magnetic resonance imaging systems.

ORIGINAL ARTICLE

On the Quantification Accuracy, Homogeneity, and Stability of Simultaneous Positron Emission Tomography/Magnetic Resonance Imaging Systems Holger Schmidt, PhD,*Þ Nina F. Schwenzer, MD,* Ilja Bezrukov, Dipl-Inform,Þþ Frederic Mantlik, Dipl-Inform,Þþ Armin Kolb, Dipl-Ing,Þ Ju¨rgen Kupferschla¨ger, PhD,§ and Bernd J. Pichler, PhDÞ Objective: A potential major application of simultaneous avalanche photodiodeYbased positron emission tomography (PET)/magnetic resonance imaging (MRI) systems are quantitative brain studies for cerebral blood flow measurements in combination with blood-oxygen-levelYdependent or perfusion MRI, requiring a high performance for both modalities. Thus, we evaluated PET quantification accuracy and homogeneity for 2 different simultaneous PET/MRI systems (wholeYbody and brain scanner) compared with those of a state-of-the-art PET detector (PET/computed tomography) using phantom studies. In addition, we investigated the long-term stability of PET and quality of functional MRI measurements of a clinical whole-body PET/MRI scanner. Materials and Methods: Phantom measurements were conducted using spheres filled with [18F]-fluoride distributed in a homogeneous cylinder phantom at different positions inside the PET field of view. Recovery values and standard deviations were extracted from resulting PET images. The influence of magnetic resonanceYbased attenuation correction and that of activity outside the PET field of view on the recovery values of these spheres was evaluated. Furthermore, long-term PET stability of the whole-body PET/ MRI system was assessed by evaluating position profiles, energy spectra, count rates, and recovery values from [68Ge]-phantom scans. Functional MRI applicability was tested in accordance with the functional Biomedical Information Research Network procedure. Results: The BrainPET system showed high recovery values (up to 99%) but also increased variability (up to 7.4%). Significant underestimations in PET quantification near activity outside the PET field of view were found (up to 80%). Using magnetic resonanceYbased attenuation correction led to an underestimation in PET activity of approximately 7%. In distinction, the wholebody PET/MRI system revealed performance similar to the PET/computed tomographic scanner (recovery values up to approximately 60% with a variability of approximately 4%). Long-term stability and fMRI performance of the whole-body PET/MRI scanner showed no degradation compared with stand-alone systems. Conclusions: Homogeneity and accuracy of avalanche photodiode-based PET detectors is comparable with those of the state-of-the-art detectors based on photomultiplier tubes. However, attenuation correction on PET/MRI systems has to be adapted carefully for quantitative PET measurements. The BrainPET Received for publication July 23, 2013; and accepted for publication, after revision, November 1, 2013. From the *Diagnostic and Interventional Radiology, Department of Radiology, †Laboratory for Preclinical Imaging and Imaging Technology of the Werner Siemens-Foundation, Department of Preclinical Imaging and Radiopharmacy, University of Tu¨bingen; ‡Department of Empirical Inference, Max Planck Institute for Intelligent Systems; and §Nuclear Medicine, Department of Radiology, University of Tu¨bingen, Tu¨bingen, Germany. Conflicts of interest and sources of funding: Supported by grants PI771/5-1, PI771/3-1, and PI771/1-1 from Deutsche Forschungsgemeinschaft. The authors report no conflicts of interest. Supplemental digital contents are available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.investigativeradiology.com). Reprints: Holger Schmidt, PhD, Diagnostic and Interventional Radiology, Department of Radiology, University of Tu¨bingen, Hoppe-Seyler-St 3, 72076 Tu¨bingen, Germany. E-mail: [email protected]. Copyright * 2014 by Lippincott Williams & Wilkins ISSN: 0020-9996/14/4906Y0373

Investigative Radiology

& Volume 49, Number 6, June 2014

system needs improved scatter correction to perform quantitative brain studies. The whole-body PET/MRI scanner, however, is applicable for quantitative brain studies. Key Words: PET, MRI, simultaneous multimodality imaging, performance, stability (Invest Radiol 2014;49: 373Y381)

C

linical positron emission tomography (PET)/magnetic resonance imaging (MRI) is an emerging new hybrid imaging modality combining the highly sensitive molecular imaging capability of PET with the superior soft-tissue contrast and functional information from MRI. Today, clinical PET/MRI scanners already exist for sequential whole-body acquisitions.1 A simultaneous acquisition of different functional parameters using PET and MRI providing multiparametric data creates enormous possibilities to study pathology and biochemical processes,2,3 and first evaluations on different clinical questions were already performed.4Y8 However, a challenge for simultaneous PET/MRI is the prevention of any adverse interaction between the 2 modalities.9 One approach is to use solid-state photosensors (avalanche photodiodes [APDs]) instead of photomultiplier tubes (PMTs) for light detection in the PET detector.10Y13 Although clinical prototype systems exist for simultaneous imaging of the brain since 2008 (BrainPET),14 the first fully integrated whole-body PET/MRI scanners are just entering the clinical field.15 Performance measurements reveal overall good results comparable with those of the state-of-the-art PET detectors and MRI scanners.15,16 However, only very little experience exist on the longterm stability of large-scale APD-based PET scanners or the image quantification homogeneity over the entire field of view (FoV). This is specifically important because the gain of APDs depends strongly on small temperature or bias voltage shifts.17 Especially for the high-resolution BrainPET, the detection of small lesions is an important clinical application. Furthermore, simultaneously acquired PET and magnetic resonance (MR) images need to provide high spatial and quantitative accuracy as well as an excellent temporal coregistration to enable the correlation of functional MRI (fMRI) and complementary PET imaging data. To achieve the required high level of quantification accuracy for such studies, the homogeneity of quantification over the entire PET FoV is essential. Our studies focused specifically on the feasibility of quantitative brain studies using short half-life radiotracers such as [11C]labeled neuroreceptor ligands or [15O]-labeled water for cerebral blood flow measurements. To assess this feasibility, we evaluated the quantification accuracy and variation of small lesions within the PET FoV by phantom studies carried out on a brain PET/MRI, a dedicated clinical brain scanner, and a fully integrated whole-body PET/MRI system. Furthermore, the presence of activity outside the PET FoV can affect the PET performance; thus, we investigated this influence using a scatter phantom placed outside the PET FoV. This issue is especially important for dynamic PET scans with short half-life radiotracers using www.investigativeradiology.com

Copyright © 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

373


Schmidt et al

high initial activity concentrations. The results were compared with those of a state-of-the-art PET/computed tomography (CT) scanner. Whereas the results on the stability of the brain PET/MRI are reported elsewhere,16 stability tests were accomplished for the first time for the whole-body PET/MRI system.

MATERIALS AND METHODS Systems The quantification accuracy and homogeneity of the brain PET detector, the whole-body PET/MRI system, and a PET/CT scanner were evaluated. Basically, all PET detectors are based on lutetium oxyorthosilicate (LSO) scintillation crystals with a length of 20 mm. The PET/MRI systems use APDs as scintillation light detectors, whereas the PET/CT uses conventional PMTs.

Brain PET/MRI The human BrainPET Insert (Siemens Healthcare, Knoxville, TN) consists of 32 detector cassettes, each comprising 6 LSO/APD block-detectors in the axial direction. A total of 192 matrices of 144 crystals with a crystal size of 2.5 2.5 20 mm3 yield a 32-cm transaxial and 19.2-cm axial FoV. The insert was placed inside the magnet bore of a clinical 3-T MRI scanner (MAGNETOM TIM Trio; Siemens Healthcare, Erlangen, Germany).

Whole-Body PET/MRI The PET detector of the whole-body PET/MRI system (Biograph mMR; Siemens Healthcare, Erlangen, Germany) consists of 56 radially arranged detector cassettes, each comprising 8 LSO/ APD block-detectors in the axial direction. A total of 448 matrices of 64 crystals with a crystal size of 4.0 4.0 20 mm3 yield a 59.4-cm transaxial and 25.8-cm axial FoV. The MRI scanner is a modified 3-T Verio system (MAGNETOM Verio; Siemens Healthcare, Erlangen, Germany). Because of the full integration of the PET detector into the MRI, the diameter of the magnet bore is reduced to 60 cm.

PET/CT The time-of-flight (ToF) PET detector of the PET/CT scanner (Biograph mCT, Siemens Healthcare, Knoxville, TN) consists of 48 detector cassettes arranged in a ring, each comprising 4 LSO/PMT block detectors in the axial direction. A total of 192 matrices of 169 crystals with a crystal size of 4.0 4.0 20 mm3 yield a 70-cm transaxial and a 21.8-cm axial FoV. The 128-slice CT scanner has an aperture of 78 cm.

PET Image Reconstruction Standard clinical reconstruction parameters using the vendorprovided algorithms and optimal reconstruction parameters (determined from the routine protocol of National Electrical Manufacturers Association [NEMA]) were applied for image reconstruction. For the BrainPET, the acquired PET data were reconstructed with an iterative 3-dimensional ordered-subset expectation maximization (OSEM) algorithm using 8 iterations and 16 subsets into a 256 256 matrix (1.25 1.25 1.25 mm3 voxel size). For attenuation correction, the CT-based attenuation map was used because no MR-based attenuation map was generated here. For the whole-body PET/MRI, acquired PET data were reconstructed with an iterative 3-dimensional OSEM algorithm using 3 iterations and 21 subsets into a 344 344 matrix (2.08 2.08 2.03-mm3 voxel size). The images were smoothed by application of a Gaussian filter of 3 mm and corrected for attenuation using an MR-based attenuation map (for PET quantification accuracy and 374

www.investigativeradiology.com


homogeneity measures only) and a CT-based attenuation map. For creating the latter map, the CT image of the phantom was used to create a CT-based attenuation connection map via MATLAB (The Mathworks Inc, Natick, MA). All CT-related objects (patient bed) were removed, and the plastic housing of the spheres was replaced by the mean Hounsfield unit (HU) value of the phantom background (0 HU) to prevent possible attenuation correction errors due to slight misalignments. The CT image was rigidly registered to a 3-dimensionalYencoded T1-weighted gradient echo sequences also used for the creation of the attenuation maps using the software 3D SLICER.18 The HU units were translated into linear attenuation coefficients according to Carney et al.19 For the PET/CT, acquired PET data were reconstructed with an iterative 3-dimensional point spread function-OSEM algorithm with ToF information using 2 iterations, 21 subsets, and a Gaussian filter of 2 mm into a 400 400 matrix (1.75 1.75 1.75-mm3 voxel size). The choice of PET reconstruction parameters influences the quality of reconstructed images. Thus, to minimize the bias in the evaluation of the different PET detectors, we used the PET reconstruction parameters that are also used in clinical routine for each scanner. These parameters are found to be the optimal values regarding image recovery according to NEMA measurements for each scanner.

Measurements and Evaluation PET Quantification Accuracy and Homogeneity Eight hollow spheres (2 spheres with inner diameters of 4, 5, 6, and 8 mm each) were filled with 70 kBq/mL of [18F]- fluoride solution. The spheres were placed at different positions inside a homogeneous cylinder phantom (diameter, 21 cm; length, 17.5 cm). The cylinder was filled with 8.75 or 17.5 kBq/mL [18F]-fluoride solution yielding a sphere-to-background ratio of 8:1 and 4:1, respectively. The PET images of the phantom were acquired at 5 different phantom positions (reference position; 90-degree axial counterclockwise rotation; flip along the z axis; 20-degree caudocranial tilt; 3-cm movement along the z axis) to evaluate the recovery for each sphere at different positions in the PET FoV. The phantom was positioned inside the MR head coil for both PET/MRI systems. The acquisition time for each position was 15 minutes (Supplemental Digital Content 1, http://links.lww.com/RLI/A130). The recovery values of the spheres depending on location and size were estimated from 3-dimensional 80% isocontour volumes of interest (VOIs) of the [18F] decay-corrected images. The VOIs were drawn in the resulting PET images using the software package PMOD (version 3.2; PMOD Technologies Ltd, Zurich, Switzerland). Recovery values were calculated by dividing the measured activity concentration by the decay-corrected true activity concentration. The background activity was measured by drawing regions of interest (ROIs) with a diameter of 6 cm at 3 different positions (caudal, center, cranial) inside the phantom and calculating the mean activity concentration of these ROIs.

Influence of Activity Outside the PET FoV Two phantoms were used in the experiment: a NEMA scatter phantom20 containing the activity outside the PET FoV and a homogeneous cylinder phantom. The scatter phantom was filled with 600 MBq [11C]-choline and placed in caudal direction close to the cylinder phantom. Two hollow spheres (inner diameters of 8 mm) were filled with 70 kBq/mL and the cylinder with 8.75 kBq/mL [18F]fluoride solution to yield a sphere-to-background ratio of 8:1. The spheres were placed in the radial center of the previously described homogeneous cylinder phantom. The phantom was placed into the PET FoV of the 3 scanners such that 1 sphere was located in close * 2014 Lippincott Williams & Wilkins




neighborhood of the activity source outside the PET FoV (caudal position), whereas the second sphere was located further away (approximately 15 cm, cranial position). Data were acquired for 2 hours in list mode and retrospectively rebinned into 5-minute frames. Three-dimensional 80% isocontour VOIs and background ROIs were drawn in the [18F] decay-corrected images, and recovery values were calculated in the same way as described in the previous experiment. The relationship between the background activity and the recovery values was quantified by the slopes of linear regression lines. The linear regression was calculated separately for each sphere.

Evaluation of Dead-Time Correction Since high levels of activity outside the PET FoV influenced the recovery measurements for the BrainPET, we applied an additional test for this system only. The aim of this experiment was to investigate whether the dead-time correction or the scatter correction was the cause for the changes in the recovery measurements. The same homogeneous cylinder phantom used for the previous experiments was used. The activity levels inside the spheres and the background (8:1 sphere-to-background ratio) were initially adapted to yield the same PET single event rate as in the previous experiment (3.5 107 counts per second). Fifteen-minute PET data were acquired in list mode after 0, 110, 220, and 330 minutes and were rebinned into 5-minute frames retrospectively. Three-dimensional 80% isocontour VOIs and background ROIs were drawn in the [18F] decay-corrected images, and recovery values were calculated in the same way as described in the first experiment. Mean recovery and standard deviations (SDs) were calculated for the 3 frames of each measurement.

PET Long-Term Stability of the Whole-Body PET/MRI System To investigate the long-term stability of the PET modality, a 5-minute scan of a [68Ge]-cylinder phantom (37 MBq, 10-cm diameter) was carried out 12 times (in the morning, afternoon, and evening on Friday as well as on the following Monday, Wednesday, and Friday). In addition, the procedure was carried out before and after the daily quality check on Monday to get an indication of possible alterations during a longer period over the weekend. The same phantom is used for the daily quality control. Prompt, random, and true counts as well as the single count rate were determined for the entire detector and for each of the 224 buckets (2 block detectors each) and corrected for decay. The SDs of the 14 measurements were determined. The mean recovery and the SD were calculated by drawing an ROI in each slice of the resulting PET image such that 70% of the axial phantom dimension was covered. The outer 5 slices were discarded because of the increasing noise at the edges of the PET FoV. In addition, the energy spectra of each crystal and position profiles of each detector block were measured. After superposition of the energy spectra from the 14 measurements for each crystal, all superposed energy spectra were visually inspected. Quality criteria were the separability of photopeak and Compton edge, the Gaussian-like shape of the photopeak, as well as the overall shape of the spectrum including height and position of the photopeak. For each row and column of each detector block, the corresponding position profiles from the 14 measurements were superposed and visually inspected regarding height and position of each peak within the row or column. The influence of MR measurements during PET acquisition was also checked with this method at 3 different time points (using a T1-weighted turbo spin echo sequence: repetition time/echo time, 580 milliseconds/12 milliseconds; excitation angle, 160 degrees; bandwidth, 269 hertz per pixel). * 2014 Lippincott Williams & Wilkins

Accuracy, Homogeneity, and Stability of PET/MRI

MRI Stability of the Whole-Body PET/MRI System Before performing fMRI studies, the MR stability should be tested according to the functional Biomedical Informatics Research Network (fBIRN) protocol.21 Thus, a stability phantom (17-cm sphere filled with agar gel) was scanned using an echo planar imaging sequence for blood-oxygenation-levelYdependent effect measurements (repetition time/echo time, 2000 milliseconds/30 milliseconds; excitation angle, 90 degrees; bandwidth, 1420 hertz per pixel; FoV, 220 mm; matrix, 64 64; 28 axial slices [4 mm + 1 mm of gap]; 200 time frames) with the dedicated 16-channel head/neck coil. The measurement was carried out 3 times with idle PET system (to evaluate the variability of the measurements), once with [68Ge]-rod sources inside the PET FoV (57 MBq, positioned outside the head/neck coil) and with the PET system turned off (detector voltage-free) to evaluate the influence of the PET system on the fMRI stability. A 20 20-voxel ROI was drawn in the center of the phantom image. The root mean square stability was calculated from the ROI values of 198 time frames (the first 2 frames were discarded) after a second-order detrending fit as percent fluctuation; the intensity drift was also calculated from the fit ([maximum ROI value Y minimum ROI value]/ mean ROI value 100). In addition, 4 images were calculated: an average image by voxelwise averaging the 200 time frames, an SD image by calculating the SD for each voxel of the 200 time frames, a noise-averaged image by subtracting the average image of the odd numbered frames from the average image of the even numbered frames, and a signal-tofluctuation-noise ratio (SFNR) image by voxelwise division of the average image by the SD image. Average signal intensity was obtained from a 20 20-voxel ROI in the average image, a noise value is received from the same ROI in the noise-averaged image, and the SFNR value from the ROI positioned in the SFNR image. Division of the average signal intensity by the noise value yielded the signal-to-noise ratio. To assure that the measured voxel noise is uncorrelated, a Weisskoff plot was performed.22 Here, the SD of a centered ROI is measured as a function of ROI size and a correlation diameter (radius of decorrelation) is estimated from a comparison of the measured limiting SD value to an ideal uncorrelated value. All calculations were carried out using MATLAB.

RESULTS PET Quantification Accuracy and Homogeneity The BrainPET revealed high recovery values (up to 99.0% for 8-mm spheres at 8:1 sphere-to-background ratio) and thus allows for the detection of smaller lesions. However, the variability of recovery values (up to 7.4%) for different sphere positions in the PET FoV was higher compared with the PET/MRI (up to 4.1%) and PET/CT (up to 4.3%) systems (Fig. 1A). The 5- and 4-mm spheres were not detectable for the 4:1 sphere-to-background ratio. In PET images of the whole-body PET/MRI and PET/CT, the 4-mm spheres could not be detected already in the 8:1 sphere-to-background ratio. Recovery values were similar for both whole-body systems but lower than for the BrainPET scanner (mean of 60.6% and 59.1%, respectively, for the 8-mm spheres at 8:1 sphere-to-background ratio; Figs. 1B, C). Using MR-derived attenuation maps for the PET image reconstruction led to an underestimation in PET quantification (up to 7.2% for our phantom) for PET/MRI as becoming apparent in the reduced recovery in the background ROIs (Fig. 1D). The MR- and CT-based attenuation correction maps are displayed in Supplemental Digital Content 2, http://links.lww.com/RLI/A131; recovery values are also given in Supplemental Digital Content 2, http://links.lww.com/RLI/A132. In addition, a slight gradient in background recovery values of approximately 2% (whole-body PET/MRI) or 3% (BrainPET and www.investigativeradiology.com


375


Schmidt et al


FIGURE 1. Recovery values for the 8 spheres and the background for the 5 different phantom positions inside the PET FoV (color gradient) from the BrainPET data (A), the whole-body PET/MRI system (B), and the PET/CT scanner (C). On the left, results for a sphere-to-background ratio of 8:1 are shown, and on the right, a sphere-to-background ratio of 4:1 was used. D, The recovery values for the whole-body PET/MRI system are shown, when an MR-based attenuation map for PET attenuation correction is used. Note the decreased recovery values in the background.

PET/CT) between the top and bottom parts of the phantom was found, most probably because of the increased attenuation and scatter effects of the table.

Influence of Activity Outside the PET FoV No influence of high activity concentrations outside the PET FoVon the spheres recovery values could be found for the whole-body PET/MRI and PET/CT systems (adjacent [distant] sphere mean [SD] recovery values of 65.3% [4.0%] [62.5% {3.6%}], slope of 0.04 [0.04] and 60.3% [3.9%] [57.8% {3.2%}], slope of 0.05 [0.05], respectively). Only a slight increase in the background activity could be found for the PET/CT at high activity concentrations in images slices adjacent to the outside activity (Figs. 2D, F). For the BrainPET, a clear correlation between activity outside the PET FoV and the recovery values as well as the measured background activity was observed (Fig. 2A). The mean (SD) recovery values were 87.3% (15.5%) for the caudal and 94.7% (8.6%) for the cranial sphere, with a slope of 0.32 and 0.16, respectively. Also, the background activity was decreased to approximately 80% for the first image frame adjacent to the outside-FoV activity source (Fig. 2B). Two possible explanations for these findings are most probable: an overcorrection of scatter events or an erroneous dead-time correction. In both cases, an underestimation of PET activity concentration in the vicinity of the outside activity source can be expected. 376


Evaluation of BrainPET Dead-Time Correction To rule out any influence by dead-time correction errors, the same experiment was repeated without the presence of any activity outside the PET FoV but with an activity concentration inside the FoV accounting for the same single count rate as the previous experiment. Assuming that the dead-time correction is erroneous, the results should be identical to the previous experiment because the same dead-times are reached. Otherwise, the scatter correction must be erroneous and cause an overcorrection of the scattered events. The result shown in Figure 3 reveals no correlation of sphere recovery values to the activity concentration inside the PET FoV. This indicates the overcorrection of the scattered events to be the reason for the influence of the outside activity on the PET quantification accuracy. This is supported by the inspection of the sinogram data shown in Supplemental Digital Content 2, http://links.lww.com/RLI/A133.

PET Long-Term Stability of the Whole-Body PET/MRI System The PET detector of the whole-body PET/MRI system showed good long-term stability. The SD of the prompt, random, and true counts as well as that of the single count rate for the 14 measurements was 0.10%, 0.10%, 0.11%, and 0.03%, respectively (Fig. 4A). The SD of the recovery values was found to be 0.3%, which is less than the mean ROI interslice variability of 0.7% (Fig. 4B). In addition, the count rates of the detector buckets showed a mean relative SD of * 2014 Lippincott Williams & Wilkins





FIGURE 2. On the left, recovery values for the spheres adjacent and distant to the outside-FoV activity source are shown as a function of outside-FoV activity concentration (A, C, E). On the right, the mean ROI values per image slice for 6 measurements are depicted (C, D, F). Note that there is dependence of the recovery values on the outside-FoV activity concentration for the BrainPET (A), whereas there is no or only marginal dependence for the whole-body PET/MRI (C) or PET/CT (E), respectively.

0.2% for the 14 measurements, with a maximum relative difference of 1.0% (bucket 48, measurement 2; Fig. 5). The energy spectra and the position profiles were of good quality and revealed no distinct deviations between the different measurements. A typical flood map, energy spectra, and position profiles are shown in Figure 6. Particularly, no influence on counts, count rates, or recovery values by simultaneous MRI could be found.

MRI Stability of the Whole-Body PET/MRI System The results of the fMRI quality assurance measurements are given in Table 1 and Supplemental Digital Content 2, http://links.lww.com/RLI/A134. Percentage fluctuation and drift are well below the recommended optimal values (0.1% and 0.4%, respectively). The signal-to-noise ratio and SFNR show similar values and are well above the recommended value of 200. Also, the radius of decorrelation indicates low autocorrelated noise of the system. No obvious PET-related changes could be found. FIGURE 3. Recovery values of the 2 spheres for decreasing activity concentration inside the BrainPET FoV (decreasing dead-time) without any outside-FoV activity. Similar to Figure 2A, no dependence of the recovery values on the activity concentration (and thus the dead-time) could be found. * 2014 Lippincott Williams & Wilkins

DISCUSSION PET/MRI is an ideal modality to study brain activation and function in neurodegenerative diseases as well as neurological and www.investigativeradiology.com


377


Schmidt et al


and relationship were contradictory even if the experiment setup was designed as similar as possible, maybe because the lack of temporal correlation led to physiological fluctuations or because of spatial misalignments of the data.23Y28 With the possibility of simultaneous PET and MRI acquisition, PET/MRI might become a perfect tool for the investigation and correlation of functional data from MRI and PET especially in brain studies. However, increased requirements on the stability of both modalities are needed for such investigations to identify small changes in fMRI signal of down to 0.8%23,24 and to acquire PET data at high initial activity

FIGURE 4. A, Total prompt, random, and true counts as well as the single count rate for the stability phantom scans. B, The mean recovery values from the ROI analysis inside the phantom image of these scans is shown. M indicates Monday; W, Wednesday; F, Friday; m, morning; a, afternoon; e, evening; superscript M, with simultaneous MR measurement; superscript Q, before daily quality check.

behavioral deficits, for neurocognitive and addiction research, monitoring of brain plasticity, or identification of functional important brain areas before surgical interventions. Many brain studies in animals and humans have been carried out in the past to correlate cognitive skills and impairments to PET water perfusion or MRI blood-oxygenationlevelYdependent data and to compare both methods. However, comparisons of PET and MRI data regarding spatial correlation, volume,

FIGURE 5. Single count rates per bucket (2 detector blocks each) for the 14 stability scans shown in Figure 4. Each scan has a different color code. Buckets are sorted according to the count rate of the first measurement (black dots). 378


FIGURE 6. A, Representative flood image of an LSO-APD PET block detector of the whole-body PET/MRI system. B, Profiles of 1 crystal row (white box in A) through the images of the 14 measurements are shown. C, Energy spectra of the 14 measurements for a center crystal (red box in A) are shown. * 2014 Lippincott Williams & Wilkins




TABLE 1. Results of the fMRI Quality Assurance Measurements at Different PET States

% fluctuation Drift Mean intensity SNR SFNR RDC

PET Idle 1

PET Idle 2

PET Idle 3

PET Activity

PET Off

0.07 0.28 1472.1 358.9 344.7 4.5

0.07 0.18 1483.9 372.0 343.5 4.6

0.07 0.23 1482.6 351.9 346.4 4.3

0.07 0.24 1488.5 340.8 350.2 4.2

0.07 0.16 1478.7 352.5 346.5 4.3

fMRI indicates functional magnetic resonance imaging; PET, positron emission tomography; RDC, radius of decorrelation; SFNR, signal-to-fluctuation-noise ratio; SNR, signal-to-noise ratio.

concentrations of several hundred MBq27,29,30 with small relative signal changes as low as 8% for [15O]-water perfusion PET scans.25 Whereas the overall performance results of simultaneous PET/ MRI scanners have already been reported,15,16 PET stability at high activity concentrations as well as the quantification accuracy and homogeneity were evaluated and compared with those of a state-of-theart PET detector in this study. In addition, the PET quantification homogeneity and long-term stability as well as the stability of f MRI measurements were tested for the whole-body PET/MRI system. Regarding the PET quantification accuracy and homogeneity of small lesions inside the PET scanner, the BrainPET system reveals a higher sensitivity for small lesions (96.4%, 63.5%, 31.3%, 19.0% for the 8-, 6-, 5-, and 4-mm spheres at 8:1 sphere-to-background ratio for BrainPET versus 66.6%, 38.4%, 23.4% for PET/MRI and 59.1%, 33.3%, 16.7% for PET/CT, respectively; the 4-mm spheres were not visible for PET/MRI and PET/CT; see Supplemental Digital Content 2, http://links.lww.com/RLI/A132), which can be explained by the higher resolution16 of up to 2 mm instead of 4 mm for the wholebody PET/MRI and PET/CT15,31 systems. However, the variability of recovery values is increased for the BrainPET as can be seen by the increased SD of the determined recovery values for the different spheres (1.1% to 7.4% for BrainPET, 1.0% to 3.7% for PET/MRI, and 1.0% to 4.3% for PET/CT; see Supplemental Digital Content 2, http://links.lww.com/RLI/A132). These differences in the recovery values are independent of position and rather caused by a global instability of the system, most probably because of small temperature variations in the cooling system of the APDs. This hypothesis is also supported by the increased variability of the background recovery values for the BrainPET in the experiments with activity outside the PET FoV (using a phantom in a fixed position) compared with the whole-body PET/MRI and PET/CT (Fig. 2). When comparing whole-body PET/MRI and PET/CT, there is no large difference in recovery values and their variability between these systems. This is specifically of importance because the PET detector of the PET/MRI is based on APDs, whereas the detector of the PET/CT is based on conventional PMTs. Of course, the use of different reconstruction parameters for the 3 scanners makes an exact quantitative comparison difficult. Nevertheless, because we used the optimal parameters for all scanners regarding recovery values, we compare the best achievable performance between the scanners. Thus, semiconductors are able to replace bulky PMTs. However, an important issue that has to be taken into account is the attenuation correction for phantom experiments. Because the acrylic glass parts of the phantom are not detected by the standard method for segmentation-based attenuation map creation from MRI data, an underestimation of PET attenuation will occur, which shifts the measured PET activity concentrations in the images toward lower values. In our case, the underestimation in recovery values measured * 2014 Lippincott Williams & Wilkins


for background activity was up to 7% (compare Fig. 1B and D) and can be even higher using larger phantoms. This has to be taken into account when performing quantitative PET measurements (also for the NEMA protocol). Thus, a CT-based attenuation map has to be acquired, carefully registered to the MRI data, and used for PET image reconstruction. This approach, on the other hand, necessarily alters the procedure from the standard method used for patient studies. In brain PET/MRI, similar problems regarding attenuation correction will occur. Because bone is ignored in the standard segmentation-based method of attenuation map creation from MR data, an underestimation in PET activity is also present in this case. Several methods have been proposed to overcome this problem32Y34 and have to be used for quantitative PET measurements. In addition, for functional brain studies, an accurate attenuation correction of the MR head coil and other ancillary objects in the PET FoV is important for quantitative PET images.35 Thus, because recovery values and deviations for whole-body PET/MRI are comparable with those of the state-of-the-art PET detectors, quantitative brain studies are possible on this system given the correct attenuation correction including bone. The BrainPET system revealed higher recovery rates allowing for the detection of smaller lesions or regions within the brain. However, the increased SD of approximately 7% could impair quantitative brain studies because typical changes in brain activation signals are in the range of 8% to 40%.23,25 High activity concentrations outside the PET FoV had no influence on sphere activity quantification for whole-body PET/MRI and only a slight influence for PET/CT. However, for the BrainPET system, an activity-dependent recovery gradient was found, resulting in an underestimation of up to approximately 80% in the vicinity of the outside activity source. Possible reasons could be scatter correction or dead-time correction. The latter could be ruled out by measuring the same count rates without any activity presence outside the PET FoV revealing no count rateYdependent recovery gradient in the resulting PET images. Thus, the scatter correction of the BrainPET system is erroneous, leading to large underestimations in PET activity concentrations. The scale of underestimation depends on the amount of and distance to the activity source outside the PET FoV. Because underlying scatter correction algorithm is similar for all 3 scanners (single scatter simulation; estimating scattered events along each line of response from emission and attenuation sinograms36), the reason for these problems in scatter correction could be caused by inaccuracies in the adaption of the algorithm to the scanner geometry and the correct consideration of attenuating objects in the PET FoV such as the MR head coil. The correction of scatter from outside the PET FoV is realized by scatter scaling.37 Here, activity regions in the uncorrected emission sinogram are masked according to the attenuation sinogram and the scatter sinogram is then scaled to the masked emission sinogram. An overestimation of scatter scaling near the source of activity outside the PET FoV (eg, by wrong masking due to incorrect consideration of attenuating objects in the PET FoV, such as the MR head coil) might explain the found underestimation in recovery. A workaround could be the use of a lead ring around the patient at the front end of the BrainPET to reduce scatter from outside the PET FoV as supposed by Herzog et al.38 The slight reduction in background activity close to the scatter phantom for high activity concentration in the images from PET/CT might also result from dead time correction or, most probably, also from scatter correction, which is more difficult to achieve for larger bore sizes. Thus, although there is no impairment due to high activity levels outside the PET FoV for the whole-body PET/MRI system, the scatter correction of the BrainPET system has to be optimized before quantitative brain studies are possible. With the current setup, a stable measurement is only possible at outside activities below www.investigativeradiology.com


379


Schmidt et al

approximately 150 MBq (corresponds to 40 minutes in Fig. 2), which is a severe limitation for all [11C]- and [15O]-based tracers. The superposition of the columns and rows of the 14 measurements from all flood maps revealed no obvious deviation in amplitude and position of each peak. Also, the superposition of the energy spectra of each crystal showed good separability of photopeak and Compton edge as well as no obvious differences in the overall shape of the spectrum including amplitude and position of the photopeak between the different measurements. Thus, no intrinsic problems in PET detector stability of the whole-body PET/MRI system could be found, such as temperature variations of the APDs, which could result in gain variation. Thus, reliable PET measurements can be performed. This is supported by the results from count rates and recovery values obtained in the PET images revealing very low deviations between the different measurements (approximately 0.1% SD for the coincidence counts, 0.03% SD for the single count rates, and 0.3% SD for the recovery values). Furthermore, no differences could be found for measurements with and without simultaneous MR scans. These results are also in agreement with that of Delso et al.15 Also, the single count rates per bucket show only a small variance between measurements (0.2% SD). However, it has to be noted that the single count rates were analyzed per bucket and not per detector block. Thus, the variance per block detector will be somewhat higher. In addition, the fBIRN protocol revealed values within the recommendations. A comparison of the fBIRN results with and without active PET component indicates that the operation of the PET detector does not influence fMRI results. Thus, PET and MR scanners of the whole-body PET/MRI system reveal a high stability allowing for quantitative PET and fMRI measurements. In conclusion, the homogeneity and accuracy of APD-based PET detectors are comparable with those of the state-of-the-art PMTbased detectors. The quantification accuracy of both evaluated PET/ MRI systems for small lesions at different positions in the PET FoV is acceptable for quantitative brain studies where the homogeneity of quantification on the entire FoV is essential. The BrainPET quantification accuracy for small lesions is higher than that for the whole-body PET/MRI and the PET/CT system; however, an increased variation in recovery values was found. Whereas the BrainPET system shows severe problems with higher amounts of activity outside the PET FoV, most probably because of erroneous scatter correction, the stability of both PET and MRI in the whole-body PET/MRI system is comparable with that of the state-of-the-art detectors. For the BrainPET system, scatter correction has to be improved before brain activation studies will become possible, whereas the whole-body PET/MRI system allows for quantitative brain studies in general. However, the attenuation correction for phantom experiments and human brain studies on PET/MRI systems has to be adapted carefully for quantitative PET measurements. ACKNOWLEDGMENTS The authors thank Ralf Ladebeck for his technical support as well as Larry Byars and Christian Michel for their discussion about the scatter correction algorithm. REFERENCES 1. Ratib O, Becker M, Vallee JP, et al. Whole body PET-MRI scanner: first experience in oncology. J Nucl Med. 2010;51(suppl 2):165. 2. Schlemmer HP, Pichler BJ, Krieg R, et al. An integrated MR/PET system: prospective applications. Abdom Imaging. 2009;34:668Y674. 3. Nekolla SG, Martinez-Moeller A, Saraste A. PET and MRI in cardiac imaging: from validation studies to integrated applications. Eur J Nucl Med Mol Imaging. 2009;36:S121YS130.

380



4. Gaertner FC, Beer AJ, Souvatzoglou M, et al. Evaluation of feasibility and image quality of 68Ga-DOTATOC positron emission tomography/magnetic resonance in comparison with positron emission tomography/computed tomography in patients with neuroendocrine tumors. Invest Radiol. 2013;48:263Y272. 5. Hirsch FW, Sattler B, Sorge I, et al. PET/MR in children. Initial clinical experience in paediatric oncology using an integrated PET/MR scanner. Pediatr Radiol. 2013;43:860Y875. 6. Bisdas S, Ritz R, Bender B, et al. Metabolic mapping of gliomas using hybrid MR-PET imaging: feasibility of the method and spatial distribution of metabolic changes. Invest Radiol. 2013;48:295Y301. 7. Platzek I, Beuthien-Baumann B, Schneider M, et al. PET/MRI in head and neck cancer: initial experience. Eur J Nucl Med Mol Imaging. 2013;40:6Y11. 8. Schmidt H, Brendle C, Schraml C, et al. Correlation of simultaneously acquired diffusion-weighted imaging and 2-deoxy-[18F] fluoro-2-D-glucose positron emission tomography of pulmonary lesions in a dedicated whole-body magnetic resonance/positron emission tomography system. Invest Radiol. 2013; 48:247Y255. 9. Pichler BJ, Judenhofer MS, Catana C, et al. Performance test of an LSO-APD detector in a 7-T MRI scanner for simultaneous PET/MRI. J Nucl Med. 2006; 47:639Y647. 10. Judenhofer MS, Wehrl HF, Newport DF, et al. Simultaneous PET-MRI: a new approach for functional and morphological imaging. Nat Med. 2008;14: 459Y465. 11. Catana C, Wu Y, Judenhofer MS, et al. Simultaneous acquisition of multislice PET and MR images: initial results with a MR-compatible PET scanner. J Nucl Med. 2006;47:1968Y1976. 12. Judenhofer MS, Catana C, Swann BK, et al. PET/MR images acquired with a compact MR-compatible PET detector in a 7-T magnet. Radiology. 2007;244: 807Y814. 13. Schlyer D, Vaska P, Woody C, et al. First images from the BNL simultaneous PET/MRI scanner. J Nucl Med. 2007;48(suppl 2):89P. 14. Schlemmer HP, Pichler BJ, Schmand M, et al. Simultaneous MR/PET imaging of the human brain: feasibility study. Radiology. 2008;248:1028Y1035. 15. Delso G, Furst S, Jakoby B, et al. Performance measurements of the Siemens mMR integrated whole-body PET/MR scanner. J Nucl Med. 2011;52:1914Y1922. 16. Kolb A, Wehrl HF, Hofmann M, et al. Technical performance evaluation of a human brain PET/MRI system. Eur Radiol. 2012;22:1776Y1788. 17. Kolb A, Lorenz E, Judenhofer MS, et al. Evaluation of Geiger-mode APDs for PET block detector designs. Phys Med Biol. 2010;55:1815Y1832. 18. Fedorov A, Beichel R, Kalpathy-Cramer J, et al. 3D slicer as an image computing platform for the quantitative imaging network. Magn Reson Imaging. 2012;30:1323Y1341. 19. Carney JP, Townsend DW, Rappoport V, et al. Method for transforming CT images for attenuation correction in PET/CT imaging. Med Phys. 2006;33:976Y983. 20. National Electrical Manufacturers Association (NEMA). Standards Publication NU 2Y2007: Performance Measurements of Positron Emission Tomographs. Rosslyn, VA: NEMA; 2007. 21. Friedman L, Glover GH. Report on a multicenter fMRI quality assurance protocol. J Magn Reson Imaging. 2006;23:827Y839. 22. Weisskoff RM. Simple measurement of scanner stability for functional NMR imaging of activation in the brain. Magn Reson Med. 1996;36:643Y645. 23. Ramsey NF, Kirkby BS, Van Gelderen P, et al. Functional mapping of human sensorimotor cortex with 3D BOLD fMRI correlates highly with H2(15)O PET rCBF. J Cereb Blood Flow Metab. 1996;16:755Y764. 24. Kinahan PE, Noll DC. A direct comparison between whole-brain PET and BOLD fMRI measurements of single-subject activation response. Neuroimage. 1999;9:430Y438. 25. Joliot M, Papathanassiou D, Mellet E, et al. fMRI and PET of self-paced finger movement: comparison of intersubject stereotaxic averaged data. Neuroimage. 1999;10:430Y447. 26. Devlin JT, Russell RP, Davis MH, et al. Susceptibility-induced loss of signal: comparing PET and fMRI on a semantic task. Neuroimage. 2000;11: 589Y600. 27. Dettmers C, Connelly A, Stephan KM, et al. Quantitative comparison of functional magnetic resonance imaging with positron emission tomography using a force-related paradigm. Neuroimage. 1996;4:201Y209. 28. Rees G, Howseman A, Josephs O, et al. Characterizing the relationship between BOLD contrast and regional cerebral blood flow measurements by varying the stimulus presentation rate. Neuroimage. 1997;6:270Y278. 29. Cherry SR, Woods RP, Hoffman EJ, et al. Improved detection of focal cerebral blood flow changes using three-dimensional positron emission tomography. J Cereb Blood Flow Metab. 1993;13:630Y638. 30. Silbersweig DA, Stern E, Frith CD, et al. Detection of thirty-second cognitive activations in single subjects with positron emission tomography: a new lowdose H2(15)O regional cerebral blood flow three-dimensional imaging technique. J Cereb Blood Flow Metab. 1993;13:617Y629. 31. Jakoby BW, Bercier Y, Conti M, et al. Physical and clinical performance of the mCT time-of-flight PET/CT scanner. Phys Med Biol. 2011;56:2375Y2389. 32. Berker Y, Franke J, Salomon A, et al. MRI-based attenuation correction for hybrid PET/MRI systems: a 4-class tissue segmentation technique using a combined ultrashort-echo-time/Dixon MRI sequence. J Nucl Med. 2012;53:796Y804.

* 2014 Lippincott Williams & Wilkins




33. Hofmann M, Steinke F, Scheel V, et al. MRI-based attenuation correction for PET/MRI: a novel approach combining pattern recognition and atlas registration. J Nucl Med. 2008;49:1875Y1883. 34. Bezrukov I, Mantlik F, Schmidt H, et al. MR-based PET attenuation correction for PET/MR imaging. Sem Nucl Med. 2013;43:45Y59. 35. Catana C, van der Kouwe A, Benner T, et al. Toward implementing an MRIbased PET attenuation-correction method for neurologic studies on the MRPET brain prototype. J Nucl Med. 2010;51:1431Y1438.

* 2014 Lippincott Williams & Wilkins


36. Watson CC, Newport D, Casey ME. A single-scatter simulation technique for scatter correction in 3D PET. In: Grangeat P, Amans J, eds. Fully ThreeDimensional Image Reconstruction in Radiology and Nuclear Medicine. Aix-les-Bains, France: Kluwer Academic Publishers; 1996. 37. Watson CC. New, faster, image-based scatter correction for 3D PET. IEEE Trans Nucl Sci. 2000;47:1587Y1594. 38. Herzog H, Langen K-J, Weirich C, et al. High resolution BrainPET combined with Simultaneous MRI. Nuklearmedizin. 2011;50:74Y82.



381

[Magnetic resonance imaging].

Pelvic magnetic resonance imaging.

Gynecologic magnetic resonance imaging.

Intracranial magnetic resonance imaging.

Shoulder magnetic resonance imaging.

Magnetic resonance imaging.

Acceptance testing of magnetic resonance imaging systems: report of AAPM Nuclear Magnetic Resonance Task Group No. 6.

Magnetic resonance imaging safety in cardiothoracic imaging.

Magnetic resonance imaging in oncology.

Artifacts in magnetic resonance imaging.

Breast magnetic resonance imaging indications.

Magnetic resonance imaging in neurocysticercosis.

Cardiac magnetic resonance imaging: update.

Magnetic resonance coronary artery imaging.

Magnetic resonance imaging of electrolysis.

Functional Magnetic Resonance Imaging Methods.

Magnetic resonance imaging in epilepsy.

Magnetic resonance imaging in hypopituitarism.

Breast magnetic resonance imaging for the interventionalist: magnetic resonance imaging-guided vacuum-assisted breast biopsy.

Cavity- and waveguide-resonators in electron paramagnetic resonance, nuclear magnetic resonance, and magnetic resonance imaging.

Tumor-targeted responsive nanoparticle-based systems for magnetic resonance imaging and therapy.

Magnetic Resonance Imaging-Guided Spine Interventions.

Magnetic resonance imaging in muscular dystrophies.

Magnetic resonance imaging in breast disease.