First Results From the High-Resolution mousespect Annular Scintillation Camera

Transcription

1 First Results From the High-Resolution mousespect Annular Scintillation Camera Andrew L. Goertzen, Douglas W. Jones, Jurgen Seidel, King Li, and Michael V. Green Abstract High resolution SPECT imaging in small animals is often limited by poor sensitivity, leading to prolonged imaging times, very large injected doses, or both. To increase sensitivity while maintaining spatial resolution, we designed and constructed a multi-pinhole collimator array to replace the parallel hole collimators of a Ceraspect TM human SPECT brain scanner. The Ceraspect scanner is comprised of an annular NaI(Tl) crystal within which the eight pinhole collimators (1 mm diameter holes) rotate while projecting non-overlapping images of the object onto the stationary annular crystal. In this manner, only 1/8 th of a collimator rotation is required to acquire a complete tomographic data set. The imaging field of view (FOV) is 2.56 cm, which is sufficient to encompass a mouse. Data is currently acquired in step-and-shoot mode, however the system is capable of list mode acquisition with the collimator continuously rotating. Images are reconstructed using a cone-beam OSEM method. The reconstructed spatial resolution of the system is 1.7 mm and the sensitivity at the centre of the FOV is 13.8 cps/microci. A whole-body bone scan of a mouse injected with Tc-99m MDP clearly revealed skeletal structures such as the ribs and vertebral bodies. These preliminary results suggest that this approach is a good tradeoff between resolution and sensitivity and, with further refinement, may permit dynamic imaging in living animals. Index Terms pinhole SPECT, mouse imaging I. INTRODUCTION IGH-RESOLUTION SPECT-imaging of small animals H has often been limited by poor sensitivity, leading to prolonged imaging times, large injected doses, and compromised spatial resolution. In general, mouse SPECT systems have been comprised of flat, position-sensitive gamma-ray detectors with a single pinhole per detector head (e.g. [1], [2]). In order to increase detection efficiency, multiple detector heads, each with its own pinhole, can be added to the system [3], [4]. This approach has the advantage Manuscript received October 17, A. L. Goertzen was with the Imaging Physics Laboratory, Clinical Center, National Institutes of Health, Bethesda, MD, 20892, USA. He is now with the McConnell Brain Imaging Centre, Montreal Neurological Institute, Montreal, QC, H3A 2B4, Canada. Phone: +1 (514) ; goertzen@bic.mni.mcgill.ca D. W. Jones is with the NIMH SPECT Lab, Clinical Brain Disorders Branch, IRP, National Institute of Mental Health, National Institutes of Health, Bethesda, MD, 20892, USA. J. Seidel is with the Imaging Physics Laboratory, Clinical Center, National Institutes of Health, Bethesda, MD, 20892, USA. K. Li is with the Imaging Physics Laboratory, Clinical Center, National Institutes of Health, Bethesda, MD, 20892, USA. M. V. Green was with the Imaging Physics Laboratory, Clinical Center, National Institutes of Health, Bethesda, MD, 20892, USA. He is now retired. of not compromising image quality but has the disadvantage of a nearly linear increase in system cost with each detector head added. Alternatively, multiple pinholes with overlapping projections can be used on a single detector head [5], [6]. This approach is very cost-effective to implement, but results in the requirement of a complex reconstruction algorithm and a compromise of resolution due to the overlapping projections. In addition to the problem of low sensitivity, one must also consider the problem of how best to image a mouse with SPECT in-vivo. First generation versions of these mouse SPECT systems often were based on a stationary gamma camera, with the mouse rotating on a turntable [1], [2]. This vertical placement of the mouse makes reproducible positioning and monitoring of mouse vital signs very difficult. Thus, most studies were performed with post-mortem, nondynamic paradigms that precluded test-retest within-subject studies. This has prohibited in vivo studies of small animal involving disease-progression, drug-treatment and genetic modification. In spite of the difficulties in imaging a mouse positioned vertically, SPECT pinhole-studies have reported significant spatial-resolution [7] and even ECG-gated temporal-resolution [3]. More recently, microspect cameras have featured a stationary mouse with a rotating camera in order to make in-vivo imaging simpler [8]. These designs featuring a rotating camera can be a challenging engineering problem, since the very heavy gamma cameras must move very precisely. Given the design constraints of high gamma detection efficiency, horizontal placement of a stationary mouse and minimal cost, we designed and constructed an eight pinhole collimator array tailored to mate with a human SPECT brain scanner, the Ceraspect TM (Digital Scintigraphs Inc., Waltham, MA). This choice was predicated on the unique annular geometry of the Ceraspect scanner in which gamma radiation is detected by a stationary annular gamma camera that surrounds the imaging volume. The collimator array is mechanically rotated inside this annulus while projecting eight non-overlapping pinhole images onto the crystal annulus. We report herein our initial experience with this modified human scanner which we have dubbed mousespect. II. MATERIAL AND METHODS A. System Description The Ceraspect TM system on which the mousespect system is based has a stationary 310 mm diameter x 130 mm wide solid sodium-iodide (NaI(Tl)) annular crystal coupled via

2 Fig. 1. Schematic of the pinhole collimator. The shaded regions indicate the eight projections onto the annular crystal. The FOV is the circle at the center of the array. a lightguide to 63 fixed photomultiplier tubes (PMTs) arranged in a cylindrical array. The intrinsic resolution of the NaI(Tl) detection system is 4.5 mm in the axial direction and 3.5 mm in the azimuthal direction of the annular crystal. The Ceraspect TM system has been previously characterized [9], [10]. Our mousespect rotating collimator-array is comprised of eight 1 mm diameter tungsten-alloy pinholes (Kulite 1850 TM, H.C. Starck, East Rutherford, NJ), evenly spaced around a circle, that rotate around a stationary FOV at up to 1 rev/10 s. The pinhole arrangement is shown in figure 1. The pinholes simultaneously project eight circular, nonoverlapping images onto the annular NaI(Tl) scintillation crystal. In as little as 1/8 th of a single rotation of the collimator, or 1.25 s, a full set of tomographic projections can be acquired. This rapid acquisition enables the possibility of dynamic imaging if the activity level is sufficiently high. The FOV is 2.56 cm, slightly larger than a. typical mouse crosssection, and the 28-mm radius of rotation (ROR) yields a magnification of approximately 4.5. This magnification, when coupled with the intrinsic resolution of the camera, led to the choice of 1 mm pinholes. A photo of the pinhole collimator assembly is shown in figure 2. The Ceraspect TM scanner with the pinhole collimator array inserted is shown in figure 3. The locations of scintillation events are mapped by the data acquisition system onto a matrix, corresponding to 1.90 mm in the azimuthal direction and 2.03 mm in the axial direction. These data were normalized using a crystal sensitivity map and then resorted into projection views similar to that of a standard rotating gamma camera with a single pinhole and a flat detector. Smoothing these flattened projections to a resolution of 4 mm (in the crystal, 0.89 mm in the FOV), improved the signal-to-noise ratio (SNR) in the Fig. 2. Photo of the collimator during assembly showing the tungsten-alloy pinhole holder and eight pinhole inserts with one partially installed (with Teflon tape to insure a snug fit). The transparent plastic tube is fixed on concentric, precision ball-bearings and guides the small animal bed into the center of the FOV. At the left and right are self-centering collets that secure the pinhole-collimator-assembly into the rotating mechanism. Fig. 3. Photo of the pinhole collimator assembly inserted in the Ceraspect scanner.

3 Fig. 4. Reconstructed image from the rod phantom. The center-to-center separation is 9 mm and the rod diameter is 1.1 mm. reconstructed images. Images were reconstructed in 3D mode using a cone-beam OSEM algorithm with 8 subsets and 3 iterations. A reconstructed voxel size of mm 3 is used. B. Performance Measurements The full width at half maximum (FWHM) resolution was measured in the reconstructed images by imaging a 3 x 3 array of capillary tubes, each with a 1.1 mm inner diameter, and filled with an aqueous solution of 99m Tc. The capillary tubes were aligned along the scanner axis and held at 9 mm center to center spacing in holes drilled in a 28 mm diameter acrylic cylinder. No correction for source size was employed in the calculation of the resolution. Additionally, nine 99m Tc-filled capillary tubes (1.1 mm inner diameter, 1.5 mm outer diameter) aligned along the scanner axis were imaged when placed next to each other, with a center to center spacing of Fig. 5. Reconstructed image of the nine capillary tubes placed side by side. 1.5 mm. A profile was drawn through the adjacent capillary tubes in the reconstructed image to look at the peak to valley ratio. One hundred and twenty tomographic projections were used in each of these studies. Sensitivity was measured using a 72.5 µci 99m Tc point source positioned at the center of the FOV. For all studies, the energy window was set at kev. C. Mouse Imaging A 46-g mouse was injected with 7 mci of 99m Tc-MDP and was euthanized 2.5 hours after injection. At the time of euthanization, the total activity remaining in the mouse was 1.74 mci. The mouse was imaged using 7 bed positions, each separated by 2 cm. At each bed position, a 30 min scan was conducted with 120 projections. Because the bed had to be moved manually, mispositioning errors were evident in the reconstructed images. Fig. 6. Plot through the centre of the capillary tubes from figure 5 showing that the peaks corresponding to individual tubes can just be resolved. III. RESULTS The measured resolution of the central 1.1-mm capillary tube was 1.7 mm FWHM. Figure 4 shows the image of the 3 3 grid of capillary tubes. Each of the capillary tubes is easily identifiable. The rods at the four corners show slight distortions away from circular symmetry due to being exactly at the edge of the FOV, and hence being only partially sampled in certain projections. Figure 5 shows the results from the scan where nine capillary tubes were placed adjacent to each other. The individual tubes can just be resolved, as confirmed by a profile drawn through the rods, shown in figure 6. The profile reveals that the measured resolution of 1.7 mm is a reasonable measure since the 1.1 mm diameter sources separated by 1.5 mm can just be resolved, but with a very poor peak to valley ratio. The sensitivity measured with

4 , with the highest count frame being centered over the kidneys and the lowest being those at the tip of the skull and the bottom of the lower legs. Fig. 7. Maximum intensity projection through the reconstructed 99m Tc-MDP mouse data set acquired on the mousespect scanner. The apparent data gap in the rib region is due to bed mispositioning errors. the point source centered in the FOV was 13.8 cps/µci. A reconstructed image of the mouse is shown in figure 7. This image is a maximum intensity projection through the reconstructed volume. The resolution of the system is demonstrated by the clear delineation of the ribs of the mouse. The band of missing data in the region of the mouse ribs is due to errors in the manual re-positioning of the bed. The total number of counts acquired in the scan was The number of counts for any one bed position ranged from IV. DISCUSSION AND CONCLUSIONS First results from the mousespect prototype are encouraging and the system design appears to be a workable tradeoff between resolution and sensitivity when imaging animals the size of mice. The initial testing showed that the sensitivity of 13.8 cps/µci and resolution of 1.7 mm closely matched our a priori estimates of 8-14 cps/µci and mm, respectively [11]. The prototype will be completed shortly with installation of a computer-controlled bed to precisely place the animal in the (small) FOV (something lacking in this preliminary evaluation). Given that the system is capable of listmode acquisition with a continuously rotating collimator, we want to use this system to investigate the possibility of dynamic SPECT of the mouse. We believe these preliminary results suggest this option is possible, particularly if lower resolution would be acceptable. To demonstrate this point, we can assume the following parameters: N = counts needed per frame (N) = 10 6 S = sensitivity = cps/mci F = fraction of mouse in the FOV = 0.2 t = duration of dynamic frame = 60 s Then the activity that would need to be injected into the mouse, A, in mci would be given by A N S F t = (1) such that for the desired 60 s frames, 6 mci would need to be injected. This is not an unreasonable amount for a microspect study of a mouse. Despite the apparent effectiveness of the mousespect design, higher spatial-resolution systems of this general type with even higher sensitivity (or temporal resolution) or both are clearly achievable if one were to design an annular system optimized for small animal imaging rather than being limited to the pre-determined geometry of the Ceraspect system. Smaller pinholes, or even different types of collimation may play a role, as may higher resolution pixilated stationary annular scintillation detectors and improved reconstruction algorithms. The present study supports the use of this type of imaging system by providing a good combination of sensitivity, spatial-resolution, low-cost and, as with all SPECT systems, the ability to fully utilize the many long-half-life isotopes amenable to SPECT imaging studies in mice. Future work with the mousespect system will explore these possibilities, including the potential for dynamical imaging and multi-isotope/multi-radiopharmaceutical imaging in the same animal. ACKNOWLEDGMENT We gratefully acknowledge Sebastian Genna of Digital Scintigraphics, Inc. and Vadim Gayshan of Scintitech Inc. for

5 useful discussions during the design-phase of this project. Pinhole-SPECT OSEM reconstruction software was generously provided by Steven Meikle of the University of Sydney and was developed in collaboration with the Thomas Jefferson National Accelerator Facility of the Southeastern Universities Research Association under DOE contract # DE- AC05-84ER REFERENCES [1] M. C. Wu, G. A. Kastis, S. J. Balzer et al., High-resolution SPECT with a CdZnTe detector array and a scintillation camera, in Conf. Rec IEEE Nuclear Science Symposium, pp [2] L. R. MacDonald, B. E. Patt, J. S. Iwanczyk, et al., Pinhole SPECT of mice using the LumaGEM gamma camera, IEEE Trans. Nuc. Sci., vol. 48, n. 3, pt. 2, pp , [3] L. R. Furenlid, D. W. Wilson, Y.-C. Chen, et al., FastSPECT II: a second-generation high-resolution dynamic SPECT imager, in Conf. Rec IEEE Nuclear Science Symposium, pp [4] T. F. Peterson, H. Kim, M. J. Crawford et al., SemiSPECT: a smallanimal imaging system based on eight CdZnTe pixel detectors, in Conf. Rec IEEE Nuclear Science Symposium, pp [5] N. U. Schramm, G. Ebel, U. Engeland et al., High-resolution SPECT using multipinhole collimation, IEEE Trans. Nuc. Sci., vol. 50, no. 3, pp , [6] S. R. Meikle, P. Kench, A. G. Weisenberger et al., A prototype coded aperture detector for small animal SPECT, IEEE Trans. Nuc. Sci., vol. 49, no. 5, pp , [7] M. C. Wu, H. R. Tang, D. W. Gao et al., "ECG-gated pinhole SPECT in mice with millimeter spatial resolution," IEEE Trans. Nuc. Sci., vol. 47, pp , [8] A. G. Weisenberger, R. Wojcik, E. L. Bradley et al., SPECT-CT system for small animal imaging, IEEE Trans. Nuc. Sci., vol. 50, no. 1, pp , [9] S. Genna and A. P. Smith, The development of ASPECT, an annular single crystal brain camera for high efficiency SPECT, IEEE Trans. Nuc. Sci., vol. 35, no. 1, pp , [10] A. P. Smith and S. Genna, Acquisition and calibration principles for ASPECT A SPECT camera using digital position analysis. IEEE Trans. Nuc. Sci., vol. 35, no. 1, pp , [11] D. W. Jones, A. L. Goertzen, S. Riboldi et al., A novel mouse SPECT scanner using an annular crystal, Mol. Imaging Biol., vol. 5, no. 3, p. 109, 2003.