Positron emission tomography with additional c-ray detectors for multipletracer

Transcription

1 Positron emission tomography with additional c-ray detectors for multipletracer imaging Tomonori Fukuchi, a) Takashi Okauchi, and Mika Shigeta RIKEN Center for Life Science Technologies, Kobe , Japan Seiichi Yamamoto Department of Radiological and Medical Laboratory Sciences, Nagoya University Graduate School of Medicine, Nagoya , Japan Yasuyoshi Watanabe and Shuichi Enomoto RIKEN Center for Life Science Technologies, Kobe , Japan (Received 21 September 2016; revised 27 January 2017; accepted for publication 29 January 2017; published 4 May 2017) Purpose: Positron emission tomography (PET) is a useful imaging modality that quantifies the physiological distributions of radiolabeled tracers in vivo in humans and animals. However, this technique is unsuitable for multiple-tracer imaging because the annihilation photons used for PET imaging have a fixed energy regardless of the selection of the radionuclide tracer. This study developed a multiisotope PET (MI-PET) system and evaluated its imaging performance. Methods: Our MI-PET system is composed of a PET system and additional c-ray detectors. The PET system consists of pixelized gadolinium orthosilicate (GSO) scintillation detectors and has a ring geometry that is 95 mm in diameter with an axial field of view of 37.5 mm. The additional detectors are eight bismuth germanium oxide (BGO) scintillation detectors, each of which is mm 3, arranged into two rings mounted on each side of the PET ring with a 92-mminner diameter. This system can distinguish between different tracers using the additional c-ray detectors to observe prompt c-rays, which are emitted after positron emission and have an energy intrinsic to each radionuclide. Our system can simultaneously acquire double- (two annihilation photons) and triple- (two annihilation photons and a prompt c-ray) coincidence events. The system s efficiency for detecting prompt de-excitation c-rays was measured using a positron-c emitter, 22 Na. Dual-radionuclide ( 18 F and 22 Na) imaging of a rod phantom and a mouse was performed to demonstrate the performance of the developed system. Our system s basic performance was evaluated by reconstructing two images, one containing both tracers and the other containing just the second tracer, from list-mode data sets that were categorized by the presence or absence of the prompt c-ray. Results: The maximum detection efficiency for 1275 kev c-rays emitted from 22 Na was approximately 7% at the scanner s center, and the minimum detection efficiency was 5.1% at the edge of the field of view. Dual-radionuclide imaging of the point sources and rod phantom revealed that our system maintained PET s intrinsic spatial resolution and quantitative nature for the second tracer. We also successfully acquired simultaneous double- and triple-coincidence events from a mouse containing 18 F-fluoro-deoxyglucose and 22 Na dissolved in water. The dual-tracer distributions in the mouse obtained by our MI-PET were reasonable from the viewpoints of physiology and pharmacokinetics. Conclusions: This study demonstrates the feasibility of multiple-tracer imaging using PET with additional c-ray detectors. This method holds promise for enabling the reconstruction of quantitative multiple-tracer images and could be very useful for analyzing multiple-molecular dynamics American Association of Physicists in Medicine [ Key words: dual-radionuclide imaging, positron emission tomography, positron gamma emitter, small animal imaging 1. INTRODUCTION Positron emission tomography (PET) is a powerful tool for noninvasive imaging of the distribution of a tracer in living organisms. However, conventional PET is restricted to singletracer imaging because it is based on the detection of positron annihilation photons, which always have an energy of 511 kev regardless of the variety of the positron emitter. If PET could image multiple tracers, it would be a much more powerful tool. In previous work, several methods have been proposed for multiple-tracer imaging using PET, such as using the difference between the half-lives of the tracers radionuclide decays or the detection of the prompt c-rays emitted after positron emission. A method using the difference in tracer decay times is not actually simultaneous multiple-tracer imaging because all such methods either assume a static tracer distribution or estimate the tracer dynamics based on pharmacokinetics. 1,2 On the other hand, a method that detect prompt c-rays, a multiisotope PET (MI-PET), is genuine simultaneous multiple Med. Phys. 44 (6), June /2017/44(6)/2257/ American Association of Physicists in Medicine 2257

2 2258 Fukuchi et al.: PET for multiple tracer imaging 2258 tracer imaging and has recently been actively studied. 3 6 A method using MI-PET combined with the administration of two tracers at different times to increase the imaging performance has also been proposed by Andreyev et al. 7 In addition to the development of a MI-PET scanner, the production of specific radionuclides for MI-PET has also been investigated. 8 SPECT also works for multiple-tracer imaging based on c-ray energy separation. However, it is still desirable to accomplish multiple-tracer imaging using PET, because it has higher sensitivity and quantitativity than other modalities including SPECT, and also because PET has its own unique tracers, such as 18 F-fluorodeoxyglucose (FDG). Additionally, although the sensitivity for the second tracer in MI-PET is lower than that of the conventional PET, MI-PET still possesses high sensitivity for both tracers. In the MI-PET method, the PET detector itself works as the prompt c-ray detector; however, employing additional detectors enhances the detection efficiency. 4 6 To demonstrate MI-PET s feasibility, we developed a prototype MI- PET system that is composed of a conventional circular ring PET and additional detectors to detect prompt c-rays. The PET system and the additional detectors consist of gadolinium orthosilicate (GSO) pixel detectors and eight largevolume ( mm 3 ) bismuth germanium oxide (BGO) detectors, respectively. There is another method for only detecting prompt c-rays using PET detectors. 3 However, the detection efficiency of PET detectors is limited, especially for high-energy c-rays, because PET detectors are designed for identifying 511 kev annihilation c-rays. Therefore, to maintain the sensitivity for high-energy c-rays, our system uses a ring configuration for the additional c-ray detectors, but other detector configurations are also possible, such as the end-cap configuration proposed by Levin et al. 5 Some of the main considerations of our configuration are the accessibility and ease of animal handling on a moving bed. The uniformity of the sensitivity for prompt c-rays in a field of view (FOV) is also considered important for normalizing the second tracer. However, further study is required to optimize the configuration of additional detectors in MI-PET. In addition to conventional double coincidence between the PET detectors, triple coincidence among the additional detectors and the PET system can be performed. Using a positron-c emitter as the second tracer, tracer identification can be made using additional detectors that locate prompt c- rays whose energy is characteristic of the radionuclide. We expect that MI-PET will be used in various new applications that are impracticable with single-tracer imaging, such as simultaneously performing multiple diagnostic scans, analyzing time sequences from multiple tracers and the effect of additional dosing with the same drug, and validating drug delivery systems by dual labeling. In this study, we investigated dual-radionuclide imaging using our newly developed MI-PET system to evaluate its imaging performance. To evaluate its basic performance, images of both tracers and just the second tracer were reconstructed from list-mode data sets that categorized data based on whether prompt c-rays were detected. 2. MATERIALS AND METHODS 2.A. Principles The positron emitters employed in PET imaging can be categorized into two types: pure positron emitters and positron-c emitters. Figure 1 shows two typical b + -decay schemes. Decay type (A) is a pure positron emitter. In such of decay, the nucleus undergoes a transition to the ground state of a daughter nucleus through b + -decay. Then, only a positron (with a neutrino) is emitted in the b + -decay. In contrast, decay type (B) is a positron-c emitter. In this type of decay, the nucleus undergoes a transition to an excited state of the daughter nucleus through b + -decay. Then, a c-ray is emitted after positron emission to transfer the energy from the excited state to the ground state. This c-ray is called a deexcitation c-ray, which has characteristic energy for each radionuclide. In some nuclides, the de-excitation c-ray is emitted with some delay from the positron emission because the nuclide goes through the isomeric state. Therefore, we use the term prompt c-ray to indicate a promptly emitted de-excitation c-ray to distinguish them from the annihilation c-rays that are produced by positron-electron pair-annihilation. MI-PET distinguishes between different tracers based on the detection of prompt c-rays. In dual-radionuclide imaging, it is possible to use a pure positron emitter and a positron-c emitter and distinguish some decay between these radioisotopes by the presence or absence of prompt c-rays. By using more than two positron-c emitters and detecting the energy of the prompt c-rays, more than three radioisotopes can be imaged simultaneously. To be used in a MI-PET, a positron-c emitter is required to have such particular characteristics, as an adequate lifetime and a large enough positron and prompt c-ray emissions per decay. Considering all of the required characteristics, the suitable radionuclides, which can be used as a positron-c emitter for MI-PET, are shown in Table I. Here, the e + ratio is the probability of the positron emission per decay, and the prompt c-ray ratio is the probability of prompt c-ray emission after positron emission 9. This table contains 22 Na whose lifetime is too long for clinical use, but since it is useful for phantom or animal imaging, we used it FIG. 1. Decay schemes for two types of b + -decay: (a) pure positron emitter and (b) positron-c emitter. Z and Z-1 are the atomic numbers of the radionuclides.

3 2259 Fukuchi et al.: PET for multiple tracer imaging 2259 TABLE I. Candidate radionuclides for use as a positron-c emitter in MI-PET imaging. Nuclide Half-life e + Ratio (%) c Energy (kev) Prompt c ratio (%) 14 O s Na yr m Cl min K min Sc 3.97 h V days Mn days m Mn 21.1 min Fe h Cu 23.7 min As 26.0 h Br 16.2 h m Rb h Y h m Tc 52.0 min m In 69.2 min I days in this study. For multiple prompt c-ray emitters, only prompt c-ray energies with strong intensities are noted. 2.B. System description Figure 2 shows a photograph and a schematic illustration of the developed system. The prototype is composed of a PET system and additional c-ray detectors. The PET system has a ring geometry and consists of pixelized GSO scintillation detectors. The additional detectors are eight BGO scintillation detectors. The PET system is based on the open-close PET developed by Yamamoto et al. 10 which can change the scanner configuration between a closed mode (ring configuration) and an open mode (two separated semicircle configuration). In this system, the detectors are a combination of mm 3 and mm 3 GSO pixel scintillation detectors with fast (35 ns) and slow (60 ns) decay times, respectively. These two types of GSO detectors were optically coupled to determine the depth of interaction from the differences in decay times. These GSO pixels are arranged into a matrix with a 0.1-mm-thick BaSO 4 reflector at a pitch of 1.7 or 2.5 mm. To configure a ring geometry, two scintillator matrices were arranged on a flat panel photomultiplier tube (FPPMT; H8500, Hamamatsu) at an angle of The energy resolution of the PET GSO detector was 28.4% (FWHM) for a 511 kev c-ray. 11 The details of the structure and the performance of this angled geometry and contact with FPPMT were described by Okumura et al. 11 Eight FPPMTs with scintillator blocks were arranged in a hexagonal shape to form a ring-like PET system whose inner diameter was 95 mm and whose axial FOV was 37.5 mm. We used the closed mode in this study. For the detection of prompt c-rays, we added eight BGO scintillation detectors with dimensions of mm 3. All were optically coupled to a 2 in. square photomultiplier tube. The measured energy resolution of a BGO detector was 10.3% (FWHM) for 1275 kev c-rays. Two rings with a 92-mm inner diameter were formed using four BGO detectors per ring and mounted on each side of the PET ring with an axial distance of 46 mm between the centers of the PET and BGO detectors. For the signal processing, we used the same front-end electronics for the block detectors as those used for the position-sensitive photomultiplier tube in a brain PET FIG. 2. Photograph (left) and schematic illustration (right) of developed MI-PET system. [Color figure can be viewed at wileyonlinelibrary.com]

4 2260 Fukuchi et al.: PET for multiple tracer imaging 2260 system developed by Yamamoto et al. 12 The system was modified to measure the triple coincidence between the PET and the BGO detectors in addition to the PET double coincidence for the annihilation photons. This triple-coincidence measurement can be performed in parallel with the PET double-coincidence measurement with list-mode data acquisition. To distinguish between double- and triple-coincidence data, a tag was recorded that indicate triple coincidence. The width of the coincidence time window was set to 20 ns for both double and triple coincidence. Although this time window is wide for the intrinsic time resolution of the GSO and BGO detectors, it is set to correspond to the time jitter originating from the energy variation of the leading edge timing and the individuality of the timing performance in each detector. For the PET detectors, the lower energy threshold level was set to 350 kev, and the upper energy threshold level remained open. In contrast, the lower- and upper-level thresholds of the energy window for the additional BGO detectors were set to 1.0 and 1.5 MeV to focus on the 1275 kev c-rays originating from the 22 Na used in this study. The energy window of the BGO detectors was set to exclude the 511 kev annihilation c-rays. Consequently, in addition to the photo-peak events, some parts of the Compton scattering events of the 1275 kev c-rays, which deposit part of their initial energy, were also detected. Therefore, the efficiency of the BGO detectors includes these Compton scattering events. However, because the PET detectors have open thresholds for the upper energy limit, they also detect 1275 kev c-rays, if the energy upper level is set, although the Compton scattering events of the 1275 kev c-rays were still detected. These events are the same as the random coincidence events in a conventional PET and can be corrected using the delayed coincidence method. Thus, in our system, only energy discrimination is used to distinguish prompt and annihilation c-rays, although another method could distinguish between them based on the collinearity of annihilation photons. 8,13 2.C. Image reconstruction The list-mode coincidence data are classified into two groups based on the presence or absence of prompt c-rays. After the classification, a conventional image reconstruction algorithm is executed for each set of the data. In our system, first, the projection data are produced using single-slice rebinning with a maximum ring difference of eight. Image reconstruction from the projected sinogram was performed using inter-update Metz-filtered ordered subset expectation maximization (IMF-OSEM) 14 in a program called Software for Tomographic Image Reconstruction (STIR). 15 The resulting image had a size of voxels with a voxel size of mm 3. For the IMF-OSEM, the subset number was fixed to be four. The images reconstructed from the data without and with the prompt c-rays coincidence are called image-d (double coincidence) and image-t (triple coincidence). Based on the delayed coincidence, the data random coincidence background was subtracted. The random coincidence between the annihilation photons and the prompt c-ray, which causes biased noise reflecting the distribution of the pure positron emitter, was also subtracted using the delayed coincidence data. Perhaps triple coincidence could be generated by scattering photons between the two sets of detectors, 16,17 but having a lower level energy threshold of 1.0 MeV for the BGO detectors, which is approximately double the energy of the 511 kev annihilation photons, removes most of this scattering from the data. From the measurements, triple-coincidence events occur very rarely (below 0.1%) when using just a pure positron emitter 18 F with activity of ~1 MBq. However, in dual-tracer experiments, random coincidence events were detected between the annihilation c-rays from either the pure positron emitter or the positron-c emitter and the prompt c-rays. The random coincidence events between the annihilation c-rays from the pure positron emitter and the prompt c-rays from the positron-c emitter produced biased noise reflecting the distribution of the pure positron emitter in the image reconstructed from triple-coincidence events where only the positron-c emitter should appear without any correction. This biased noise can also be subtracted using a delayed coincidence method. In a mouse experiment using 18 F (810 kbq) and 22 Na (410 kbq) in Section 3.C, the delayed coincidence fraction of the true triple coincidence was 14.1%. 2.D. Image normalization In MI-PET, normalization of the positional dependence of the prompt c-ray detection efficiency is necessary in addition to the conventional image normalization of PET. Therefore, a long-period scan was performed using a cylindrical 22 Na phantom (diameter: 78 mm; length: 180 mm). Sodium-22 emits both a positron and a 1275 kev c-ray (Table I). The phantom had 1 MBq of total activity, which corresponds to 1.1 kbq/ml. A total of double- and triple-coincidence events were recorded over 87 h of measurement. Projection data were produced for each set of data and used for the normalization of the double- and triple-coincidence images. This normalization measurement was for the 1275 kev c-rays of 22 Na used in this study; if another radionuclide is used as a tracer, a specific measurement is necessary for normalization with it. If the measurement is difficult, some simulations or interpolations for different energies can be executed instead. In this study, c-ray absorption in the imaged objects was ignored, because the phantom and animal had small volumes. 3. PERFORMANCE EVALUATION 3.A. Point sources To evaluate the prompt c-ray detection efficiency and the spatial resolution for dual-radionuclide imaging, we made a 0.3-mm-diameter 22 Na point source with 42 kbq activity.

5 2261 Fukuchi et al.: PET for multiple tracer imaging A.1. Detection efficiency for prompt c-rays The positional dependence of the prompt c-ray detection efficiency was measured in the transverse (x-axis) and sagittal (z-axis) directions in 2.5-mm increments from the scanner s center. The detection efficiency of the triple coincidence (P t ) in each position is defined as P t ¼ P d P g ; where P d is the detection efficiency of the double coincidence for the annihilation c-rays and P g is the prompt c-ray detection efficiency. P g is the ratio of the double and triple coincidences: Pg ¼ P t =P d : We must consider the emission fraction of prompt c-rays, but 22 Na, which was used for efficiency measurements, emits a prompt c-ray after positron emission 99.9% of the time (Table I), and when PET detects annihilation photons generated by positron emission, a prompt c-ray is also emitted at the same time. Then, efficiency P g is the same as the absolute efficiency of the prompt c-ray detection in the first approximation. The acquisition time for each measurement was 5 min. We also measured the positional dependence along a line in the x-y plane at 45 to the x- and y-axes (the x-y direction for brevity). This radial direction differs from the x-direction, since all the BGO detectors are mounted, so they point in a direction 45 to the x- and y-axes. To obtain a global view, the prompt c-ray detection efficiency on the x-z plane was also measured on a 5-mm mesh. These measured points are depicted in Fig. 3. In addition to the measurement using 22 Na, to investigate the energy dependence of the prompt c-ray efficiency (P g ), we also measured the efficiency of the BGO detectors using single c-ray emitters 54 Mn (835 kev) and 137 Cs (662 kev). 3.A.2. Dual-radionuclide imaging To evaluate the spatial resolution of the second tracer ( 22 Na, positron fraction 90.4%, Table I) when used in combination with another positron emitter, we prepared an 18 F (positron fraction 96.7%) point source in addition to the 22 Na point source. The 18 F point source had a diameter of 0.8 mm with 47 kbq activity (comparable with the activity of the 22 Na point source) at the start of the measurement. The two point sources were arranged along the x-axis at 20-mm intervals in three arrangements: (a) 22 Na at the center, (b) the two point sources equidistant from the center, and (c) 18 F at the center. With respect to the z-axis, the point sources were arranged in the same way as on the x-axis with a 10-mm interval (arrangements (d)-(f)). These arrangements are shown in Fig 4. Measurements were performed for 5 min in each arrangement. After image reconstruction using three-dimensional filtered backprojection, the spatial resolutions for the second tracer ( 22 Na) were evaluated for each source arrangement. 3.B. Rod phantom To make a quantitative assessment of the imaging ability of the MI-PET system, we scanned a dual-radionuclide rod phantom that had three cylindrical rods with a diameter of 10 mm and a length of 76 mm. Fluorine-18 and 22 Na dissolved in water were poured into the rods. The first rod had 18 F (686 kbq at the start of the measurement) and 22 Na (585 kbq) activities. The second and third rods had 18 F (636 kbq at the start of the measurement) and 22 Na (585 kbq) activities, respectively. The three rods were arranged in an equilateral triangle at a distance of twice the rod diameter. The rods were positioned at the scanner s center parallel to the scanner s axial FIG. 3. Measurement positions for prompt c-ray detection efficiency using a 22 Na point source. Z-axis is axial direction, and x- and y-axes are transverse horizontal and vertical directions, respectively. The line labeled x-y lies in the x-y plane. The origin of coordinate axes is the scanner s center. FIG. 4. Arrangements of point sources for dual-radionuclide imaging. (a) (c) and (d) (f), respectively show x-axis (transverse) and z-axis (axial) arrangements.

6 2262 Fukuchi et al.: PET for multiple tracer imaging 2262 FIG. 5. Illustration of rod phantom configuration. Three rods with activity 18 F (636 kbq) + 22 Na (585 kbq), 18 F (636 kbq), and 22 Na (585 kbq) were called rods 1, 2, and 3, respectively. [Color figure can be viewed at wileyonlinelibrary.com] direction. The configuration and activities of the rod phantom are shown in Fig. 5. A 30-min scan was performed for the double and triple coincidences in parallel. For each of the three rods, the regions of interest (ROIs) were chosen to make a quantitative comparison between the original activities and those deduced from the PET images. FIG. 6. Axial (z-axis) positional dependence of efficiencies of additional detectors for 1275 kev c-rays of 22 Na. 3.C. Small animal study To test the practical performance of the developed MI- PET system, we carried out small animal imaging. In this experiment, 410 kbq 22 Na and 810 kbq 18 F-FDG were administered to an 8-week-old normal male mouse by oral administration and tail vein injection, respectively. Ten minutes after administration, a 30-min scan with bed motion was performed under anesthesia. Image reconstructions using the IMF-OSEM were carried out for data with and without the detection of the prompt c-rays. This animal experiment was performed in accordance with the Principles of Laboratory Animal Care (NIH Publication No , revised 1985) and approved by the Institutional Animal Care and Use Committee (IACUC) of RIKEN, Kobe Branch. FIG. 7. X- and x-y direction dependence of efficiency of additional detectors for 1275 kev c-rays of 22 Na. 4. RESULTS 4.A. Point sources 4.A.1. Prompt c-ray detection efficiency In the measurements of prompt c-ray detection efficiency, approximately 400,000 events were accumulated at the center position in 5-min measurements. Accumulated event numbers at the other positions were of the same order. The prompt c-ray detection efficiency in each position (P g ) contains accidental coincidence events between the annihilation photons and the prompt c-rays. However, the rate of accidental coincidence calculated using the coincidence window (20 ns) of the PET and the additional detectors count rate (PET: 450 cps; BGO: 3000 cps) in these measurements were less than 0.03 cps, which is negligible. The prompt c- FIG. 8. Measured efficiency of additional detectors for 1275 kev c-rays of 22 Na in x-z plane with a 5-mm mesh. ray detection efficiency for a 1275 kev c-ray at the center position was approximately 7%. Figures 6 and 7 show the axial (z-axis) and transverse (xaxis) dependence of the prompt c-ray detection efficiency. Figure 8 shows the prompt c-ray detection efficiency on the x-z plane with a 5-mm mesh. It monotonically decreased in both

7 2263 Fukuchi et al.: PET for multiple tracer imaging 2263 FIG. 9. Examples of dual-radionuclide images and profiles for point sources in arrangement 2 (a) and arrangement 4 (b). Images and profiles are center slices of image-d (left) and image-t (right). These images have a voxel size of mm 3. the x- and z-directions as the distance from the scanner s center increased, and so a position at the edge of the FOV should have the lowest prompt c-ray detection efficiency. The ratio of the maximum detection efficiency (7.0% at the scanner s center) and the minimum detection efficiency (5.1% at the edge of the FOV, x = 30 mm, z = 15 mm) was about From the measurements for energy dependence, the c-ray photo-peak efficiencies for 54 Mn (835 kev) and 137 Cs (662 kev) in the center position were 7.8% and 9.4%, respectively. 4.A.2. Dual-radionuclide imaging In dual-radionuclide imaging, the total accumulated numbers of double- and triple-coincidence events for the layout shown in Fig. 4 (a) were and , respectively. The double-coincidence number includes events originating from the annihilation c-rays from the pure positron emitter ( 18 F). For the other layouts in Figs. 4(b) (f), the total accumulated counts were approximately the same as those for the layout in Fig. 4(a). Figure 9 shows a slice image of the 18 F and 22 Na point sources. Image-D (left) and image-t (right) were reconstructed from data without and with the detection of the prompt c-rays, respectively. The upper and lower images are those of point sources placed along the x- and z-axes, respectively, equidistant from the scanner s center (Figs. 4(b) and 4(e)). In both layouts, we can clearly observe the 18 F and 22 Na point sources in the image-d and the 22 Na point source in image-t. Table II summarizes the spatial

8 2264 Fukuchi et al.: PET for multiple tracer imaging 2264 TABLE II. Spatial resolutions for a 22 Na point source in image-d and image-t. Layout Axis x x x z z z Image-D (mm) Image-T (mm) Values are full widths at half-maximum in millimeters. Values in layouts 1 3 are x-y plane spatial resolutions and those for layouts 4 6 are z-projection spatial resolutions. Layouts of 18 F and 22 Na point sources are shown in Fig. 4. resolutions of the 22 Na point source in image-d and image-t in the presence of another radionuclide ( 18 F). The spatial resolutions for the arrangements in Figs. 4(a) (c) in image-t are for the x-y plane, and those for the arrangements in Figs. 4(d) (f) are for the z projection. At the center, the x-y spatial resolution in image-t of the layout in Fig. 4(a) was 2.6 mm and the z-projection resolution was 2.4 mm. 4.B. Rod phantom In the rod phantom measurement, the count rate at the starting time was 15.5 kcps double- and triple-coincidence events were acquired. The double-coincidence number includes events originating from the annihilation c-rays from the pure positron emitter ( 18 F). Images were reconstructed using IMF-OSEM with six iterations and four subsets. In this image reconstruction, image normalization and delayed background subtraction were also executed. Figure 10 shows the x-y plane projection images of the rod phantom in image-d (left) and image-t (right). In both images, we can clearly observe the 18 F and 22 Na activities with reasonable intensities. Comparisons between ROIs and original activities are shown in Table III. The ROIs were set to the whole region of each rod in both images. The values in image-d and image-t were normalized by the measured total activity of rod 1 in image-d. From these values, the original activities of both radioisotopes were well reconstructed in both image-d and image-t. The ROI activity of rod 2 in image-t, which does not have 22 Na, is small. This nonzero activity is conceivably due to the accidental coincidence between the annihilation photons of 18 F and the 1275 kev c-rays of 22 Na that are TABLE III. Measured rod activities for image-d and image-t in dual-radionuclide rod phantom. Layout Rod 1 ( 18 F+22Na) Rod 2 ( 18 F) Rod 3 ( 22 Na) Image-D Image-T Activities were normalized by total activity of rod 1 in image-d. not fully removed by the delayed background correction. However, the amount of remaining activity is comparable to the error of the activities in other ROIs. The activities in every slice of the ROIs were also evaluated to check the uniformity along the z-axis. Figures 11 and 12 show the activities in each z slice for image-d and image-t, respectively. The nonuniformity of image-d was in the range of a few percent and this value is sufficiently small. The nonuniformity of image-t is rather large compared with that in image-d, but it is also acceptable. 4.C. Small animal study In the mouse experiment, the initial double- and triplecoincidence count rates were 9.7 kcps and 340 cps, leading to a total of accumulated events in a 30-min scan of FIG. 11. Uniformity of measured rod activities in z-axis slices of image-d. FIG. 10. X-y plane projection of rod phantom in image-d (left) and image-t (right).

9 2265 Fukuchi et al.: PET for multiple tracer imaging 2265 FIG. 12. Uniformity of measured rod activities in z-axis slices of image-t and counts, respectively. The doublecoincidence number includes events originating from the annihilation c-rays from the pure positron emitter ( 18 F). Images were reconstructed using IMF-OSEM with six iterations and four subsets. Figure 13 shows the results of dualradionuclide small animal imaging. We can clearly see 22 Na in the esophagus and stomach, whereas 18 F-FDG is distributed in the brain, heart, kidneys, and urinary bladder. 5. DISCUSSION For prompt c-ray detection efficiency, the sensitivity of the open-close PET to annihilation photons decreases proportionally with the distance from the scanner s center 10. In contrast, the sensitivity of the additional detectors for the prompt c-rays also decreases proportionally with the distance from the scanner s center but more drastically. This tendency may reflect the fact that the additional detectors on the far side are shadowed by the PET detectors. The transverse prompt c-ray detection efficiency in Fig. 7 also decreases proportionally with the distance from the scanner s center, but its variation is slower than the variation in the axial position. This difference is presumably due to the fact that the prompt c-ray detectors on both sides are not shadowed in this case. The detection efficiency along the slanting line in the x-y plane in Fig. 7 decreases more slowly than it does along the x-axis. This can be explained by the fact that the x-y direction is parallel to the axis of one of the additional BGO detectors. The minimum prompt c-ray detection efficiency of approximately 5% indicates that it takes at least 20 times longer to obtain a second tracer image with comparable quality to conventional PET imaging. However, Cal-Gonzalez et al. showed that the triple-coincidence fraction in a small animal PET (Argus small animal PET/CT scanner) for 22 Na (1275 kev) is about 2.5%. 17 Because the PET detectors were designed for 511 kev annihilation c-rays, the detection efficiency for highenergy c-rays is lower. Furthermore, the prompt c-ray efficiency of the Argus scanner for 76 Br (559 kev) is smaller than that of our system s 9.4% for 137 Cs (662 kev). The spatial resolution of the second tracer is comparable to the intrinsic open-close PET scanner performance of 2.4- mm full width at half-maximum of the z-projection image in the center position 10. The spatial resolutions away from the center also behave in the same manner as in single-tracer imaging with an open-close PET. As a result, we found that the spatial resolutions for the second tracer in dual-radionuclide imaging are comparable with the scanner s intrinsic spatial resolutions. This means that the first tracer does not affect the spatial resolution of the second positron-c emitter in a reconstructed image with prompt c-ray detection (image-t). The spatial resolution of image-t probably only depends on the accumulated number of events. In the rod phantom measurement, we reconstructed the original activity of the dual radioisotopes. The values in image-t for rod 2, which has no activity in the original phantom, are not zero. As discussed above, this might be because of incomplete delayed background subtraction, and the FIG. 13. Results of dual-radionuclide imaging of a mouse. 18 F-FDG and 22 Na were administered by tail vein injection and oral administration, respectively. Images are maximum intensity projections.

10 2266 Fukuchi et al.: PET for multiple tracer imaging 2266 magnitude of this remaining activity is comparable with the error in the other rods activities. From this result, we concluded that our normalization and delayed coincidence methods, which are the same as those used in conventional PET image reconstructions, are applicable for MI-PET. In this study, 22 Na was employed as the second tracer, and the data for normalization were also measured using a 1275 kev c-ray from 22 Na. However, if another radionuclide is used, such as one of those in Table I, measurements or simulations at specific energy for these radionuclides are necessary for image reconstruction. A dual-radionuclide experiment on a small animal shows that our developed system and the image reconstruction method work for distributed tracers in a living organism. The measured distributions of both the first and second tracers are reasonable from a physiological viewpoint. The additional detectors are sensitive to the prompt c-rays from out of the FOV. These c-rays increase the random coincidence noise, which, however, was subtracted using a delayed coincidence method, at least, at the activity levels in our measurements. To handle higher activities, further investigation and the development of equipment appear necessary, such as the addition of a shield for the c-rays from out of the FOV. In both the phantom and animal studies, we reconstructed images for the presence and absence of prompt c-ray detection. Therefore, a superposed image of the first ( 18 F: pure positron emitter) and the second ( 22 Na: positron-c emitter) tracers and an image of only the second tracer were reconstructed. To reconstruct an independent image of a pure positron emitter, we must differentiate the first tracer image from the images with and without prompt c-ray detection and perhaps develop another iterative method. 7 We intend to develop optimal image reconstruction methods for our system in future work. 6. CONCLUSION We developed an MI-PET system and successfully performed dual-radionuclide imaging for a phantom and a small animal. In this study, we reconstructed the superposed distribution of the first and second tracers and the distribution of the second tracer. Even though there are applications for which superposed tracer distributions may be useful, we need to develop a specific image reconstruction method for MI- PET to obtain independent images of the tracers. We will also use our developed MI-PET system for various applications of multiple-tracer imaging. ACKNOWLEDGMENTS The authors thank the members of the RIKEN Center for Life Science Technologies for their support. This work was supported by JSPS KAKENHI Grant Numbers JP15H04770 and JP CONFLICT OF INTEREST The authors have no relevant conflicts of interest to disclose. a) Author to whom correspondence should be addressed. Electronic mail: tfukuchi@riken.jp. REFERENCES 1. Kadrmas DJ. Feasibility of rapid multitracer PET tumor imaging. IEEE Trans Nucl Sci. 2005;52: Ben-Haim S, Kacperski K, Hain S, et al. Simultaneous dual-radionuclide myocardial perfusion imaging with a solid-state dedicated cardiac camera. Eur J Nucl Med Mol Imaging. 2010;37: Andreyev A, Celler A. Dual-isotope PET using positron-gamma emitters. Phys Med Biol. 2011;56: Miyaoka RS, Hunter WCJ, Andreyev A, et al. Dual-radioisotope PET data acquisition and analysis. In: Proceedings of the IEEE Nuclear Science Symposium and Medical Imaging Conference Valencia, Spain: IEEE; 2011: (2012). 5. Gonzalez E, Olcott PD, Bieniosek M, Levin CS. Methods for increasing the sensitivity of simultaneous multi-isotope positron emission tomography. In: Proceedings of the IEEE Nuclear Science Symposium and Medical Imaging Conference Valencia, Spain: IEEE; 2011: (2012). 6. Olcott PD, Levin CS. Methods and systems for increasing the sensitivity of simultaneous multi-isotope positron emission tomography. Patent US B2; Andreyev A, Sitek A, Celler A. EM reconstruction of dual isotope PET using staggered injections and prompt gamma positron emitters. Med Phys. 2014;41: Herraiz JL, Lage E, Parot V, et al. Production of positron-gamma emitters for multiplexed PET (mpet) imaging. Proc IEEE NNS/MIC. 2014;2013: National Nuclear Data Center. Brookhaven National Laboratory Yamamoto S, Okumura S, Watabe T, et al. Development of a prototype Open-close positron emission tomography system. Rev Sci Instrum. 2015;86: Okamura S, Yamamoto S, Watabe H, Kato N, Hamamura H. Development of dual-layer GSO depth-of-interaction block detector using angled optical fiber. Nucl Inst Meth Phys Res Sect A. 2015;781: Yamamoto S, Honda M, Oohashi T, Shimizu K, Senda M. Development of a Brain PET System, PET-Hat: a wearable PET system for Brain Research. IEEE Trans Nucl Sci. 2011;58: Lage E, Parot V, Moore SC, et al. Recovery and normalization of triple coincidences in PET. Med Phys. 2015;42: Jacobson M, Levkopvitz R, Ben-Tal A, et al. Enhanced 3D PET OSEM reconstruction using inter-update Metz filtering. Phys Med Biol. 2000;45: Thielemans K, Tsoumpas C, Mustafovic S, et al. STIR: software for tomographic image reconstruction release 2. Phys Med Biol. 2012;57: Gillam JE, Solevi P, Oliver JF, et al. Sensitivity recovery for the AX- PET prototype using inter-crystal scattering events. Phys Med Biol. 2014;59: Cal-Gonzalez J, Lage E, Herranz E, et al. Simulation of triple coincidences in PET. Phys Med Biol. 2015;60: