LSO PET/CT Pico Performance Improvements with Ultra Hi-Rez Option Y. Bercier, Member, IEEE, M. Casey, Member, IEEE, J. Young, Member, IEEE, T. Wheelock, Member, IEEE, T. Gremillion Abstract-- Factors which may improve image quality in PET systems include spatial resolution and scatter fraction. By reducing the energy window, the size of the crystal elements and thickness, the spatial resolution and scatter fraction can be improved. The LSO PET/CT Pico system, which consists of an 8x8 crystal array per detector block, operates on an energy window of 400 to 650 kev. The new LSO PET/CT Hi-Rez system includes a 13x13 crystal array per detector block, a decrease of 20% in crystal thickness and a narrower energy window of 425 to 650 kev. Improvements in spatial resolution and in scatter fraction are therefore expected. Although a reduction in sensitivity may result, the image quality of the system should not be affected. In order to evaluate the overall performance of the new system, tests from the NEMA NU 2-2001 protocol were conducted on LSO PET/CT Pico and Hi-Rez systems, and their results were compared. The Hi-Rez system presented an improvement in spatial resolution of 27% at 1cm and 28% at 10 cm. The scatter fraction on the Pico and Hi-Rez systems indicate average values of 41% and 34% respectively, equivalent to a decrease of 17% for the Hi-Rez system. The sensitivity of the Hi- Rez system diminished by 22% and changes of less than 4% were observed for the peak NEC rate. The improvements in spatial resolution and scatter fraction obtained in our measurements may create significant advances in the physician s ability to detect small lesions in PET/CT images. T I. INTRODUCTION he introduction of a new PET system requires a performance evaluation in order to compare the new system s capabilities to those of its predecessors [1]. Documents, such as the NEMA NU 2-1994 [2], have been developed by experienced research groups with the purpose of providing measurement protocols for comparative performance evaluation of different systems. In order to consider recent advancements in PET instrumentation, the National Electrical Manufacturers Association (NEMA) committee has released a revised document. The NEMA NU 2-2001 [3] protocol suggests testing methods to evaluate the following system characteristics: 1) spatial resolution, 2) sensitivity, 3) scatter fraction, count losses and randoms measurement, 4) accuracy of corrections for count losses and randoms and 5) image quality. For each testing method, the document presents All authors are with CPS Innovations, Knoxville, TN 37932 USA (telephone: 865-218-2000). details concerning the data acquisition and processing procedures, data reporting, as well as source distribution and activity levels required at the start of data acquisition. With the advent of the ultra Hi-Rez option, NEMA NU 2-2001 tests were conducted and the results are discussed in the following sections. II. METHODS AND MATERIALS The systems which were tested are the LSO PET/CT Pico 3D [4] and the LSO PET/CT Pico 3D Hi-Rez [5],[6] (CPS Innovations, Knoxville, TN, USA). Both systems have LSO crystals, Pico electronics and 3D functionality (no septa). They differ primarily in crystal size, crystal thickness and lower level energy discriminator (LLD). TABLE I SCANNER SPECIFICATIONS Scanner Detector material LSO LSO Transaxial field of view 585 585 (mm) Axial field of view (mm) 162 162 Coincidence time window 4.5 4.5 (nsec) Upper level energy 650 650 discriminator (kev) Lower level energy 400 425 discriminator (kev) Image planes 47 81 Crystal dimensions (mm 3 ) 6.45x6.45x25 4x4x20 Crystal array 8x8 13x13 Plane spacing (mm) 3.375 2 Due to the increased LLD and reduced crystal width and depth of the Hi-Rez system, resolution and scatter fraction are expected to improve, however, the overall system sensitivity is expected to decrease. In order to quantify the difference in performance between the two systems, NEMA tests were conducted and the results are given in the following section. A. Spatial resolution The purpose of this test is to measure the full width half maximum (FWHM) of the point spread function (PSF) of a reconstructed point source. The point source assembly
consists of F18 absorbing resin of size 1x1x1 mm 3 is inserted in a glass capillary of 1.1 mm inner diameter, 0.20 mm wall thickness and 75 mm length (Fig. 1). A high-density polyethylene (HDPT) material is located within the tube, in front and behind the resin, in order to keep the resin from moving. The 7.4 MBq point source is then prepared by pulling F18 solution through the assembly. A fixture (Fig. 1), specifically designed for point source positioning, is secured to the patient handling system (PHS). The fixture enables the source assembly to be positioned laterally or vertically every 1 cm within the field of view (FOV). As x and y define the transaxial plane and z defines the axial direction, the NEMA protocol requires data to be acquired at the following 6 (x,y,z) cm locations: (0,1,½FOV z ), (0,1,¾FOV z ), (10,0,½FOV z ), (10,0,¾FOV z ), (0,10, ½FOV z ), (0,10, ¾FOV z ). For each location, 2E6 nettrues were acquired. several activity levels, by providing information concerning peak true countrate and peak noise equivalent countrate. A polyethylene solid right circular cylinder is required for this test (Fig. 3). This phantom has 1) a 203 mm diameter, 2) a 700 mm length and 3) a 700 mm long hole parallel to the central axis of the cylinder, located 45 mm below the center of the phantom. With the phantom centered on the PHS, a 700 mm F18 line source of 1.04 GBq, calibrated for the beginning of the test, is positioned within the hole. Data is acquired over 35 frames with a 600 second acquisition time and a 900 second delay time. Fig. 2. NEMA NU2-2001 sensitivity sleeves and fixtures. The sensitivity sleeves are suspended by the fixtures within the view. Fig. 1. Point source assembly and fixture. The point source assembly is suspended by the fixture within the view. B. Sensitivity The sensitivity test quantifies the system s ability to detect positron annihilations at two positions within the transaxial field of view: 1) at the center and 2) at a radial offset of 10 cm. A 700 mm line source of low activity, calibrated for 3.7 MBq of F18 at the beginning of the test, was used. Five 700 mm aluminum sleeves, used as attenuating material surrounding the line source, are suspended in the field of view (Fig. 2). The sleeves are of the following inner and outer diameter sizes in mm: 1) 3.9 and 6.4, 2) 7.0 and 9.5, 3) 10.2 and 12.7, 4) 13.4 and 15.9 and 5) 16.6 and 19.1. Data was acquired for 400 seconds for each attenuation thickness in order to measure, by extrapolation, the system s sensitivity without interfering attenuation. C. Scatter fraction, count losses and randoms measurement This test measures 1) the system s sensitivity to scattered radiation by providing a scatter fraction, and 2) the effect of system dead-time and the generation of random events at Fig. 3. NEMA NU2-2001 scatter phantom. The phantom is positioned on the PHS and centered using the lasers.
D. Image quality, accuracy of attenuation and scatter corrections The purpose of this test is to produce images simulating those obtained from patient total body imaging examinations. Two phantoms are required for this test: 1) the image quality phantom and 2) the scatter phantom. The scatter phantom simulates scatter originating from the human body during an examination. The inside of the image quality phantom includes 6 spheres of different sizes and a cylinder containing lung equivalent material. The sphere internal diameters are the following: 1) 10 mm, 2) 13 mm, 3) 17 mm, 4) 22 mm, 5) 28 mm and 6) 37 mm (Fig. 4). At the beginning of the scan, the background region must contain an activity concentration of 5.3 kbq/cc of F18. The 4 smallest spheres must contain an activity concentration of N times that of the background, where N=8 and N=4. The 2 largest diameter spheres shall be cold. The line source in the scatter phantom must have an activity of 116 MBq at the start of the scan. The data acquisition time was set to 432 seconds in order to simulate total body imaging: 100 cm total axial imaging distance in 60 minutes. After image reconstruction, regions of interest (ROIs) of diameters equal to those of the spheres are drawn on the spheres. An additional 12 ROIs are drawn on the background region. The percent contrast and percent background variability are reported for each sphere. TABLE II NEMA NU2-2001 SPATIAL RESOLUTION Radial position and 1 cm offset Transverse [mm] 6.2 4.2 Axial [mm] 5.6 4.5 10 cm offset Transverse tangential [mm] 6.9 4.6 Transverse radial [mm] 7.3 5.0 Axial [mm] 6.9 5.5 B. Sensitivity The results of the sensitivity test are given in Table III. The overall system sensitivity of the Hi-Rez system is 22 % lower than that of the Pico, due to the decrease in crystal depth and increase in LLD. TABLE III NEMA NU2-2001 SENSITIVITY Radial position 0 cm offset [cps/mbq] 5657 4489 10 cm offset [cps/mbq] 5766 4452 C. Scatter fraction, count losses and randoms measurement The scatter fraction on the Pico and Hi-Rez systems indicate average values of 41% and 34% respectively, equivalent to a decrease of 17%. These results are shown in Fig. 5. Changes of less than 4% were observed for the peak NEC rate, as shown in Fig. 6. TABLE IV NEMA NU2-2001 SCATTER Fig. 4. NEMA NU2-2001 image quality phantom. Within the phantom are six different size spheres and a cylinder containing lung equivalent material. Scatter fraction [%] 41 34 Peak NEC rate [cps] 9.74E+04 9.28E+04 III. RESULTS NEMA tests were performed on three different scanners of each system in order to obtain an average performance. On each Pico and Hi-Rez scanners, one and two datasets were acquired respectively. A. Spatial resolution The images were reconstructed 336x336 using a DIFT algorithm. An average of the obtained results is given in Table II. The Hi-Rez system presents an improvement in spatial resolution of 27% at 1cm and 28% at 10 cm. Fig. 5. NEMA NU2-2001 scatter fraction graphs of a Pico scanner (a) and a Hi-Rez scanner (b).
TABLE V NEMA NU2-2001 IMAGE QUALITY (a) Fig. 6. NEMA NU2-2001 noise equivalent countrate graphs of a Pico scanner (a) and Hi-Rez scanner (b). D. Image quality, accuracy of attenuation and scatter corrections As common for clinical studies, images were reconstructed 128x128 by attenuation-weighted FORE OSEM, with 4 iterations, 8 subsets, and a Gaussian filter with FWHM of 5 mm. Images are given in Fig. 7 for the N=8 and N=4 activity concentration ratios. The average contrast and background variability between the N=8 and N=4 ratios are presented in Table V. The results in the table show an improved average contrast and background variability on the Hi-Rez system. (a) (b) (b) Object Average contrast between N=8 and N=4 ratios [%] 10 mm sphere 23 26 13 mm sphere 35 40 17 mm sphere 56 60 22 mm sphere 65 74 28 mm sphere 60 60 37 mm sphere 66 66 Average background variability between N=8 and N=4 ratios [%] 10 mm sphere 7 5 13 mm sphere 6 5 17 mm sphere 5 4 22 mm sphere 5 4 28 mm sphere 4 3 37 mm sphere 3 3 The spatial resolution was evaluated at 6 different locations within the field of view. Compared to the Pico, the Hi-Rez system has an improved and more constant spatial resolution across the field of view. An average 28% improvement in resolution was reported, which may have considerable advantages in clinical environments. The narrower energy window and shallower crystals have decreased the overall system sensitivity and scatter fraction by 22% and 17% respectively. The improved resolution and scatter fraction may result in an increase in image contrast, which allows a more accurate localization and delineation of lesions. The results obtained from the image quality test have shown an improvement in the reproduction of hot and cold spots, ranging from 10 to 37 mm diameter. (c) Fig. 7. NEMA NU2-2001 image quality images. The N=8 and N=4 Pico images are given in (a) and (b) respectively and the N=8 and N=4 Hi-Rez images are given in (c) and (d) respectively. IV. DISCUSSION In order to allow a comparable evaluation of different systems, measurements protocols have been developed by experienced research groups. The NEMA NU 2-2001 protocol has been released in order to consider recent advances in PET instrumentation. Due to the advent of the Hi-Rez system, NEMA tests have been conducted in order to evaluate its performance and to compare it to previously released systems such as the Pico system. This paper has presented the results obtained from the testing methods suggested by the NEMA committee. (d) V. ACKNOWLEDGMENT We would like to thank Tim Mulnix and Curtis Howe, from CPS Innovations, Knoxville, TN, USA, for providing software to process the NEMA data. We would also like to thank Lutz Tellmann, from the Institute of Medicine, Juelich, Germany, as well as Maria-Jose Martinez, from the Technische Universitaet, Munich, Germany, for their help with the acquisition of the image quality data for the Pico system. VI. REFERENCES [1] H. Herzog, L. Tellmann, C. Hocke, U. Pietrzyk, M.E. Casey and T. Kuwert, NEMA NU2-2001 Guided Performance Evaluation of Four Siemens ECAT PET Scanners, IEEE Transactions on Nuclear Science, vol. 51, no. 5, pp. 2262-2269, October 2004. [2] National Electrical Manufacturers Association. NEMA Standards Publication NU 2-1994. Performance Measurements of Positron Emission Tomographs,. Washington, DC, National Electrical Manufacturers Association, 1994. [3] National Electrical Manufacturers Association. NEMA Standards Publication NU 2-2001. Performance Measurements of Positron Emission Tomographs,. Rosslin, VA, National Electrical Manufacturers Association, 2001. [4] M.S. Musrock, J.W. Young, J.C. Moyers, J.E. Breeding, M.E. Casey, J.M. Rochelle et al., Performance Characteristics of a New Generation
of Processing Circuits for PET Applications, IEEE Transactions on Nuclear Science, vol. 50, no. 4, pp. 974-978, August 2003. [5] M. E. Casey, J. Young, T. Wheelock, M. Schmand, B. Bendriem, R. Nutt, Physical Performance of a High Resolution PET/CT Scanner, presented at the 51 st annual meeting of the SNM in Philadelphia. June 20-23, 2004. [6] D. W. Townsend, J. P. Carney, J. T. Yap, M. Long, N. C. Hall, J. Young et al., Clinical Performance of a High Resolution 16-Slice LSO PET/CT Scanner, presented at the 51 st annual meeting of the SNM in Philadelphia. June 20-23, 2004.