Yale University’s Animal Care Committee.

3.1.

Testing and Calibration of the Beta Detection System

The performance of the OCT system has been demonstrated elsewhere.^{26, 27, 32} In this part, we present the testing and calibration results of the beta detection system.

FDG labeled with ${}^{18}\mathrm{F}$ is the most commonly used radiopharmaceutical in positron emission nuclear imaging. However, ${}^{18}\mathrm{FDG}$ is inconvenient for use as a calibration source due to its preparation, transportation, and relatively short half life for repeated calibration purposes. Instead, we have used a sealed $0.7\text{‐}\mu \mathrm{Ci}$ thallium TI-204 isotope with a half life of $3.77\text{years}$ for most calibration tests presented in this work.

Several critical system parameters related to the ex vivo and in vivo use of this scintillating catheter were evaluated experimentally. These parameters are signal-to-background ratio versus multichannel analyzer (MCA) lower level discriminator (LLD) setup, minimal counting time, detection sensitivity, and ratio of probe response to beta and gamma.

3.1.1.

Signal-to-background ratio versus MCA LLD setup

The PMT dark noise spectrum³⁶ measured with minimum LLD is shown in Fig. 5(a) (the solid curve). The dashed curve in Fig. 5(a) is the measured gross beta spectrum when the TI-204 source is placed $1\mathrm{mm}$ away from the beta probe. The pure beta spectrum (dotted curve) obtained by subtracting the background from the gross counting shows a continuous energy distribution up to a definite maximum³⁷ that corresponds to $765\mathrm{keV}$ for TI-204 beta emission.

Fig. 5

Determination of LLD for optimum SBR. (a) Spectrum of TI-204 beta emission. (b) The dependence of SBR and measurement error on LLD setup.

The signal-to-background ratio (SBR) may be calculated from the total counts of TI-204 source relative to the background counts. The dependence of SBR to the LLD of MCA is shown in Fig. 5(b) (solid curve). It is shown that as LLD increases, SBR reaches a maximum and starts to decrease. This indicates that there is an optimum LLD setup for maximum SBR. However, the counting statistics must be taken into account for setting up the optimum LLD. As the LLD increases, the total counts decrease, resulting in an increased relative probable error (RPE)³⁶ that is determined by $\mathrm{RPE}=\sqrt{N}∕N$ , where $N$ is the total photon count. The RPE curve in Fig. 5(b) (dot-dashed line) is inversely plotted to show the decrease in the measurement accuracy as LLD increases. It is apparent that there is a tradeoff between SBR and RPE at a lower LLD side of the SBR maximum.

TI-204 has maximum beta decay energy of $765\mathrm{keV}$ , while ${}^{18}\text{Fdeoxyglucose}$ has maximum positron emission energy of $635\mathrm{keV}$ . If the same system is used to measure the positron radiation from ${}^{18}\mathrm{FDG}$ , the beta spectrum maximum will shift to the lower side. The ${}^{18}\mathrm{FDG}$ spectrum can be approximated by squeezing the TI-204 spectrum from $765\text{to}635\mathrm{keV}$ . This approximation is intended for assisting the determination of optimum LLD in ${}^{18}\mathrm{FDG}$ -based photon counting. The SBR of ${}^{18}\mathrm{FDG}$ as a function of LLD is shown in Fig. 5(b) as the dashed curve below the TI-204 SBR. The optimal LLD for approximated ${}^{18}\mathrm{FDG}$ counting shifts to the lower side, which agrees with the fact that ${}^{18}\mathrm{FDG}$ beta emission has lower maximum energy compared to TI-204. The finalized LLD setup indicated by the vertical dashed line intersecting with the abscissa around 60 may provide a balance between SBR and RPE for both TI-204-based counting in calibration and ${}^{18}\mathrm{FDG}$ -based-counting in ex vivo and in vivo experiments.

Further tests were conducted to validate the dependence of SBR on the LLD setting. Shown in Fig. 6 are the photon counting results at different LLD levels by translating the TI-204 source parallel to the beta probe axis. The absolute counting results are shown in Fig. 6(a), while the normalized results are shown in Fig. 6(b). The curves from top to bottom correspond to gradually increased LLD. As LLD increases, the total counts at positions near and far from the source all decrease; nevertheless, the SBR improves gradually so that the source position is resolved better. The second dashed curve from the bottom in Fig. 6(b) corresponds to the optimal LLD determined in Fig. 5(b). It is worth mentioning that, in the source position-resolved test, each measurement curve provides a convolution between the point spread function of the $2\text{‐}\mathrm{mm}$ scintillating probe and the point spread function of the $4\text{‐}\mathrm{mm}$ source at a different LLD level.

Fig. 6

Photon counting results at different LLD levels by translating the TI-204 source parallel to the beta probe axis. (a) The absolute and (b) normalized counting results.

3.1.2.

Counting time

It is critical to reduce the counting time for in vivo use of the beta probe. However, a minimum counting time is needed for accuracy. Figure 7(a) shows the total events measured at different counting times from $1\mathrm{s}\text{to}2\min$ with the optimum LLD setting discussed early. Each data point was averaged for 10 repeated measurements. It shows excellent linearity between total counts and counting time. However, the counting accuracy degrades dramatically when the counting time is less than $10\mathrm{s}$ [see Fig. 7(b)]. When the counting time is not less than $30\mathrm{s}$ , the RPE remains relatively uniform. The position-resolved photon counting at counting times of 10 and $30\mathrm{s}$ , respectively, is compared in Fig. 7(c). Although the source can be resolved at both settings, the $10\text{‐}\mathrm{s}$ counting shows much larger measurement fluctuation than the $30\text{‐}\mathrm{s}$ counting. Based on these calibration tests with our current system, we choose $30\mathrm{s}$ as the counting time in the experiments thereafter.

Fig. 7

Determining the counting time with the measurement error taken into account. (a) Counts versus counting time. (b) Relative probable error versus counting time. (c) Localization of the source position at 10 and $30\mathrm{s}$ counting times.

3.1.3.

Detection sensitivity

System sensitivity was measured by placing aluminum absorbers with different density thickness over the TI-204 source. For beta emission of $E=765\mathrm{keV}$ by TI-204, the half-thickness of aluminum is $25.4\mathrm{mg}∕{\mathrm{cm}}^2$ based on Libby’s equation.³⁸ We have used aluminum absorbers with densities of 13.8 and $34.9\mathrm{mg}∕{\mathrm{cm}}^2$ to attenuate the $0.63\text{‐}\mu \mathrm{Ci}$ TI-204 source, resulting in source activities of $0.43\mu \mathrm{Ci}$ and $0.24\mathrm{Ci}$ , respectively. Shown in Fig. 8(a) are the average position-resolved results obtained from 10 sets of repeated measurements. Note that the activity of the $0.7\text{‐}\mu \mathrm{Ci}$ TI-204 source was reduced to $0.63\mu \mathrm{Ci}$ at the time when Fig. 8(a) was revised. In these tests, the source without∕with an absorber was placed $1\mathrm{mm}$ away from the beta probe and the source position can be clearly resolved for a $0.24\text{‐}\mu \mathrm{Ci}$ TI-source.

Fig. 8

Detection sensitivity test. (a) Sealed TI-204 source. (b) ${}^{18}\mathrm{FDG}$ .

The detection sensitivity was also verified with different doses of ${}^{18}\mathrm{FDG}$ [see Fig. 8(b)]. The ${}^{18}\mathrm{FDG}$ tests were done by manually pulling the catheter backward inside a vessel. The vessel was attached to tapered tubings on each side, and the catheter was inserted up to the distal tapered tubing. The tubings were connected to a plastic tube filled with saline water. By pulling back the catheter, we could maintain the vessel position and shape. Each time the catheter had been pulled back, we must carefully push the catheter back into the distal tubing. As shown in Fig. 8(b), three curves were aliened reasonably well in position by carefully controlling the initial position, and pulling steps that could vary for each set of measurements. As a result, a repeated measurement was not applicable for FDG tests.

3.1.4.

Ratio of probe response to beta and gamma

The reduction of the associated gamma background is essential for in vivo beta imaging, especially for imaging coronary arteries. The beta∕gamma ratio is measured from the SBR of TI-204 with and without the presence of gamma background. As shown in Fig. 9(a) , a $0.7\text{‐}\mu \mathrm{Ci}$ gamma source Cs-137 was placed $1\mathrm{mm}$ away from the beta probe at the opposite side of the TI-204 source. The activities of these gamma and beta sources are the same, and the ratio between gamma and beta energies is $662∕765$ , close to the $511∕635$ of ${}^{18}\mathrm{FDG}$ . From the SBG results with and without gamma background [see Fig. 9(b)], we have calculated the ratio of the detection response between beta and gamma, that is

$$\gamma =\frac{\text{Response}̱\text{beta}}{\text{Response}̱\text{gamma}}=\frac{\text{beta}\text{‐}\text{background}}{\text{gamma}\text{‐}\text{background}}.$$

The $\gamma$ is found to be between 10 and 13.

Fig. 9

The probe response to beta and gamma. (a) Testing setup. (b) SBR results.

3.2.

Coregistered Beta∕OCT Imaging Using Sealed Beta Source

Using the coregistered beta∕OCT imaging probe, we acquired coregistered OCT and beta detection data from freshly excised pieces of bovine coronary arteries. The sealed beta source is positioned close to a $3\text{‐}\mathrm{mm}$ -diam hole left at the artery. The examined artery is securely positioned, and an OCT cross section image and radiation counting are acquired simultaneously after each $1\text{‐}\mathrm{mm}$ displacement of combined probe. The acquired coregistered beta∕OCT images shown in Fig. 10 are quite similar to those reported in Ref. 32, whereas the beta detection resolution is improved significantly due to shorter length of the scintillating tip and optimized counting parameters.

Fig. 10

Coregistered beta∕OCT imaging. Top curve: beta counting. Bottom image locations: top row for 1, 5, 9, 13; row two for 2, 6, 10, 14; row three for 3, 7, 11, 15; and row four for 4, 8, 12, 16.

3.3.

Coregistered Beta∕OCT Imaging Result Obtained from Injured Rabbit Models

The injured rabbit was injected intravenously with ${}^{18}\mathrm{FDG}$ $\left(3.5\mathrm{mCi}\right)$ . After one hour, the rabbit was euthanized and the entire abdominal aorta and iliac arteries were removed, cut open, and fixed to the board. Several thin pieces were cut fresh before the combined imaging for histological analysis. A photograph of the sample is shown in Fig. 11(a) . The scintillating fiber tip is positioned above the sample and is aligned parallel to the long axis of the sample. The OCT probe, or the GRIN lens, is deployed vertically beside the beta probe. The OCT probe scans transversely across the vessel, and the stationary beta probe detects the local beta activity. The sample is translated parallel to the scintillating fiber axis after each beta-OCT scan. Since the half-life of ${}^{18}\mathrm{FDG}$ is short, the radiation detection was conducted first at a step of $0.8\mathrm{cm}$ and with $30\text{‐}\mathrm{s}$ counting at each position. The total beta scanning time for the entire vessel was about $8\min$ . After beta scanning, the sample was repositioned to the origin by moving the stage, and coregistered OCT scans were acquired in $0.4\text{‐}\mathrm{cm}$ step. At the gap positions where no sample was underneath, OCT images were acquired in $0.2\text{‐}\mathrm{cm}$ step.

Fig. 11

(a) Photograph of the artery. (b) Detection curve obtained from the coregistered scintillating probe. The first peak from the left (marked with shot arrow) corresponds to the injured site, and the second peak from the left corresponds to the extensive atherosclerotic areas with high lipid contents. First point at $0\mathrm{cm}$ and last point at $11\mathrm{cm}$ are outside the vessel. (c) Beta autoradiography obtained with overnight exposure. (d) The mean intensity per pixel of the autoradiography.

The detection curve obtained from a coregistered scintillating probe is shown in Fig. 11(b). The first point at $0\mathrm{cm}$ and the last point at $11\mathrm{cm}$ are outside of the vessel sample. The first peak from the left (marked with a short arrow) corresponds to the injury site, and the second peak from the left corresponds to the extensive atherosclerosis area with high lipid content (pointed by an arrow). The beta autoradiography obtained with overnight exposure is shown in Fig. 11(c). The two dark regions pointed by the white arrows from left to right correspond to the site of injury and the area of extensive atherosclerosis. The autoradiography was scanned on a high-resolution scanner (Seiko Epson Corporation). The mean intensity per pixel of the autoradiography was measured using Kodak 1D software (Eastman Kodak Company, Rochester, New York) at regular intervals along the artery, background corrected, and expressed in arbitrary numbers as shown in Fig. 11(d). Linearity of the signal was confirmed using serial dilutions of the tracer. The mean intensity curve quantitatively agrees with the detection curve shown in Fig. 11(b). The coregistered sequence of corresponding OCT images is shown in Fig. 12 . In the site of injury, the lipid shadowing is pronounced, as pointed by white arrows. In the area with extensive atherosclerosis, the lipid shadowing appears again, as pointed by white arrows, and the fibrocalcific aortic plaque with very low light scattering or signal-poor interior is circled in the last figure. Similar signal-poor regions with sharply delineated upper and∕or lower borders were reported in Ref. 15 with samples obtained from a human autopsy. Representative histological slices and OCT images obtained from the sites of injury and the extensive atherosclerosis are shown in Figs. 12(b) and 12(c), respectively. Histological slices were stained with Elastic Van-Gieson (EVG) at 200-fold magnification. At both the injury site and the extensive atherosclerosis regions, the intima layer was thickened and OCT images showed significant lipid shadowing.

Fig. 12

Coregistered OCT images numbered from left to right. The spatial dimension from top to bottom is $4.8\mathrm{mm}$ and the depth is $2\mathrm{mm}$ . The first group of white arrows in the $3.6\text{‐}\mathrm{cm}$ region points to lipid shadowing regions corresponding to the injured site. The second group of white arrows in the $9.6\text{‐}\mathrm{cm}$ region points to lipid shadowing regions corresponding to extensive lipid content area toward the aorta. In the last figure the fibrocalcific aortic plaque is circled. (b) Representative histology ( $\times 200$ magnification) and OCT image obtained from the injured site. Elastic fibers under Elastic Van-Gieson (EVG) staining appear dark and the elastic membrane was thickened compared with noninjured sites. The OCT image showed lipid shadowing with less light penetration. (c) Representative histology and OCT image obtained from extensive plaque regions. The OCT image showed significant lipid shadowing with less light penetration.