Home
Search
Collections
Journals
About
Contact us
My IOPscience
A novel optically transparent RF shielding for fully integrated PET/MRI systems
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2017 Phys. Med. Biol. 62 7357 (http://iopscience.iop.org/0031-9155/62/18/7357) View the table of contents for this issue, or go to the journal homepage for more Download details: IP Address: 132.174.250.220 This content was downloaded on 03/09/2017 at 10:13 Please note that terms and conditions apply.
You may also be interested in: MR-compatibility of a high-resolution small animal PET insert operating inside a 7 T MRI J D Thiessen, E Shams, G Stortz et al. PET-MRI: a review of challenges and solutions in the development of integrated multimodality imaging Stefaan Vandenberghe and Paul K Marsden Small animal simultaneous PET/MRI Sri Harsha Maramraju, S David Smith, Sachin S Junnarkar et al. FPGA-based RF interference reduction techniques for simultaneous PET–MRI P Gebhardt, J Wehner, B Weissler et al. MR compatibility aspects of a silicon photomultiplier-based PET/RF insert with integrated digitisation Bjoern Weissler, Pierre Gebhardt, Christoph W. Lerche et al. MR-compatibility assessment of the first preclinical PET-MRI insert equipped with digital silicon photomultipliers J Wehner, B Weissler, P M Dueppenbecker et al.
Institute of Physics and Engineering in Medicine Phys. Med. Biol. 62 (2017) 7357–7378
Physics in Medicine & Biology https://doi.org/10.1088/1361-6560/aa8384
A novel optically transparent RF shielding for fully integrated PET/MRI systems C Parl, A Kolb, A M Schmid, H F Wehrl, J A Disselhorst, P D Soubiran, D Stricker-Shaver and B J Pichler1 Department of Preclinical Imaging and Radiopharmacy, Werner Siemens Imaging Center, Eberhard Karls University Tuebingen, Roentgenweg 13, 72076 Tuebingen, Germany E-mail: [email protected] and [email protected] Received 11 October 2016, revised 21 July 2017 Accepted for publication 2 August 2017 Published 1 September 2017 Abstract
Preclinical imaging benefits from simultaneous acquisition of high-resolution anatomical and molecular data. Additionally, PET/MRI systems can provide functional PET and functional MRI data. To optimize PET sensitivity, we propose a system design that fully integrates the MRI coil into the PET system. This allows positioning the scintillators near the object but requires an optimized design of the MRI coil and PET detector. It further requires a new approach in realizing the radiofrequency (RF) shielding. Thus, we propose the use of an optically transparent RF shielding material between the PET scintillator and the light sensor, suppressing the interference between both systems. We evaluated two conductive foils (ITO, 9900) and a wire mesh. The PET performance was tested on a dual-layer scintillator consisting of 12 × 12 LSO matrices, shifted by half a pitch. The pixel size was 0.9 × 0.9 mm2; the lengths were 10.0 mm and 5.0 mm, respectively. For a light sensor, we used a 4 × 4 SiPM array. The RF attenuation was measured from 320 kHz to 420 MHz using two pick-up coils. MRI-compatibility and shielding effect of the materials were evaluated with an MRI system. The average FWHM energy resolution at 511 keV of all 144 crystals of the layer next to the SiPM was deteriorated from 15.73 ± 0.24% to 16.32 ± 0.13%, 16.60 ± 0.25%, and 19.16 ± 0.21% by the ITO foil, 9900 foil, mesh material, respectively. The average peak-to-valley ratio of the PET detector changed from 5.77 ± 0.29 to 4.50 ± 0.39, 4.78 ± 0.48, 3.62 ± 0.16, respectively. The ITO, 9900, mesh attenuated the scintillation light by 11.3 ± 1.6%, 11.0 ± 1.8%, 54.3 ± 0.4%, respectively. To attenuate the RF from 20 MHz to 200 MHz, mesh performed better than copper. The results show that an RF shielding 1
Author to whom any correspondence should be addressed.
1361-6560/17/187357+22$33.00 © 2017 Institute of Physics and Engineering in Medicine Printed in the UK
7357
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
material that is sufficiently transparent for scintillation light and is MRI compatible can be obtained. This result enables the development of a fully integrated PET detector and MRI coil assembly. Keywords: PET/MRI, RF shielding, SiPM, PET, preclinical imaging, small animal PET/MRI, PET/RF (Some figures may appear in colour only in the online journal) Introduction Preclinical imaging studies generally benefit from a simultaneous acquisition of high-resolution anatomical data and molecular information (Judenhofer et al 2008). Integrated positron emission tomography and magnetic resonance imaging (PET/MRI) systems provide complementarily quantitative information about the PET tracer distribution and detailed morphology, and they enable the correlation of these data with spatially and temporally matched functional MRI (Bailey et al 2013, 2014, Sander et al 2013, Wehrl et al 2013, 2014, Maier et al 2014, Griessinger et al 2015). Therefore, the design and optimization of preclinical and clinical PET/ MRI systems is the focus of ongoing research in medical imaging science (Hofmann et al 2008, Wu et al 2009, Kolb et al 2012, Bezrukov et al 2013, Levin et al 2013, Wehner et al 2014). State-of-the-art PET/MRI systems aim to ensure that the overall performance is not hampered compared to stand-alone MRI or PET systems. The MRI performance should enable the application of all standard and advanced functional imaging sequences and provide stateof-the-art imaging performance (Wehrl et al 2011). For PET, the important specifications are spatial resolution (which depends on the crystal size and reconstruction method), quantification accuracy and detection sensitivity. These factors heavily depend on the PET signal quality, namely, the energy and timing resolution. Hence, electronic noise can seriously degrade these PET imaging performance parameters. Furthermore, the detection sensitivity of a PET system is improved when the distance between the object to be measured and the PET detectors is minimized. These limitations must be carefully considered when designing a PET/MRI system, specifically for high-resolution preclinical applications and organ-specific human systems. In the more recent simultaneous PET/MRI systems, the PET detectors use avalanche photo diodes or silicon photo-multiplier (SiPMs) (Catana et al 2006, Maramraju et al 2011, Yamamoto et al 2012, Yoon et al 2012, Wehrl et al 2013, Weissler et al 2014, Kang et al 2015, Olcott et al 2015, Ko et al 2016) and are completely inserted into the open bore of the MRI gantry, between the gradient system and the transmit/receive MRI coil. However, the operation of an MRI system requires the use of strong radio frequency (RF) pulses. In a PET system designed for simultaneous PET/MR imaging, these pulses can easily disturb the PET system performance by introducing electronic noise. Thus, the entire PET/MRI setup must be carefully designed (Pichler et al 2006), and the sources for potential electro-magnetic system interferences must be understood, evaluated and minimized. Restrictions regarding the PET detector size must also be considered when the detector is between the MRI coil and the gradient system. In addition to the design considerations for the PET, the MR image quality and signal-tonoise ratio (SNR) are dominated by the quality of the transmitted and received RF signals. In MR imaging, the noise is mostly intrinsic, e.g. due to in-homogeneities of the magnetic field introduced by the object to measure or electromagnetic interferences (EMI) of the environment. Thus, it cannot be easily reduced (Mispelter and Lupu 2008). However, the amplitude of the received signal can be increased by reducing the distance between the coil and the object to be measured. Hence, similar to PET detectors, MRI coils should be as near to the object as possible 7358
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 1. Conventional PET insert (left) with a separated shielded MRI coil. In this
setup, the bore size dsys is limited by the free space between the MRI coil and its shielding. Our PET-insert approach with an integrated MRI coil (right): The PET bore size dsys is maintained, the shielding is located between the scintillation crystals and the light sensor (SiPM). A certain distance of the shielding to the MRI coil is provided, which prevents mutual EMI interferences.
(Hyde 2007). Typically, MRI coils are shielded to limit radiation losses and avoid parasitic coupling to common mode currents (Mispelter and Lupu 2008, Peng et al 2014). Furthermore, the shielding prevents the capacitive coupling to surrounding conductive parts. In the case of PET/ MRI, the MRI coil shielding also prevents the EMI of RF pulses on PET electronics (Yamamoto et al 2011), which may cause incorrect energy determination or fake events. To maintain the quality and homogeneity of the electro-magnetic field inside the RF coil, the shielding must be placed at a certain distance from the coil (Collins et al 1997, Mispelter and Lupu 2008). In cur rent small-animal PET/MRI systems, the shielded RF coil is separated and located in the PET bore. This setup reduces the useable bore size of the PET system (figure 1, left) and prevents an optimization of the solid angle coverage of the PET system by reducing the system diameter. Because most MRI sequences target 1H, the highest EMI on the PET electronics is expected at its Larmor frequency, which is approximately 300 MHz at a field strength of 7 T (Peng et al 2014). Assuming that the bandwidth of the electronics of a preclinical, non-time-offlight PET system is limited from 1 kHz to 200 MHz, the EMI issue is more critical at lower field strengths because the Larmor frequency may be in the bandwidth of the PET signals. However, MR gradient switching also causes EMI at lower frequencies, which depend on the MR sequence (Yamamoto et al 2011, Peng et al 2014, Wehner et al 2015). Most current preclinical PET/MRI systems (Judenhofer et al 2008, Yamamoto et al 2011, Yoon et al 2012, Hong et al 2013, Wehrl et al 2013, Wehner et al 2014) use standard shielded MRI coils. In addition, the entire PET front-end such as electronics, scintillators and light sensors, are commonly covered with a conductive material, which is connected to the system ground to avoid EMI. Another approach is to exclude the light sensors from the RF shielded housing of the PET electronics (Hong et al 2012, Yamaya et al 2015, Nishikido et al 2016). To achieve a high level of integration, compactness, maximization of the open combined system bore size and PET sensitivity without degrading the MR image quality, we propose a system design that integrates the MRI coil into the PET detector system. Because common PET scintillators are nonconductive and have low magnetic susceptibility (Strul et al 2003), they do not affect the quality of the electro-magnetic field. Thus, our approach (figure 1, right) is to radially assemble the MRI coil in the PET scintillator blocks. The shielding is placed between the scintillator blocks and the SiPM light sensor. Thus, a sufficient distance to the MRI coil is provided to avoid parasitic capture, which may detune the MRI coil. 7359
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
However, the shielding material must exhibit optimal EMI protection for the PET electronics, prevent the MRI from the noise introduced by the PET, and be transparent for the scintillation light. In this study, we evaluated this essential component of this innovative detector approach, which enables a high degree of integration for a PET detector and an MRI coil by testing three different shielding materials. Material and methods Several conductive materials that enable light transmission are commercially available. We compared three different materials, as shown in table 1 and figure 2. The performed tests measured the efficiency of an RF shielding, which was inserted between the PET scintillator crystals and the SiPM light sensor, on the PET detector performance parameters, such as the peak-to-valley ratio of the crystal position profile (PVR), energy resolution at 511 keV, and timing resolution. Furthermore, we assessed the shielding efficiency of the materials between 320 kHz and 420 MHz in comparison to a 9 µm-thick continuous layer of copper. Investigated shielding materials
• ITO: a polyethylene terephthalate foil evaporated with indium tin oxide. It is widely used in solar cell technology (Richard Woehr, Hoefen/Enz, Germany). • 9900: a polyethylene terephthalate foil coated with a thin conductive layer. It is commonly used to shield touch screens from electromagnetic interference (Holland Shielding Systems, Dordrecht, Netherlands). • mesh: a mesh of stainless steel wires, 100 wires per square inch, each having a diameter of 50 µm. The wire mesh was enclosed in polyethylene terephthalate and had an overall thickness of 280 µm. It was developed to shield displays and windows of RF cabinets such as MRI rooms or EMI measurement chambers (Holland Shielding Systems, Dordrecht, Netherlands). Further specifications of the evaluated samples with a common size of 51 × 110 mm2 are listed in table 1. As a reference, we used a printed circuit board that consisted of a core of 500 µm glass-reinforced epoxy laminate (FR-4), which was continuously covered with 9 µm of copper (Goettle Leiterplattentechnik, Koenigsbrunn, Germany). This material (CuPCB) was used to shield our conventional, custom-built, whole-body preclinical PET/MRI insert (Wehrl et al 2013). The analysis on the acquired data has been performed in MATLAB (MathWorks, Natick, MA, USA). The statistical significance of the measurements is stated as probability values (P-value). These were determined as follows: we began by testing each dataset for normal distribution by a One-sample Kolmogorov–Smirnov test with a significance level of 5%. In case of a positive decision on the normal distribution of all datasets within a category, we performed a t-test. In case of a negative decision on one of the datasets within a category we performed a non-parametric Wilcoxon-rank-sum test (Hollander et al 2015) for the entire category. PET performance measurements
The PET test setup is illustrated in figure 3. We used a PET detector that consisted of two stacked 12-by-12 cerium-doped lutetium oxyorthosilicate (LSO) matrices (Agile Engineering, 7360
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Table 1. Properties of the investigated optically transparent shielding materials.
Short description Typea Light transmissiona Sq. resistancea Resistanceb Attenuationa Layer thickness a b
% Ω/25.4 mm2 Ω/(51 × 110 mm2) dB (100 MHz) µm
ITO
9900
mesh
CuPCB
NV-CT-100 70 80–120 367.5 — 180
9900 68 50 136.5 50 120
9000 48 0.1 2.8 68 280
— 0 — 0.5 — 500
Manufacturer specifications. Measured by a 2-point ohmmeter.
Figure 2. Picture of the three RF-attenuation materials: ITO (left), 9900 (center), and mesh (right). To connect the materials electrically, they had a highly conductive region along 2 edges. For the ITO material it was made of copper, for 9900 and mesh silver was used. The attached copper stripe was used to connect the materials to the common ground of the RF attenuation setup. The reference material (CuPCB) for the RF attenuation is shown in figure 6.
Knoxville, TN, USA). The inner array, which was next to the FoV of the PET system, was shifted by half a pitch in both directions relative to the outer layer. The faces of all LSO pixels were 0.9 × 0.9 mm2; their thicknesses were 5.0 mm and 10.0 mm for the inner and outer crystal layers, respectively. The individual crystals of both layers were optically isolated by a wide-spectral-range reflective foil (Vikuiti, 3M, St. Paul, MN, USA). This foil was also attached to the top of the inner layer to reflect the scintillation light toward the outer layer and light sensor. The light-transmitting surfaces of the layers were optically coupled using optical grease (BC630, Saint-Gobain Crystals, Nemours Cedex, France). The scintillation light of the stacked LSO matrices was spread by a solid light guide (13.6 × 13.6 × 1.0 mm3, borosilicate glass) across a 4 × 4 SiPM array (S11828-3344M, Hamamatsu Photonics, Hamamatsu City, Japan). To distinguish the layers and determine the timing resolution, we used a 40.0 mm-long tungsten collimator with a slit of 1.0 × 25.0 mm2 and a reference detector. This detector consisted of a PMT module (H6779-01, Hamamatsu Photonics) with an optically coupled (BC630, Saint-Gobain Crystals) 3.5 × 3.5 × 20.0 mm3 LSO crystal. The scintillator was wrapped in three layers of white Polytetrafluoroethylene (PTFE) tape. The 16 SiPM elements of the block detector were operated at 73.5 V; this was 1.3 V above the average recommended operation voltage from the manufacturer. The output signals were passively multiplexed to four channels using a resistor network. These four signals were fed to current to the voltage converters (ADA4895, Analog Devices, Inc., Norwood, MA, USA) with a conversion factor of 182 V A−1. These four signals were fed to a summation circuit to provide an analog sum output. All output signals were digitized with 12-bit resolution and 7361
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 3. PET test setup for the coincidence measurements: The slit-collimated 22Na source irradiated the reference detector (PMT) and outer LSO layer of the SiPM block detector. The inner LSO layer was shifted by half a pitch in both directions relative to the outer layer. The scintillation light of the crystals was spread by a solid light guide across the 4 × 4 SiPM array. The RF shielding materials were placed between the light guide and the SiPM, whose signals were multiplexed by a resistive network into four channels and amplified. The four encoding channels, their sum signal, and the PMT signal were fed into the digitizer; the acquired data of coincident events were sent to a PC.
250 MS s−1 (V1720, CEAN, Viareggio, Italy). The PMT detector was directly connected to another digitizer channel. The digitizer was set to acquire only coincident events of the block detector and PMT module. Because of different propagation delays of both detectors, the coincidence time window was adjusted to 40 ns. After the signal exceeded a threshold of 9.8 mV, the waveform was digitized from 40 ns before to 160 ns after the trigger crossing (200 samples). First, all digitized waveforms were baseline-corrected using the average value of the first four samples. In this process, all 200 samples were integrated to determine the energy of the respective channel. For all digitized energy values we did not correct for the saturation of the light sensor. Subsequently, a software constant fraction discriminator (CFD) calculated the timestamp more precisely than the sampling speed of the digitizer (4 ns). The position of every event in the block detector was calculated using Anger logic (Anger 1958) and sorted into a crystal position profile. A crystal lookup map was generated using a Watershed algorithm (Roerdink and Meijster 2000) to discriminate individual crystals and compute the individual full width at half maximum (FWHM) energy resolution for the 511 keV photo peak. The PVRs were calculated from the position profiles for specific crystal rows and columns in the center and on the edge of the block detector. The maxima and minima were selected by a peak-finding algorithm. The timing resolution was computed from the timestamps of the sum channel of the SiPMs and the PMT. The PMT signal was neither shaped nor calibrated in time, and only a lower energy threshold of 9 keV was applied. The determined variation of the differences in time between the block detector and the PMT represented the timing resolution of the folded signals of both detectors. All measurements were performed with and without the respective shielding materials between the LSO scintillator block and the SiPM light sensor. To measure the relative attenuation of the scintillation light by the respective shielding materials, we determined the position of the full 511 keV energy peaks in a single crystal energy spectra and calculated the ratio of the 511 keV peak positions when the shielding mat erials were used relative to a measurement with no shielding material of a single, designated crystal. 7362
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
All measurements were performed in a dark environment at 25 ± 0.1 °C. To irradiate the detectors, a 22Na point source was used with a residual activity of approximately 950 kBq (Eckert & Ziegler, Berlin, Germany). Then, 106 coincident events were acquired for each measurement. The individual shielding materials were placed between the SiPM array and the light guide, as illustrated in figure 3. This study focused on comparing the PET detector performance with different shielding materials. Thus, we only investigated the outer crystal layer, which was closer to the light sensor. All PET measurements have been repeated three times. The optical coupling of the scintillator block, the light guide and the RF shielding mat erial was replaced before every measurement. RF attenuation performance of the shielding materials
To investigate the RF attenuation performance of the shielding materials we investigated the effect of an MRI sequence on the count rate of the PET detector block. The setup is depicted in figure 4. We used the same scintillator setup and electronics as in the PET performance evaluation. To increase the mechanical stability, we used solid transparent silicone rubber (EJ-560, Eljen Technologies, Sweetwater, TX, USA) with a size of 13.6 × 13.6 × 1.0 mm3 as a light guide. A nonconductive housing enclosed the detector setup to prevent the detection of light from the environment. The power supplies, the ADC and the PC were located outside the RF cabin of the MRI. Due to the ambient temperature of approximately 23 °C, we adjusted the bias voltage of the SiPM to 72.9 V. In this experiment we acquired the sum signal of the PET detector at a threshold of 9.8 mV. The average count rate was calculated based on the ADC data over 30 s during the MRI system was idle or running a Turbo RARE sequence. We used a 7 T MRI system (BioSpec 70/30, Bruker, Ettlingen, Germany). The MRI signals were transmitted and received by a surface coil (T6615V3, Bruker) that was located at a distance of approximately 1.5 mm to the detector setup. We included a solid, 1 mm thick phantom (water and 2% Agarose) on the coil. The respective shielding materials were placed on top of the detector setup. We used a Turbo RARE sequence with a repetition time (TR) of 800 ms, an echo time (TE) of 5.14 ms and a rare factor of 16, which lead to an effective TE of 30.82 ms. The power of the excitation pulse was set manually to 20 W. We did not use a γ-source for this experiment, the events in the PET detector were generated by the inherent radiation of the LSO crystals. All measurements have been repeated five times for the respective shielding materials. In a separate test setup, we evaluated the RF attenuation of the three shielding materials over a wider frequency range than the 7 T MRI system could provide. We placed the shielding materials in the center of two pick-up coils, which faced each other at a distance of 20 mm. Figure 5 shows a scheme of the setup. In MRI technology, pick-up coils are antennas that are commonly used as surface coils to receive RF signals. We used this type of coil to transmit and receive RF signals. The pick-up coils were built from a coaxial cable (RG-59/U, Belden, Inc., Indianapolis, PA, USA), where its core was formed to a circular loop with a diameter of 9 mm. The ground sheath of the cables was removed at the coils and connected to the shielding material in the test condition using a 5 mm-wide and 40 µm-thick copper stripe (Scotch 1181, 3M, St. Paul, MN, USA), which covered the entire longer edge of the material. We used a network analyzer (A6051, Agilent Technologies, Inc., Santa Clara, CA, USA) to determine the forward transmission coefficient, which was the level of signal that coil B received relative to the RF stimulus applied to coil A. As a stimulus, we used a sine wave with a frequency from 320 kHz to 410 MHz to have a broad coverage of the 1H Larmor frequencies at an MRI field strength up to 9.4 T. Because the materials under investigation were placed between the coils, the forward transmission coefficient was an indicator of achieved 7363
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 4. Setup to investigate the influence of a MRI sequence on the count rate of the
PET detector.
Figure 5. Scheme of the RF test setup. The shielding material was placed between two coils and connected to the common ground. A network analyzer determined the forward transmission coefficient between coil A and coil B over a frequency domain from 320 kHz to 420 MHz.
RF attenuation. All measured data were smoothed using a moving average algorithm. To determine the effect of the environment on our setup, we included a blank measurement with no shielding material between the coils. MRI compatibility To investigate the MR compatibility of the shielding materials, we performed measurements to determine the influence of the presence of the material on the B0 field of the MRI. These measurements include the RF noise spectrum and imaging sequences, such as echo planar imaging (EPI), diffusion weighted imaging (DWI), and magnetic resonance spectroscopy (MRS). We used a 7 T MRI system (BioSpec 70/30, Bruker). The system was equipped with a B-GA 20S gradient system. To adapt the test setup to our future mouse brain PET/MRI system, we placed the test material radially outside an MRI coil (T20063V3, Bruker). The MRI coil offered an inner diameter of 23 mm and was used in transmit- and receive-mode. For all measurements, we used a plastic tube with an inner/outer diameter of 9.7/11.8 mm filled with 7364
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 6. MRI compatibility setup: The foam inside the water phantom minimized the
amount and size of air bubbles in the water. During the measurements, additional foam was used to radially center the phantom in the MRI coil. The coil was used in transmit and receive mode. The RF shielding materials were located around the coil. As an RF shielding material, the picture shows exemplarily CuPCB. A Euro Cent is shown as a reference in dimension.
water, covering approximately 45 mm length. A piece of plastic foam was used to minimize the number and size of air bubbles in the FoV. The phantom itself was centered inside the MRI coil using foam. Figure 6 shows a picture of the test setup. The parameters of the used sequences are summarized in table 2. The effect of the shielding material and the PET detector on the MRI was investigated with the setup depicted in figure 4. We performed a spin echo sequence (SE) in which the RF power was set to zero. Thus, no RF pulse was applied; however, the MRI recorded the receive channel over a bandwidth of 50 kHz around the imaging frequency of ~300 MHz with a resolution of 763 Hz. The receiver gain was set to the same value for all measurements. This procedure has been performed twice, with the PET detector switched off and while it was powered on and acquiring events. To evaluate the effect of the shielding materials on the MRS, a 5 × 5 × 10 mm3 ROI in the center of the phantom was selected and a spectroscopy sequence was performed for this region. Within the absolute values of the resulted spectra, we calculated the FWHM and the full width at tenth maximum (FWTM) of the water peak. Additionally, we determined the SNR by the ratio of the peak value and the mean of 100 values at the left noise area of the spectra. The influence of the shielding material on the homogeneity of the main magnetic field (B0) was evaluated by a field map sequence. A FoV of 40.03 × 40.03 × 40.03 mm3 was scanned in an isotropic resolution of (0.417 mm)3. For every material, a B0 map was generated by averaging the signal intensity of every pixel over five repetitions of the measurement. Additionally, we performed a similar measurement without any additional material. To find the region of the phantom within the B0 dataset we performed a T1 sequence with an TE of 6 ms, a TR of 80 ms, a flip angle (FA) of 90° and the same resolution as the field map. The image of the T1 sequence was then discriminated by a threshold of 20% of the maximum signal intensity, and the contour of the same slice was overlaid onto the B0 map. Within this dataset, we selected the coronal center slice of the phantom in which we calculated the average and standard deviation of the magnetic field within the T1 contour. 7365
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Table 2. MR sequences and parameters.
Sequence
Repetition Echo time, time, TR (ms) TE (ms)
Turbo RARE 800 B0 field map 35 PRESS MRS 2500 SE DWI 1100 SE EPI 2000 a b
5.14/30.82a 2.34 16.60 42.00 30.00
Voxel size (mm3)
Matrix size
Slices
Flip angle, FA (°)
0.313 × 0.313 × 0.313 0.417 × 0.417 × 0.417 5.000 × 5.000 × 10.000 0.391 × 0.391 × 1.000 0.391 × 0.391 × 2.000
32 × 32 96 × 96 1 × 1 64 × 64 64 × 64
32 96 1 5 5
—b 30 90 90 90
Effective echo time, rare factor 16. The excitation pulse power was manually set to 20 W.
We also investigated the effect of the shielding materials on advanced imaging sequences, such as DWI and EPI. With the DWI method, we acquired five transversally adjacent slices in the center of the phantom covering a FoV of about 25 × 25 × 5 mm3. The voxel size was 0.391 × 0.391 × 1.000 mm3. We used a TR of 1100 ms and a TE of 42 ms. Five images have been acquired for b-values from 0 to 1000. Since we expected the highest impact on the SNR at the highest b-value, only this measurement is included in the results. We defined a signal-ROI of 12 × 12 voxels inside the phantom and determined the maximum signal intensity Smax and minimum value Smin. According to Wehrl et al (2011) we calculated the homogeneity H as follows: Smax − Smin H = 1− × 100%. Smax + Smin
For the determination of the SNR we defined an additional noise-ROI of 12 × 12 voxels at the lower right edge of the image. Afterwards, we calculated the ratio of the average signal intensity S of all pixels inside the signal-ROI and the standard deviation σ of the signal intensities in the noise-ROI. To account for the noise distribution of the magnitude images of the MRI (Henkelman 1985); the SNR was then calculated as: SNR = 0.655 ×
S . σ
The image analysis was performed on the center slice, the other slices have been controlled by visual inspection to ensure the comparability of the results. For the EPI sequence we also acquired five transversally adjacent slices in the center of the phantom to cover a FoV of 25 × 25 × 10 mm3. Per slice, 200 images have been generated. The TR was set to 2000 ms and the TE to 30 ms. We determined the drift of the signal intensity by averaging the values of the signal-ROI within the individual images. These averages were then plotted over the series of 200 EPI-images and a linear fit was applied. All imaging sequences have been applied five times. We determined the statistical significance of the measured parameters between the respective shielding materials and the measurements without any additional material. Results PET performance
One of the crystal position profiles and PVR histograms for the respective shielding material is shown in figure 7. We summarized the results for the PVR of the row and column in the 7366
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 7. Position profiles and PVR histograms of a reference measurement (A) with no additional shielding material. (B)–(D) Show the results with the ITO-foil, 9900foil, and mesh material, respectively. The data that has been used to generate the PVR histograms is indicated by the pathways in the position profiles. The circles identify the crystal that was used to investigate the attenuation of the scintillation light; see figure 8.
center and on the edge in average across three measurements in average and standard deviation in table 3. The energy resolution, the timing resolution and the photo peak position of all 144 crystals across the measurements are also listed in table 3. Figure 8 compares the energy spectra of the seventh crystal in the fourth row for each material, exemplarily. The respective crystal is indicated by a circle in the crystal position profile in figure 7. For this crystal, the average light attenuation and standard deviation across the measurements of ITO foil, 9900 foil, and mesh were 11.3 ± 1.6%, 11.0 ± 1.8%, and 54.3 ± 0.4%. For an overview, figure 9 compares the photo peak positions which have been derived from the individual energy spectra- of the 144 crystals for each material in equal color scale. The results of the influence of the MRI on the PET detector count rate are summarized as mean across five measurements and its standard deviation in table 4. The respective shielding material was used to shield the RF signals of the MRI (figure 4). The measurements in this section were performed without a precise temperature stabilization of the PET detector, this may have caused a slight fluctuation of the mean count rates. 7367
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Table 3. Average PET performance parameters across three repetitions of the same
measurement and their standard deviation.
Material
PVR center
PVR edge
Energy resolution (%, 511 keV)
non ITO 9900 mesh
5.69 ± 0.29 4.50 ± 0.39 4.78 ± 0.48 3.62 ± 0.16
5.36 ± 0.20 4.75 ± 0.27 4.39 ± 0.23 3.73 ± 0.25
15.73 ± 0.24 16.32 ± 0.13 16.60 ± 0.25 19.16 ± 0.21
Timing resolution (ns)
Photo peak position (a.u., 511 keV)
2.39 ± 0.06 2.54 ± 0.06 2.51 ± 0.07 2.80 ± 0.08
45663 ± 277 40359 ± 728 40369 ± 750 20997 ± 120
Figure 8. Single-crystal energy spectra of the shielding materials (ITO, 9900, mesh) and
a reference measurement (non) with no additional material. The respective crystals are marked by the circles in figure 7. The position of the full energy peak of the individual spectra indicates the attenuation of the scintillation light by the respective material. The photo peak positions are labeled in percent of the non-measurement.
RF attenuation performance
Figure 10 shows the RF attenuation performance of the tested materials versus frequency. From 320 kHz up to approximately 6 MHz, all materials (ITO, 9900, mesh, and CuPCB) had similar RF attenuations. From 6 MHz to approximately 200 MHz, the mesh shielding showed better attenuation than the CuPCB, except at approximately the 3 T Larmor frequency (~130 MHz). MRI compatibility
To address the MRI compatibility of the shielding materials we wrapped a commercial MRI coil in the respective material and performed MRI routines to evaluate the B0 field map, the RF noise, and MRS as well as imaging sequences as DWI and EPI. Figure 11 compares the results of the RF noise measurement for the on- and off-state of the PET detector. The average noise level and root mean square (RMS) across the spectra are summarized in table 5. Over all, no prominent peaks in the received signal could be observed, regardless of the shielding material used and the detector status. In all cases we could observe minor changes in the noise level and the RMS depending on the detector status. 7368
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 9. Photo peak position of all crystals for the respective materials ((B)–(D)) and
a measurement without any additional material (A) in equal color scale.
Table 4. PET detector count rate during the MRI is idle and running a Turbo RARE
sequence and while the MRI was in idle state. The respective shielding materials were placed between the MRI coil and the detector. The setup is illustrated in figure 4. Detector count rate (cps)
Material
MRI status
Mean ± SD
non
Sequence Idle Sequence Idle Sequence Idle Sequence Idle Sequence Idle
1431 ± 16 683 ± 21 670 ± 14 649 ± 42 716 ± 14 706 ± 28 657 ± 19 676 ± 7 611 ± 3 605 ± 6
ITO 9900 mesh CuPCB
7369
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 10. RF attenuation level (L) of each optically transparent shielding material (ITO, 9900, mesh) versus frequency (f). As a reference, a 500 µm-thick epoxy core with a 9 µm-thick copper layer (CuPCB) and a blank measurement (non) with no shielding material are included. The frequency is logarithmically scaled from 320 kHz to 420 MHz. The vertical lines indicate the approximate Larmor frequencies of 1H at magnetic field strengths of 1.0, 1.5, 3.0, 7.0, and 9.4 Tesla.
Figure 11. RF noise spectra of the MRI receiver channel in arbitrary units (a.u.) over the resonance frequency f0 = ~300 MHz ± 50 kHz, acquired with no additional material around the MRI coil (non) and four RF-shielding materials: ITO, 9900, mesh, CuPCB. The receiver gain was constant at all measurements.
Exemplarily, figure 12 depicts one of the MRS spectra of the water peak for the respective shielding materials and a measurement without any additional material (non). The average SNR, FWHM, FWTM across five repetitions of the measurement for the respective shielding material are listed in table 6. The shielding materials did not degrade the width of the water peak and the SNR. The investigation of the homogeneity of the main magnetic B0 field is shown in figure 13. In all measurements, the variations over the upper 70% of the phantom were
Search
Collections
Journals
About
Contact us
My IOPscience
A novel optically transparent RF shielding for fully integrated PET/MRI systems
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2017 Phys. Med. Biol. 62 7357 (http://iopscience.iop.org/0031-9155/62/18/7357) View the table of contents for this issue, or go to the journal homepage for more Download details: IP Address: 132.174.250.220 This content was downloaded on 03/09/2017 at 10:13 Please note that terms and conditions apply.
You may also be interested in: MR-compatibility of a high-resolution small animal PET insert operating inside a 7 T MRI J D Thiessen, E Shams, G Stortz et al. PET-MRI: a review of challenges and solutions in the development of integrated multimodality imaging Stefaan Vandenberghe and Paul K Marsden Small animal simultaneous PET/MRI Sri Harsha Maramraju, S David Smith, Sachin S Junnarkar et al. FPGA-based RF interference reduction techniques for simultaneous PET–MRI P Gebhardt, J Wehner, B Weissler et al. MR compatibility aspects of a silicon photomultiplier-based PET/RF insert with integrated digitisation Bjoern Weissler, Pierre Gebhardt, Christoph W. Lerche et al. MR-compatibility assessment of the first preclinical PET-MRI insert equipped with digital silicon photomultipliers J Wehner, B Weissler, P M Dueppenbecker et al.
Institute of Physics and Engineering in Medicine Phys. Med. Biol. 62 (2017) 7357–7378
Physics in Medicine & Biology https://doi.org/10.1088/1361-6560/aa8384
A novel optically transparent RF shielding for fully integrated PET/MRI systems C Parl, A Kolb, A M Schmid, H F Wehrl, J A Disselhorst, P D Soubiran, D Stricker-Shaver and B J Pichler1 Department of Preclinical Imaging and Radiopharmacy, Werner Siemens Imaging Center, Eberhard Karls University Tuebingen, Roentgenweg 13, 72076 Tuebingen, Germany E-mail: [email protected] and [email protected] Received 11 October 2016, revised 21 July 2017 Accepted for publication 2 August 2017 Published 1 September 2017 Abstract
Preclinical imaging benefits from simultaneous acquisition of high-resolution anatomical and molecular data. Additionally, PET/MRI systems can provide functional PET and functional MRI data. To optimize PET sensitivity, we propose a system design that fully integrates the MRI coil into the PET system. This allows positioning the scintillators near the object but requires an optimized design of the MRI coil and PET detector. It further requires a new approach in realizing the radiofrequency (RF) shielding. Thus, we propose the use of an optically transparent RF shielding material between the PET scintillator and the light sensor, suppressing the interference between both systems. We evaluated two conductive foils (ITO, 9900) and a wire mesh. The PET performance was tested on a dual-layer scintillator consisting of 12 × 12 LSO matrices, shifted by half a pitch. The pixel size was 0.9 × 0.9 mm2; the lengths were 10.0 mm and 5.0 mm, respectively. For a light sensor, we used a 4 × 4 SiPM array. The RF attenuation was measured from 320 kHz to 420 MHz using two pick-up coils. MRI-compatibility and shielding effect of the materials were evaluated with an MRI system. The average FWHM energy resolution at 511 keV of all 144 crystals of the layer next to the SiPM was deteriorated from 15.73 ± 0.24% to 16.32 ± 0.13%, 16.60 ± 0.25%, and 19.16 ± 0.21% by the ITO foil, 9900 foil, mesh material, respectively. The average peak-to-valley ratio of the PET detector changed from 5.77 ± 0.29 to 4.50 ± 0.39, 4.78 ± 0.48, 3.62 ± 0.16, respectively. The ITO, 9900, mesh attenuated the scintillation light by 11.3 ± 1.6%, 11.0 ± 1.8%, 54.3 ± 0.4%, respectively. To attenuate the RF from 20 MHz to 200 MHz, mesh performed better than copper. The results show that an RF shielding 1
Author to whom any correspondence should be addressed.
1361-6560/17/187357+22$33.00 © 2017 Institute of Physics and Engineering in Medicine Printed in the UK
7357
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
material that is sufficiently transparent for scintillation light and is MRI compatible can be obtained. This result enables the development of a fully integrated PET detector and MRI coil assembly. Keywords: PET/MRI, RF shielding, SiPM, PET, preclinical imaging, small animal PET/MRI, PET/RF (Some figures may appear in colour only in the online journal) Introduction Preclinical imaging studies generally benefit from a simultaneous acquisition of high-resolution anatomical data and molecular information (Judenhofer et al 2008). Integrated positron emission tomography and magnetic resonance imaging (PET/MRI) systems provide complementarily quantitative information about the PET tracer distribution and detailed morphology, and they enable the correlation of these data with spatially and temporally matched functional MRI (Bailey et al 2013, 2014, Sander et al 2013, Wehrl et al 2013, 2014, Maier et al 2014, Griessinger et al 2015). Therefore, the design and optimization of preclinical and clinical PET/ MRI systems is the focus of ongoing research in medical imaging science (Hofmann et al 2008, Wu et al 2009, Kolb et al 2012, Bezrukov et al 2013, Levin et al 2013, Wehner et al 2014). State-of-the-art PET/MRI systems aim to ensure that the overall performance is not hampered compared to stand-alone MRI or PET systems. The MRI performance should enable the application of all standard and advanced functional imaging sequences and provide stateof-the-art imaging performance (Wehrl et al 2011). For PET, the important specifications are spatial resolution (which depends on the crystal size and reconstruction method), quantification accuracy and detection sensitivity. These factors heavily depend on the PET signal quality, namely, the energy and timing resolution. Hence, electronic noise can seriously degrade these PET imaging performance parameters. Furthermore, the detection sensitivity of a PET system is improved when the distance between the object to be measured and the PET detectors is minimized. These limitations must be carefully considered when designing a PET/MRI system, specifically for high-resolution preclinical applications and organ-specific human systems. In the more recent simultaneous PET/MRI systems, the PET detectors use avalanche photo diodes or silicon photo-multiplier (SiPMs) (Catana et al 2006, Maramraju et al 2011, Yamamoto et al 2012, Yoon et al 2012, Wehrl et al 2013, Weissler et al 2014, Kang et al 2015, Olcott et al 2015, Ko et al 2016) and are completely inserted into the open bore of the MRI gantry, between the gradient system and the transmit/receive MRI coil. However, the operation of an MRI system requires the use of strong radio frequency (RF) pulses. In a PET system designed for simultaneous PET/MR imaging, these pulses can easily disturb the PET system performance by introducing electronic noise. Thus, the entire PET/MRI setup must be carefully designed (Pichler et al 2006), and the sources for potential electro-magnetic system interferences must be understood, evaluated and minimized. Restrictions regarding the PET detector size must also be considered when the detector is between the MRI coil and the gradient system. In addition to the design considerations for the PET, the MR image quality and signal-tonoise ratio (SNR) are dominated by the quality of the transmitted and received RF signals. In MR imaging, the noise is mostly intrinsic, e.g. due to in-homogeneities of the magnetic field introduced by the object to measure or electromagnetic interferences (EMI) of the environment. Thus, it cannot be easily reduced (Mispelter and Lupu 2008). However, the amplitude of the received signal can be increased by reducing the distance between the coil and the object to be measured. Hence, similar to PET detectors, MRI coils should be as near to the object as possible 7358
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 1. Conventional PET insert (left) with a separated shielded MRI coil. In this
setup, the bore size dsys is limited by the free space between the MRI coil and its shielding. Our PET-insert approach with an integrated MRI coil (right): The PET bore size dsys is maintained, the shielding is located between the scintillation crystals and the light sensor (SiPM). A certain distance of the shielding to the MRI coil is provided, which prevents mutual EMI interferences.
(Hyde 2007). Typically, MRI coils are shielded to limit radiation losses and avoid parasitic coupling to common mode currents (Mispelter and Lupu 2008, Peng et al 2014). Furthermore, the shielding prevents the capacitive coupling to surrounding conductive parts. In the case of PET/ MRI, the MRI coil shielding also prevents the EMI of RF pulses on PET electronics (Yamamoto et al 2011), which may cause incorrect energy determination or fake events. To maintain the quality and homogeneity of the electro-magnetic field inside the RF coil, the shielding must be placed at a certain distance from the coil (Collins et al 1997, Mispelter and Lupu 2008). In cur rent small-animal PET/MRI systems, the shielded RF coil is separated and located in the PET bore. This setup reduces the useable bore size of the PET system (figure 1, left) and prevents an optimization of the solid angle coverage of the PET system by reducing the system diameter. Because most MRI sequences target 1H, the highest EMI on the PET electronics is expected at its Larmor frequency, which is approximately 300 MHz at a field strength of 7 T (Peng et al 2014). Assuming that the bandwidth of the electronics of a preclinical, non-time-offlight PET system is limited from 1 kHz to 200 MHz, the EMI issue is more critical at lower field strengths because the Larmor frequency may be in the bandwidth of the PET signals. However, MR gradient switching also causes EMI at lower frequencies, which depend on the MR sequence (Yamamoto et al 2011, Peng et al 2014, Wehner et al 2015). Most current preclinical PET/MRI systems (Judenhofer et al 2008, Yamamoto et al 2011, Yoon et al 2012, Hong et al 2013, Wehrl et al 2013, Wehner et al 2014) use standard shielded MRI coils. In addition, the entire PET front-end such as electronics, scintillators and light sensors, are commonly covered with a conductive material, which is connected to the system ground to avoid EMI. Another approach is to exclude the light sensors from the RF shielded housing of the PET electronics (Hong et al 2012, Yamaya et al 2015, Nishikido et al 2016). To achieve a high level of integration, compactness, maximization of the open combined system bore size and PET sensitivity without degrading the MR image quality, we propose a system design that integrates the MRI coil into the PET detector system. Because common PET scintillators are nonconductive and have low magnetic susceptibility (Strul et al 2003), they do not affect the quality of the electro-magnetic field. Thus, our approach (figure 1, right) is to radially assemble the MRI coil in the PET scintillator blocks. The shielding is placed between the scintillator blocks and the SiPM light sensor. Thus, a sufficient distance to the MRI coil is provided to avoid parasitic capture, which may detune the MRI coil. 7359
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
However, the shielding material must exhibit optimal EMI protection for the PET electronics, prevent the MRI from the noise introduced by the PET, and be transparent for the scintillation light. In this study, we evaluated this essential component of this innovative detector approach, which enables a high degree of integration for a PET detector and an MRI coil by testing three different shielding materials. Material and methods Several conductive materials that enable light transmission are commercially available. We compared three different materials, as shown in table 1 and figure 2. The performed tests measured the efficiency of an RF shielding, which was inserted between the PET scintillator crystals and the SiPM light sensor, on the PET detector performance parameters, such as the peak-to-valley ratio of the crystal position profile (PVR), energy resolution at 511 keV, and timing resolution. Furthermore, we assessed the shielding efficiency of the materials between 320 kHz and 420 MHz in comparison to a 9 µm-thick continuous layer of copper. Investigated shielding materials
• ITO: a polyethylene terephthalate foil evaporated with indium tin oxide. It is widely used in solar cell technology (Richard Woehr, Hoefen/Enz, Germany). • 9900: a polyethylene terephthalate foil coated with a thin conductive layer. It is commonly used to shield touch screens from electromagnetic interference (Holland Shielding Systems, Dordrecht, Netherlands). • mesh: a mesh of stainless steel wires, 100 wires per square inch, each having a diameter of 50 µm. The wire mesh was enclosed in polyethylene terephthalate and had an overall thickness of 280 µm. It was developed to shield displays and windows of RF cabinets such as MRI rooms or EMI measurement chambers (Holland Shielding Systems, Dordrecht, Netherlands). Further specifications of the evaluated samples with a common size of 51 × 110 mm2 are listed in table 1. As a reference, we used a printed circuit board that consisted of a core of 500 µm glass-reinforced epoxy laminate (FR-4), which was continuously covered with 9 µm of copper (Goettle Leiterplattentechnik, Koenigsbrunn, Germany). This material (CuPCB) was used to shield our conventional, custom-built, whole-body preclinical PET/MRI insert (Wehrl et al 2013). The analysis on the acquired data has been performed in MATLAB (MathWorks, Natick, MA, USA). The statistical significance of the measurements is stated as probability values (P-value). These were determined as follows: we began by testing each dataset for normal distribution by a One-sample Kolmogorov–Smirnov test with a significance level of 5%. In case of a positive decision on the normal distribution of all datasets within a category, we performed a t-test. In case of a negative decision on one of the datasets within a category we performed a non-parametric Wilcoxon-rank-sum test (Hollander et al 2015) for the entire category. PET performance measurements
The PET test setup is illustrated in figure 3. We used a PET detector that consisted of two stacked 12-by-12 cerium-doped lutetium oxyorthosilicate (LSO) matrices (Agile Engineering, 7360
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Table 1. Properties of the investigated optically transparent shielding materials.
Short description Typea Light transmissiona Sq. resistancea Resistanceb Attenuationa Layer thickness a b
% Ω/25.4 mm2 Ω/(51 × 110 mm2) dB (100 MHz) µm
ITO
9900
mesh
CuPCB
NV-CT-100 70 80–120 367.5 — 180
9900 68 50 136.5 50 120
9000 48 0.1 2.8 68 280
— 0 — 0.5 — 500
Manufacturer specifications. Measured by a 2-point ohmmeter.
Figure 2. Picture of the three RF-attenuation materials: ITO (left), 9900 (center), and mesh (right). To connect the materials electrically, they had a highly conductive region along 2 edges. For the ITO material it was made of copper, for 9900 and mesh silver was used. The attached copper stripe was used to connect the materials to the common ground of the RF attenuation setup. The reference material (CuPCB) for the RF attenuation is shown in figure 6.
Knoxville, TN, USA). The inner array, which was next to the FoV of the PET system, was shifted by half a pitch in both directions relative to the outer layer. The faces of all LSO pixels were 0.9 × 0.9 mm2; their thicknesses were 5.0 mm and 10.0 mm for the inner and outer crystal layers, respectively. The individual crystals of both layers were optically isolated by a wide-spectral-range reflective foil (Vikuiti, 3M, St. Paul, MN, USA). This foil was also attached to the top of the inner layer to reflect the scintillation light toward the outer layer and light sensor. The light-transmitting surfaces of the layers were optically coupled using optical grease (BC630, Saint-Gobain Crystals, Nemours Cedex, France). The scintillation light of the stacked LSO matrices was spread by a solid light guide (13.6 × 13.6 × 1.0 mm3, borosilicate glass) across a 4 × 4 SiPM array (S11828-3344M, Hamamatsu Photonics, Hamamatsu City, Japan). To distinguish the layers and determine the timing resolution, we used a 40.0 mm-long tungsten collimator with a slit of 1.0 × 25.0 mm2 and a reference detector. This detector consisted of a PMT module (H6779-01, Hamamatsu Photonics) with an optically coupled (BC630, Saint-Gobain Crystals) 3.5 × 3.5 × 20.0 mm3 LSO crystal. The scintillator was wrapped in three layers of white Polytetrafluoroethylene (PTFE) tape. The 16 SiPM elements of the block detector were operated at 73.5 V; this was 1.3 V above the average recommended operation voltage from the manufacturer. The output signals were passively multiplexed to four channels using a resistor network. These four signals were fed to current to the voltage converters (ADA4895, Analog Devices, Inc., Norwood, MA, USA) with a conversion factor of 182 V A−1. These four signals were fed to a summation circuit to provide an analog sum output. All output signals were digitized with 12-bit resolution and 7361
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 3. PET test setup for the coincidence measurements: The slit-collimated 22Na source irradiated the reference detector (PMT) and outer LSO layer of the SiPM block detector. The inner LSO layer was shifted by half a pitch in both directions relative to the outer layer. The scintillation light of the crystals was spread by a solid light guide across the 4 × 4 SiPM array. The RF shielding materials were placed between the light guide and the SiPM, whose signals were multiplexed by a resistive network into four channels and amplified. The four encoding channels, their sum signal, and the PMT signal were fed into the digitizer; the acquired data of coincident events were sent to a PC.
250 MS s−1 (V1720, CEAN, Viareggio, Italy). The PMT detector was directly connected to another digitizer channel. The digitizer was set to acquire only coincident events of the block detector and PMT module. Because of different propagation delays of both detectors, the coincidence time window was adjusted to 40 ns. After the signal exceeded a threshold of 9.8 mV, the waveform was digitized from 40 ns before to 160 ns after the trigger crossing (200 samples). First, all digitized waveforms were baseline-corrected using the average value of the first four samples. In this process, all 200 samples were integrated to determine the energy of the respective channel. For all digitized energy values we did not correct for the saturation of the light sensor. Subsequently, a software constant fraction discriminator (CFD) calculated the timestamp more precisely than the sampling speed of the digitizer (4 ns). The position of every event in the block detector was calculated using Anger logic (Anger 1958) and sorted into a crystal position profile. A crystal lookup map was generated using a Watershed algorithm (Roerdink and Meijster 2000) to discriminate individual crystals and compute the individual full width at half maximum (FWHM) energy resolution for the 511 keV photo peak. The PVRs were calculated from the position profiles for specific crystal rows and columns in the center and on the edge of the block detector. The maxima and minima were selected by a peak-finding algorithm. The timing resolution was computed from the timestamps of the sum channel of the SiPMs and the PMT. The PMT signal was neither shaped nor calibrated in time, and only a lower energy threshold of 9 keV was applied. The determined variation of the differences in time between the block detector and the PMT represented the timing resolution of the folded signals of both detectors. All measurements were performed with and without the respective shielding materials between the LSO scintillator block and the SiPM light sensor. To measure the relative attenuation of the scintillation light by the respective shielding materials, we determined the position of the full 511 keV energy peaks in a single crystal energy spectra and calculated the ratio of the 511 keV peak positions when the shielding mat erials were used relative to a measurement with no shielding material of a single, designated crystal. 7362
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
All measurements were performed in a dark environment at 25 ± 0.1 °C. To irradiate the detectors, a 22Na point source was used with a residual activity of approximately 950 kBq (Eckert & Ziegler, Berlin, Germany). Then, 106 coincident events were acquired for each measurement. The individual shielding materials were placed between the SiPM array and the light guide, as illustrated in figure 3. This study focused on comparing the PET detector performance with different shielding materials. Thus, we only investigated the outer crystal layer, which was closer to the light sensor. All PET measurements have been repeated three times. The optical coupling of the scintillator block, the light guide and the RF shielding mat erial was replaced before every measurement. RF attenuation performance of the shielding materials
To investigate the RF attenuation performance of the shielding materials we investigated the effect of an MRI sequence on the count rate of the PET detector block. The setup is depicted in figure 4. We used the same scintillator setup and electronics as in the PET performance evaluation. To increase the mechanical stability, we used solid transparent silicone rubber (EJ-560, Eljen Technologies, Sweetwater, TX, USA) with a size of 13.6 × 13.6 × 1.0 mm3 as a light guide. A nonconductive housing enclosed the detector setup to prevent the detection of light from the environment. The power supplies, the ADC and the PC were located outside the RF cabin of the MRI. Due to the ambient temperature of approximately 23 °C, we adjusted the bias voltage of the SiPM to 72.9 V. In this experiment we acquired the sum signal of the PET detector at a threshold of 9.8 mV. The average count rate was calculated based on the ADC data over 30 s during the MRI system was idle or running a Turbo RARE sequence. We used a 7 T MRI system (BioSpec 70/30, Bruker, Ettlingen, Germany). The MRI signals were transmitted and received by a surface coil (T6615V3, Bruker) that was located at a distance of approximately 1.5 mm to the detector setup. We included a solid, 1 mm thick phantom (water and 2% Agarose) on the coil. The respective shielding materials were placed on top of the detector setup. We used a Turbo RARE sequence with a repetition time (TR) of 800 ms, an echo time (TE) of 5.14 ms and a rare factor of 16, which lead to an effective TE of 30.82 ms. The power of the excitation pulse was set manually to 20 W. We did not use a γ-source for this experiment, the events in the PET detector were generated by the inherent radiation of the LSO crystals. All measurements have been repeated five times for the respective shielding materials. In a separate test setup, we evaluated the RF attenuation of the three shielding materials over a wider frequency range than the 7 T MRI system could provide. We placed the shielding materials in the center of two pick-up coils, which faced each other at a distance of 20 mm. Figure 5 shows a scheme of the setup. In MRI technology, pick-up coils are antennas that are commonly used as surface coils to receive RF signals. We used this type of coil to transmit and receive RF signals. The pick-up coils were built from a coaxial cable (RG-59/U, Belden, Inc., Indianapolis, PA, USA), where its core was formed to a circular loop with a diameter of 9 mm. The ground sheath of the cables was removed at the coils and connected to the shielding material in the test condition using a 5 mm-wide and 40 µm-thick copper stripe (Scotch 1181, 3M, St. Paul, MN, USA), which covered the entire longer edge of the material. We used a network analyzer (A6051, Agilent Technologies, Inc., Santa Clara, CA, USA) to determine the forward transmission coefficient, which was the level of signal that coil B received relative to the RF stimulus applied to coil A. As a stimulus, we used a sine wave with a frequency from 320 kHz to 410 MHz to have a broad coverage of the 1H Larmor frequencies at an MRI field strength up to 9.4 T. Because the materials under investigation were placed between the coils, the forward transmission coefficient was an indicator of achieved 7363
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 4. Setup to investigate the influence of a MRI sequence on the count rate of the
PET detector.
Figure 5. Scheme of the RF test setup. The shielding material was placed between two coils and connected to the common ground. A network analyzer determined the forward transmission coefficient between coil A and coil B over a frequency domain from 320 kHz to 420 MHz.
RF attenuation. All measured data were smoothed using a moving average algorithm. To determine the effect of the environment on our setup, we included a blank measurement with no shielding material between the coils. MRI compatibility To investigate the MR compatibility of the shielding materials, we performed measurements to determine the influence of the presence of the material on the B0 field of the MRI. These measurements include the RF noise spectrum and imaging sequences, such as echo planar imaging (EPI), diffusion weighted imaging (DWI), and magnetic resonance spectroscopy (MRS). We used a 7 T MRI system (BioSpec 70/30, Bruker). The system was equipped with a B-GA 20S gradient system. To adapt the test setup to our future mouse brain PET/MRI system, we placed the test material radially outside an MRI coil (T20063V3, Bruker). The MRI coil offered an inner diameter of 23 mm and was used in transmit- and receive-mode. For all measurements, we used a plastic tube with an inner/outer diameter of 9.7/11.8 mm filled with 7364
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 6. MRI compatibility setup: The foam inside the water phantom minimized the
amount and size of air bubbles in the water. During the measurements, additional foam was used to radially center the phantom in the MRI coil. The coil was used in transmit and receive mode. The RF shielding materials were located around the coil. As an RF shielding material, the picture shows exemplarily CuPCB. A Euro Cent is shown as a reference in dimension.
water, covering approximately 45 mm length. A piece of plastic foam was used to minimize the number and size of air bubbles in the FoV. The phantom itself was centered inside the MRI coil using foam. Figure 6 shows a picture of the test setup. The parameters of the used sequences are summarized in table 2. The effect of the shielding material and the PET detector on the MRI was investigated with the setup depicted in figure 4. We performed a spin echo sequence (SE) in which the RF power was set to zero. Thus, no RF pulse was applied; however, the MRI recorded the receive channel over a bandwidth of 50 kHz around the imaging frequency of ~300 MHz with a resolution of 763 Hz. The receiver gain was set to the same value for all measurements. This procedure has been performed twice, with the PET detector switched off and while it was powered on and acquiring events. To evaluate the effect of the shielding materials on the MRS, a 5 × 5 × 10 mm3 ROI in the center of the phantom was selected and a spectroscopy sequence was performed for this region. Within the absolute values of the resulted spectra, we calculated the FWHM and the full width at tenth maximum (FWTM) of the water peak. Additionally, we determined the SNR by the ratio of the peak value and the mean of 100 values at the left noise area of the spectra. The influence of the shielding material on the homogeneity of the main magnetic field (B0) was evaluated by a field map sequence. A FoV of 40.03 × 40.03 × 40.03 mm3 was scanned in an isotropic resolution of (0.417 mm)3. For every material, a B0 map was generated by averaging the signal intensity of every pixel over five repetitions of the measurement. Additionally, we performed a similar measurement without any additional material. To find the region of the phantom within the B0 dataset we performed a T1 sequence with an TE of 6 ms, a TR of 80 ms, a flip angle (FA) of 90° and the same resolution as the field map. The image of the T1 sequence was then discriminated by a threshold of 20% of the maximum signal intensity, and the contour of the same slice was overlaid onto the B0 map. Within this dataset, we selected the coronal center slice of the phantom in which we calculated the average and standard deviation of the magnetic field within the T1 contour. 7365
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Table 2. MR sequences and parameters.
Sequence
Repetition Echo time, time, TR (ms) TE (ms)
Turbo RARE 800 B0 field map 35 PRESS MRS 2500 SE DWI 1100 SE EPI 2000 a b
5.14/30.82a 2.34 16.60 42.00 30.00
Voxel size (mm3)
Matrix size
Slices
Flip angle, FA (°)
0.313 × 0.313 × 0.313 0.417 × 0.417 × 0.417 5.000 × 5.000 × 10.000 0.391 × 0.391 × 1.000 0.391 × 0.391 × 2.000
32 × 32 96 × 96 1 × 1 64 × 64 64 × 64
32 96 1 5 5
—b 30 90 90 90
Effective echo time, rare factor 16. The excitation pulse power was manually set to 20 W.
We also investigated the effect of the shielding materials on advanced imaging sequences, such as DWI and EPI. With the DWI method, we acquired five transversally adjacent slices in the center of the phantom covering a FoV of about 25 × 25 × 5 mm3. The voxel size was 0.391 × 0.391 × 1.000 mm3. We used a TR of 1100 ms and a TE of 42 ms. Five images have been acquired for b-values from 0 to 1000. Since we expected the highest impact on the SNR at the highest b-value, only this measurement is included in the results. We defined a signal-ROI of 12 × 12 voxels inside the phantom and determined the maximum signal intensity Smax and minimum value Smin. According to Wehrl et al (2011) we calculated the homogeneity H as follows: Smax − Smin H = 1− × 100%. Smax + Smin
For the determination of the SNR we defined an additional noise-ROI of 12 × 12 voxels at the lower right edge of the image. Afterwards, we calculated the ratio of the average signal intensity S of all pixels inside the signal-ROI and the standard deviation σ of the signal intensities in the noise-ROI. To account for the noise distribution of the magnitude images of the MRI (Henkelman 1985); the SNR was then calculated as: SNR = 0.655 ×
S . σ
The image analysis was performed on the center slice, the other slices have been controlled by visual inspection to ensure the comparability of the results. For the EPI sequence we also acquired five transversally adjacent slices in the center of the phantom to cover a FoV of 25 × 25 × 10 mm3. Per slice, 200 images have been generated. The TR was set to 2000 ms and the TE to 30 ms. We determined the drift of the signal intensity by averaging the values of the signal-ROI within the individual images. These averages were then plotted over the series of 200 EPI-images and a linear fit was applied. All imaging sequences have been applied five times. We determined the statistical significance of the measured parameters between the respective shielding materials and the measurements without any additional material. Results PET performance
One of the crystal position profiles and PVR histograms for the respective shielding material is shown in figure 7. We summarized the results for the PVR of the row and column in the 7366
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 7. Position profiles and PVR histograms of a reference measurement (A) with no additional shielding material. (B)–(D) Show the results with the ITO-foil, 9900foil, and mesh material, respectively. The data that has been used to generate the PVR histograms is indicated by the pathways in the position profiles. The circles identify the crystal that was used to investigate the attenuation of the scintillation light; see figure 8.
center and on the edge in average across three measurements in average and standard deviation in table 3. The energy resolution, the timing resolution and the photo peak position of all 144 crystals across the measurements are also listed in table 3. Figure 8 compares the energy spectra of the seventh crystal in the fourth row for each material, exemplarily. The respective crystal is indicated by a circle in the crystal position profile in figure 7. For this crystal, the average light attenuation and standard deviation across the measurements of ITO foil, 9900 foil, and mesh were 11.3 ± 1.6%, 11.0 ± 1.8%, and 54.3 ± 0.4%. For an overview, figure 9 compares the photo peak positions which have been derived from the individual energy spectra- of the 144 crystals for each material in equal color scale. The results of the influence of the MRI on the PET detector count rate are summarized as mean across five measurements and its standard deviation in table 4. The respective shielding material was used to shield the RF signals of the MRI (figure 4). The measurements in this section were performed without a precise temperature stabilization of the PET detector, this may have caused a slight fluctuation of the mean count rates. 7367
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Table 3. Average PET performance parameters across three repetitions of the same
measurement and their standard deviation.
Material
PVR center
PVR edge
Energy resolution (%, 511 keV)
non ITO 9900 mesh
5.69 ± 0.29 4.50 ± 0.39 4.78 ± 0.48 3.62 ± 0.16
5.36 ± 0.20 4.75 ± 0.27 4.39 ± 0.23 3.73 ± 0.25
15.73 ± 0.24 16.32 ± 0.13 16.60 ± 0.25 19.16 ± 0.21
Timing resolution (ns)
Photo peak position (a.u., 511 keV)
2.39 ± 0.06 2.54 ± 0.06 2.51 ± 0.07 2.80 ± 0.08
45663 ± 277 40359 ± 728 40369 ± 750 20997 ± 120
Figure 8. Single-crystal energy spectra of the shielding materials (ITO, 9900, mesh) and
a reference measurement (non) with no additional material. The respective crystals are marked by the circles in figure 7. The position of the full energy peak of the individual spectra indicates the attenuation of the scintillation light by the respective material. The photo peak positions are labeled in percent of the non-measurement.
RF attenuation performance
Figure 10 shows the RF attenuation performance of the tested materials versus frequency. From 320 kHz up to approximately 6 MHz, all materials (ITO, 9900, mesh, and CuPCB) had similar RF attenuations. From 6 MHz to approximately 200 MHz, the mesh shielding showed better attenuation than the CuPCB, except at approximately the 3 T Larmor frequency (~130 MHz). MRI compatibility
To address the MRI compatibility of the shielding materials we wrapped a commercial MRI coil in the respective material and performed MRI routines to evaluate the B0 field map, the RF noise, and MRS as well as imaging sequences as DWI and EPI. Figure 11 compares the results of the RF noise measurement for the on- and off-state of the PET detector. The average noise level and root mean square (RMS) across the spectra are summarized in table 5. Over all, no prominent peaks in the received signal could be observed, regardless of the shielding material used and the detector status. In all cases we could observe minor changes in the noise level and the RMS depending on the detector status. 7368
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 9. Photo peak position of all crystals for the respective materials ((B)–(D)) and
a measurement without any additional material (A) in equal color scale.
Table 4. PET detector count rate during the MRI is idle and running a Turbo RARE
sequence and while the MRI was in idle state. The respective shielding materials were placed between the MRI coil and the detector. The setup is illustrated in figure 4. Detector count rate (cps)
Material
MRI status
Mean ± SD
non
Sequence Idle Sequence Idle Sequence Idle Sequence Idle Sequence Idle
1431 ± 16 683 ± 21 670 ± 14 649 ± 42 716 ± 14 706 ± 28 657 ± 19 676 ± 7 611 ± 3 605 ± 6
ITO 9900 mesh CuPCB
7369
C Parl et al
Phys. Med. Biol. 62 (2017) 7357
Figure 10. RF attenuation level (L) of each optically transparent shielding material (ITO, 9900, mesh) versus frequency (f). As a reference, a 500 µm-thick epoxy core with a 9 µm-thick copper layer (CuPCB) and a blank measurement (non) with no shielding material are included. The frequency is logarithmically scaled from 320 kHz to 420 MHz. The vertical lines indicate the approximate Larmor frequencies of 1H at magnetic field strengths of 1.0, 1.5, 3.0, 7.0, and 9.4 Tesla.
Figure 11. RF noise spectra of the MRI receiver channel in arbitrary units (a.u.) over the resonance frequency f0 = ~300 MHz ± 50 kHz, acquired with no additional material around the MRI coil (non) and four RF-shielding materials: ITO, 9900, mesh, CuPCB. The receiver gain was constant at all measurements.
Exemplarily, figure 12 depicts one of the MRS spectra of the water peak for the respective shielding materials and a measurement without any additional material (non). The average SNR, FWHM, FWTM across five repetitions of the measurement for the respective shielding material are listed in table 6. The shielding materials did not degrade the width of the water peak and the SNR. The investigation of the homogeneity of the main magnetic B0 field is shown in figure 13. In all measurements, the variations over the upper 70% of the phantom were
