M5C1  High Resolution and Preclinical Systems

Friday, Nov. 6  14:00-16:00  Golden Pacific Ballroom

Session Chair:  Scott Metzler, University of Pennsylvania, United States; Emilie Roncali, University of California-Davis, United States

Show/Hide All Abstracts

(14:00) M5C1-1, One Step Calibration of Monolithic Gamma-Ray Detectors Using a Collimator Grid

C. Bouckaert, R. Van Holen, K. Deprez, S. Vandenberghe

Dept. of Electronics and Information Systems, MEDISIP - Ghent University - iMinds Medical IT, Ghent, Belgium
Advanced positioning algorithms such as maximum likelihood position estimation or k-nearest neighbor, are required to accurately position the data from monolithic scintillator detectors. These methods require calibration data that relate a known source position to the detector response. The classical method to acquire calibration data makes use of a robot-driven beam source that scans a Cartesian grid across the detector surface and is very time-consuming. To get sufficient statistics (10000 events per positions), the acquisition of e.g. a 28 by 28 grid takes approximately 5 hours. Since, in most cases, each detector needs to be calibrated separately, this method becomes impractical when a full system is considered. To simplify and speed up the calibration process, we propose a method that makes use of calibration collimators. This will allow simultaneous calibration of all detectors in a system. Our calibration is performed using a fan-beam collimator containing 28 by 28 holes spaced 1.15 mm apart and a line source placed in the focal line of the collimator. Using this method, it only takes 3.5 hours to acquire 10000 events per collimator hole. After positioning the data using an old calibration set obtained for a different detector, the events corresponding to each collimator hole need to be separated from each other. The events corresponding to each collimator hole are then used to determine the detector response functions. The performance of the calibration method was tested using a resolution collimator that contains six region of holes with sizes ranging from 0.4 to 1.4 mm. The detector images were generated making use of both the beam source-based and the collimator-based calibration data. It was found that all 0.6 mm holes and some of the 0.4 mm holes could be discriminated with both methods. The proposed method is thus a practical and fast alternative for the current calibration method based on beam sources.

(14:15) M5C1-2, Scintillator and Pinhole Insert for Nuclear Imaging with a Preclinical Optical Imaging System

G. S. Mitchell1, S. R. Miller2, E. Roncali1, V. V. Nagarkar2, S. R. Cherry1

1Biomedical Engineering, UC Davis, Davis, CA, USA
2Radiation Monitoring Devices, Inc., Watertown, MA, USA
Planar images of gamma-ray emitting radiotracers can be obtained by retrofitting a CCD-based small animal optical imaging system with a pinhole collimator and a large area CsI:Tl scintillator with a novel morphology. Recent progress at Radiation Monitoring Devices (RMD) has led to thick, transparent, crystalline microcolumnar structure (CMS) scintillator detectors of CsI:Tl that simultaneously provide high gamma-ray absorption efficiency, high intrinsic spatial resolution, and bright light emission. The ability to use an existing commercial preclinical optical imaging system to rapidly acquire planar gamma ray images with good spatial resolution of one or possibly multiple animals would be a new and useful tool for high throughput screening of nuclear imaging radiotracers. We are designing an insert which can be placed in an existing optical imaging system, where the insert uses a collimator and scintillator to enable gamma rays to be imaged by the existing sensitive CCD camera. Preliminary data were obtained using a commercial optical imaging system, with a field of view containing a 10 cm by 10 cm scintillator, and the tungsten pinhole collimator and scintillator located above the imaging subject for 1:1 magnification. With a 1 mm diameter capillary tube filled with 1 mCi of Tc-99m, a 0.5 mm collimator resulted in a FWHM image line spread function (LSF) of 1.5 mm, and a 2 mm collimator resulted in a 3.5 mm FWHM LSF. With the 2 mm diameter pinhole collimator, acquisition of dynamic images (30 s exposures obtained every 45 s) of a mouse injected with 1 mCi Tc-99m MAG-3 allows the changing distribution of the radiotracer to be readily observed.

(14:30) M5C1-3, First Results with an Interventional Handheld PET

B. Frisch1, D. Cortinovis2, A. Cserkaszky3, R. Bugalho4, F. Ben Mimoun5, H. Chen6, Z. Liu7, R. Martinez7, Y. Munwes6, M. Pizzichemi7, J. C. da Silva4, A. Shah1, V. Stankova6, C. Zorraquino Gaston4, M. Zvolsky2, E. Auffray5, H.-C. Schulz-Coulon6, J. Varela4, E. Garutti8, P.-H. Rolland9, N. Navab1, P. Lecoq5

1Technische Universität München, Munich, Germany
2DESY, Hamburg, Germany
3SurgicEye, Munich, Germany
4LIP, Lisbon, Portugal
5CERN, Geneva, Switzerland
6Universität Heidelberg, Heidelberg, Germany
7University Milano Biccoca, Milan, Italy
8Universität Hamburg, Hamburg, Germany
9Aix-Marseille Université, Marseilles, France
Radioguided surgery has proven to be a valuable tool for the interventional localisation of radioactive hotspots, guiding the physician to target tissue sampling for a biopsy or to identify cancerous tissue for resection. It is however mostly restricted to the use of low-energy gamma-emitting radiotracers. Indeed, while Positron Emission Tomography (PET) is an established functional imaging method for cancer diagnosis and staging, it is limited by its most common design in a gantry-based setup as well as the long time to acquire sufficient data to reconstruct an image, preventing its use as an intra-operative device. We develop a system which, composed of an external detector plate held by a mechanic arm and a handheld light-weight detector coupled to an ultrasound transducer, would allow for the live acquisition of multimodal PET/US images in an interventional environment. The detectors are based on matrices of 4x4 LYSO:Ce crystals coupled to arrays of 4x4 SiPMs. In the plate, 256 matrices are grouped by 4 and read out by 64-channel ASICs, of which 16 are hosted by each front-end board. The endoscope consists in 2 matrices read out in a similar way. A PCIe data acquisition card concentrates event data for coincidence sorting. The position of both detectors is recorded with an optical tracking system. An iterative ML-EM algorithm reconstructs the image. We present energy spectra and reconstructed images acquired on a phantom where four cylindrical sources filled with FDG can be clearly identified.

(14:45) M5C1-4, Open-Field Mouse Brain PET: Design Considerations and Detector Development

A. Z. Kyme1,2, K. Gong1, M. S. Judenhofer1, J. Qi1, S. R. Meikle2, S. R. Cherry1

1Biomedical Engineering, University of California Davis, Davis CA, USA
2Faculty of Health Sciences and Brain & Mind Research Institute, University of Sydney, Sydney NSW, Australia
‘Open-field’ PET imaging offers the key capability of correlating functional changes in the brain of an awake animal with its behavioral response to environmental or pharmacologic challenges. The feasibility of this concept has been demonstrated in rats using motion compensation techniques, however our current system is not suitable for imaging the mouse brain due to limitations imposed by the use of a commercial PET scanner. Therefore, we are designing a purpose-built scanner which optimizes the geometry, motion tracking and imaging performance for open-field imaging of the mouse brain.

We simulated the sensitivity and spatial resolution performance of four candidate scanner designs: ring, parallel plate, and two box designs. The block detector was a LSO array comprising either 1x1x20 or 0.8x0.8x20 mm3 crystals. A ML-EM reconstruction with DoI capability was used to determine the DoI resolution necessary to achieve approximately uniform and isotropic sub-millimeter spatial resolution throughout the FoV.

The results showed that 2-3 mm DoI resolution is necessary to achieve the required spatial resolution performance for all scanners except the parallel-plate design. However, the sensitivity advantage of the overlapping box design (peak 14% and 30% improvement over the ring design) suggests this unconventional design is favored for imaging the mouse brain. We have also designed a dual-ended readout DoI-encoding detector module based on tiled through-silicon-via SiPM arrays to meet the DoI requirement. The next stage of work is to characterize its performance and use it to build a prototype bench-top scanner for initial testing of the concept.

For the final design we propose to slide the scanner axially on rails according to the animal’s motion, rather than move the animal (as in our previous design). This allows faster motion without disturbing the animal’s behavior and achieves a large axial FoV for animal movement at minimal cost.

(15:00) M5C1-5, Performance Comparison Between a SiPM- and PSAPD-Based PET Insert for Simultaneous High Resolution Small-Animal PET/MR Imaging

M. S. Judenhofer, J. Zhou, X. Bai, A. Kyme, J. Walton, J. Qi, S. R. Cherry

Biomdecial Engineering, University of California, Davis, Davis, CA, USA
In the past we developed a PET insert based on position-sensitive APDs (PSAPDs) and evaluated its performance. However, significant limitations were identified and we have now modified the system to use silicon photomultipliers (SiPMs) as the photodetector while using the same crystal arrays, geometry and processing electronics. The PET system uses 96 detector blocks (10x10 LSO array, 1.2x1.2x14mm3) arranged in 4 rings. Each LSO array is coupled to the photodetector which was initially a PSAPD (14x14 mm2) with a charge sensitive pre-amplifier and is now a size-matched custom made 3x3-SiPM array with resistive network and simple buffer amplifier. Signals were processed by the same commercial PET electronics. Comparative performance measurements of both PET systems were done, focusing on PET imaging performance with and without the MRI operating and on MR image quality. PET performance was evaluated based on measurements of image quality, detection sensitivity and detector performance, both with and without standard MR sequences running. MR performance was evaluated for different sequences interrogating image SNR and homogeneity. PET images were acquired (15 min) at an activity of ~8.5MBq and reconstructed using a MLEM algorithm. MR images of a cylinder were acquired with and without the respective PET insert installed to assess MR performance. Results showed that the SiPM-based PET insert provided an improved average energy resolution of 18% (22% PSAPD) and system timing resolution of < 3ns (PSAPD 12ns). The PSAPD system had a reconstructed spatial resolution of 1.25 mm without MR sequences but showed degradations due to MR gradients and loss of counts (sensitivity 2.5%-1.0%). The SiPM system showed no interference from the MR sequences with 1.0 mm spatial resolution and 3.7% sensitivity. MR images with the PSAPD-PET showed degradation due to the outside copper shield (5-60% loss in SNR). The SiPM-PET does not require shielding and had unperturbed MR performance.

(15:15) M5C1-6, NEMA NU4-2008 Performance Measurements of a Preclinical PET/MR Insert with Digital Silicon Photomultiplier Technology

P. Hallen1, D. Schug1, C. Lerche2, B. Weissler3, P. Gebhardt4, B. Goldschmidt1, A. Salomon3, P. Düppenbecker4, F. Kießling5, V. Schulz1

1Physics of Molecular Imaging Systems, RWTH Aachen University, Aachen, Germany
2Medical Imaging Physics, Forschungszentrum Jülich, Jülich, Germany
3Clinical Application Research, Philips Research, Aachen, Germany
4Imaging Sciences and Biomedical Engineering, King's College London, London, United Kingdom
5Experimental Molecular Imaging, RWTH Aachen University, Aachen, Germany
We have developed a preclinical digital PET/MR insert with a ring of 10 singles detection modules with a hybrid FOV of 160?mm × 96.6?mm (transaxial × axial). The detector stack comprises a pixelated array of 30 × 30 LYSO scintillators with a pitch of 1?mm and a thickness of 12?mm, coupled to a digital silicon photomultiplier via a 2?mm lightguide for light sharing. In this work, we present performance measurements of this PET/MR insert following the NEMA NU4-2008 standard. To determine the spatial resolution, we placed a 22Na point source at different positions inside the scanner's FOV and reconstructed the activity using a 2D filtered backprojection. The measured spatial resolution is about 0.8?mm FWHM in the isocenter and degrades to about 2?mm FWHM at larger radial and axial distances. The system sensitivity is measured with a 22Na point source at different axial positions. The peak system sensitivity is measured to be 2.6% in the scanner's isocenter. Measurements of the NEMA image quality phantom yield recovery coefficients of 0.2 ± 0.04 for a rod diameter of 1?mm, 0.6 ± 0.05 for a diameter of 2?mm and 0.8 ± 0.06 for a diameter of 3?mm. For larger rod diameters the recovery coefficients approach 1. The spill-over ratios are of 0.038 ± 0.003 for the water-filled cylinder and 0.0055 ± 0.0007 for the air-filled cylinder.

(15:30) M5C1-7, A Novel Platform for Low-Light Radioluminescence Microscopy

T. J. Kim, S. Tuerkcan, G. Pratx

Radiation Oncology/Medical Physics, Stanford University, Palo Alto, USA
Background- Radioluminescence microscopy is a new type of technique that uses radionuclides to study the biomolecular transport of small molecules. While the technique provides radioactive decay information at the cellular level, scintillation signals are difficult to capture due to its low brightness. Moreover, since radioluminescence microscopy is a new technique, there is currently no microscope that is customized for such imaging. Methods- In this paper, we designed an infinity-corrected microscope that can efficiently collect scintillation signals by replacing the tube lens component with a commercially available microscope objective lens (Nikon ? 4x). The brightness of the system was evaluated by measuring the signal to noise ratio (SNR), and the low light imaging capability was tested by taking radioluminescence, bioluminescence and fluorescence images of live cells. Results- While the oil-immersion objective lens yielded the highest average SNR among the tested microscope objectives, the intensity distribution was the broadest as well, indicating possible vignetting or loss of emission signals. Also, the SNR was similar between microscope objectives with different magnifications but with the same numerical aperture. Performance test of radioluminescence imaging showed good results in capturing the dim scintillation signals, and the radioactive decay was traceable with single-cell resolution. The Nikon ? system was also able to efficiently capture low light bioluminescence and fluorescence images. Conclusions- Given the brightness and the quality of radioluminescence images, the Nikon ? microscope system is deemed to be suitable for low-light radioluminescence imaging. Moreover, we expect this modular microscope to be generally suitable for bioluminescence and low-excitation fluorescence imaging as well.

(15:45) M5C1-8, MADPET4 - A 3D Printed, Unshielded, MRI Compatible Dual Layer PET Insert: First Simultaneous PET/MRI Results

F. R. Schneider1,2, G. Topping1, J. Cabello1, N. Omidvari1,2, S. Paul2, S. I. Ziegler1

1Klinikum rechts der Isar, Nuklearmedizinische Klinik und Poliklinik, Technische Universität München, Munich, Germany
2Physik Department E18, Technische Universität München, Garching, Germany
MADPET4 (Munich Avalanche Photodiode PET) is an insert for a 7T small animal MRI scanner, matching inside the Tx/Rx coil with a diameter of 150mm. It is a dual layer system allowing DOI (depth of interaction) with individually read out LYSO crystals with an end face of 1.5mm x 1.5mm. Their lengths are 6mm for the inner and 14mm for the outer layer. The signals of the non-magnetic SiPMs (Silicon Photomultiplier, active area 1.2mm x 1.2mm, gain >8M, 32V bias) are currently digitized by SADCs (Sampling Analog to Digital Converter, 10bit, 80MHz) outside of the MRI cabin. Although the insert is assembled with all crystals, SiPMs and cables in their final configuration, at the moment only two opposing PET modules (40 ch each) can be read out by The SADCs (covered FOV 5.05mm x 5.05mm x 19.7mm). The structure of the insert is fully made from 3D printed parts, the PET components have no electromagnetic shielding. MRI images are acquired with a FGRE (Fast Gradient Echo) sequence. The PET reconstruction is an iterative OSEM (Ordered Subset Expectation Maximization) algorithm with a Monte Carlo generated system matrix. In a study of a cubic grid MRI phantom filled with Gd(aq.), only a slight degradation of the MRI image quality and SNR (Signal to Noise Ratio) with the insert is visible. A comparison of the PET energy spectra (Na-22) with and without running FGRE sequence shows no difference. A hybrid PET/MRI phantom with four Gd(aq.) filled syringes and two Na-22 point sources (in the center FOV, distance 5.2mm) is successfully reconstructed, with the PET data acquisition running simultaneously with the MRI FGRE sequence. This proves, that with MADPET4 simultaneous PET/MRI imaging is possible. Up to our knowledge, this is at the moment the only PET/MRI compatible insert for a 7T MRI based on SiPMs. Currently a platform to read out all 2640 channels based on time-over-threshold ASICs (Application Specific Integrated Circuit) is under development.