High resolution BrainPET combined with ...

74

© Schattauer 2011

Original article

High resolution BrainPET combined with simultaneous MRI H. Herzog1; K.-J. Langen1; C. Weirich1; E. Rota Kops1; J. Kaffanke1; L. Tellmann1; J. Scheins1; I. Neuner1,2; G. Stoffels1; K. Fischer1; L. Caldeira3; H. H. Coenen4, N. J. Shah1,5 1Institute

of Neuroscience and Medicine – 4, Forschungszentrum Jülich, Germany; 2Department of Psychiatry, Faculty of Medicine, JARA, RWTH Aachen University, Germany; 3Institute of Biophysics and Biomedical Engineering, Science Faculty of University of Lisbon, Portugal; 4Institute of Neuroscience and Medicine – 5, Forschungszentrum Jülich, Germany; 5Department of Neurology, Faculty of Medicine, JARA, RWTH Aachen University, Germany

Keywords Positron emission tomography, magnetic resonance imaging, PET, MRI, PET/MRI, MRPET, multimodality, hybrid imaging

Summary After the successful clinical introduction of PET/CT, a novel hybrid imaging technology combining PET with the versatile attributes of MRI is emerging. At the Forschungszentrum Jülich, one of four prototypes available worldwide combining a commercial 3T MRI with a newly developed BrainPET insert has been installed, allowing simultaneous data acquisition with PET and MRI. The BrainPET is equipped with LSO crystals of 2.5 mm width and Avalanche photodiodes (APD) as readout electronics. Here we report on some performance characteristics obtained by phantom studies and also on the initial BrainPET studies on various patients as compared with a conventional HR+ PET-only scanner. Material, methods: The radiotracers [18F]-fluoroethyl-tyrosine (FET), [11C]-flumazenil and [18F]-FP-CIT were applied. Results: Comparing

Correspondence to: Prof. Dr.-Ing. Hans Herzog Institute of Neuroscience and Medicine – 4 Forschungszentrum Jülich 52425 Jülich, Germany Tel. +49/(0)24 61/61 59 13; Fax +49/(0)24 61/61 83 10 E-mail: [email protected]

As positron emission tomography (PET) has grown over the past few decades, dedicated brain scanners (15, 16, 21) have been developed and manufactured commercially in parallel with whole-body PET scanners (27, 29, 33). After the success of PET/CT, dedicated brain PET scanners are Nuklearmedizin 2/2011

the PET data obtained with the BrainPET to those of the HR+ scanner demonstrated the high image quality and the superior resolution capability of the BrainPET. Furthermore, it is shown that various MR images of excellent quality could be acquired simultaneously with BrainPET scans without any relevant artefacts. Discussion, conclusion: Initial experiences with the hybrid MRI/BrainPET indicate a promising basis for further developments of this unique technique allowing simultaneous PET imaging combined with both anatomical and functional MRI.

Schlüsselwörter Positronenemissionstomographie, Magnetresonanztomographie, PET, MRT, PET/MRT, MRPET, Multimodalität, Hybridbildgebung

Zusammenfassung Nach der erfolgreichen klinischen Einführung von PET/CT zeichnet sich eine neue bildgebende Hybridtechnologie ab, die PET mit den umfangreichen Möglichkeiten der MRT kom-

Hochauflösende BrainPET kombiniert mit simultaner MRT Nuklearmedizin 2011; 50: 74–82 doi:10.3413/Nukmed-0347-10-09 received: September 1, 2010 accepted in revised form: January 5, 2011 prepublished online: February 2, 2011

biniert. Im Forschungszentrum Jülich wurde einer von vier weltweit verfügbaren Prototypen aufgestellt, die ein kommerzielles 3T MRT mit einem neu entwickelten BrainPET-Einsatz vereinen, so dass simultane Messungen von PET und MRT möglich sind. Der BrainPET ist mit 2,5 mm breiten LSO-Kristallen ausgestattet, die von Avalanche Photodioden (APD) ausgelesen werden. In dieser Arbeit beschreiben wir erstens einige mithilfe von Phantomuntersuchungen gemessene Leistungsparameter und zweitens erste BrainPET-Aufnahmen bei verschiedenen Patienten in Vergleich zu Messungen mit einem konventionellen HR+ PET. Material, Methoden: Als Radiotracer wurden [18F]-Fluorethyl-Tyrosin (FET), [11C]-Flumazenil und [18F]-FP-CIT eingesetzt. Ergebnisse: Der Vergleich der mit dem BrainPET gewonnenen Daten mit denen des HR+ PET macht die sehr gute Bildqualität und die verbesserte Auflösung des BrainPET deutlich. Weiterhin wird gezeigt, dass unterschiedliche MR-Sequenzen simultan mit den BrainPETAufnahmen durchgeführt werden können, wobei eine ausgezeichnete Qualität ohne relevante Beeinträchtigungen erzielt wird. Diskussion, Schlussfolgerungen: Die anfänglichen Erfahrungen mit dem hybriden MRT/ BrainPET sind eine viel versprechende Basis für Weiterentwicklungen dieser einzigartigen Technik, die funktionelle PET-Bildgebung mit der anatomischen und funktionellen MRT kombiniert.

no longer commercially available, although a prototype of a high-resolution research tomograph (HRRT) has demonstrated its utility in neurological PET imaging.

limited to 4–5 mm (3, 34, 37), the HRRT achieves a resolution of about 2.5 mm which considerably reduces partial volume effects in brain imaging (7, 40).

While the intrinsic resolution of PET images from current human PET/CT scanners is

Following on from the success of PET/CT (2, 36) and more recently SPECT/CT (5,

H. Herzog et al.: HR BrainPET with MRI

10) by providing corresponding functional and structural information in one examination, a new hybrid imaging technology combining PET and magnetic resonance imaging (MRI) for human studies is under development (9, 13, 31, 32). As a first step towards this new technology, some prototypes of hybrid PET/MRI scanners for brain studies were installed by Siemens Medical Solutions Inc. at four PET centres in Germany and the United States, that is, in ● Tübingen, ● Boston, ● Jülich and ● Atlanta. Using new solid-state based magneto-insensitive detector electronics, the prototype combines a dedicated BrainPET with a standard 3T MR scanner for the simultaneous imaging of the human brain (31). With a crystal width of 2.5 mm, the BrainPET is expected to offer high-resolution PET imaging. The combination of the BrainPET with MRI opens new horizons for bimodal imaging where MRI is not restricted to structural imaging but includes the full spectrum of functional MRI techniques. This paper describes the technology and some performance parameters of the 3TMR-BrainPET installed at the Forschungszentrum Jülich in autumn 2008. A number of simultaneous PET/MRI studies in patients with various brain pathologies are illustrated, demonstrating the imaging potential and the feasibility of the new hybrid imaging technology.

with a horizontal radial offset of 0, 25 mm, 50 mm, 75 mm and 100 mm. Each measurement took 300 s. Transversal images of the line source were reconstructed within the central plane (z = 0 mm) and two planes located at z = ± 40 mm. For the axial resolution, a [22Na] point source of 300 kBq was placed in air within the central plane and 40 planes separated by 1 mm each at radii of 0 and 50 mm. Each measurement took 300 s. Using the STIR software library (35) images with 256 × 256 pixels each sized 1.25 × 1.25 mm2 were reconstructed by filtered back projection with a ramp filter of 0.5 cycles/pixel. Resolution is expressed as the full width at half maximum (FWHM) of the Gaussian curves fitting the profiles through the line and point source, respectively. The sensitivity was determined as the point source sensitivity. For this purpose a point source of 7.6 MBq [18F]-solution was positioned in the centre of the BrainPET and scanned for 60 seconds. The total number of measured prompt coincidences Ctot was obtained by single slice rebinning of all 1399 straight and oblique sinograms. The sensitivity was calculated as the ratio of decay-corrected Ctot and the activity in the point source. An equivalent test was recorded on a Siemens ECAT Exact HR+ PET scanner (4) in our laboratory. Besides the patient measurements described below, a realistic brain phantom for a phantom-based comparison of the image quality of BrainPET and HR+ PET scanner was employed. The brain phantom was

constructed by Iida and co-workers (18) from a photo-curable polymer with density of 1.07 g/ml by using a laser-modelling technique. It simulates detailed grey matter and the skull of a young healthy volunteer. It was filled with 40 MBq [18F]-solution and measured in the BrainPET for 10 min and directly afterwards in the HR+ PET for another 10 min. Corrections and reconstructions were performed as described below for patient studies. In the case of the BrainPET, attenuation correction was based on a transmission measurement of the phantom recorded in the HR+ PET scanner.

Patients and radiopharmaceuticals To date, most PET/MRI measurements have been performed in brain tumour patients referred to our PET facility for a diagnostic scan with [18F]-fluoro-ethyl-tyrosine (FET) (20, 26, 39). The diagnostic scan was carried out on an HR+ PET scanner. For this purpose, around 200 MBq FET were injected at the start of the PET acquisition. [11C]-flumazenil was applied in a patient in order to localize a seizure focus (25). One of the studies reported here used [18F]-FP-CIT for a differential diagnosis of Parkinson’s disease (19). Further information on the comparative measurements with the BrainPET and the HR+ are detailed below together with the reports of the individual cases. Since the 3TMRBrainPET is not a commercial product but

Material, methods Performance measurements of the BrainPET Since not all performance parameters according to the NEMA protocols are of interest when comparing brain studies, this report focuses on the image resolution and the sensitivity of the imaging device. For measurements of the transversal resolution, a line source filled with 11 MBq [18F]-solution and an inner diameter of 0.5 mm was positioned along the z-axis in air © Schattauer 2011

Fig. 1 The 3TMR-BrainPET consisting of a MAGNETOM Trio MRI and the BrainPET insert shown also separately with removed inner cover so that the cassettes become visible.

Nuklearmedizin 2/2011

75

76


a prototype, it can only be used in human studies under the conditions of a clinical study according to §20 of the German Medizinproduktegesetz (German Act on Medical Devices). The studies were approved by the local ethical committee and written, prior informed consent was obtained from each patient measured in the 3TMR-BrainPET.

3TMR-BrainPET The two components of the 3TMR-BrainPET are a Siemens 3T MRI scanner MAGNETOM Trio and the newly constructed BrainPET as an insert in the bore of the magnet (씰Fig. 1). Without the BrainPET insert, the MAGNETOM Trio is CE-certified and can be operated as a standard scanner. When the BrainPET is inserted into the MRI scanner, the standard patient bed has to be replaced by a bed of a fixed height and two accompanying head coils. In this mode, the body coil commonly dedicated for transmitting the high frequency pulses is switched off. Thus, the outer birdcage head coil can be used for both transmitting radiofrequency pulses and receiving the signals, whereas the inner coil is an 8-channel coil for receive only. These coils fit into the BrainPET (inner diameter of 36 cm). They are optimised for PET with respect to minimal attenuation of radiation. Since the 3TMR-BrainPET with its slightly modified MR components and the newly constructed BrainPET is not a

commercial product and not CE-certified, it can only be operated in human studies under the conditions of a clinical study as mentioned above. The BrainPET is a compact cylinder with a length of 72 cm and an outer diameter of 60 cm, fitting snugly into the bore of the magnet. It consists of 32 detector cassettes with a 6 mm gap between the cassettes. Each cassette contains 6 detector modules covering an axial field-of-view (FOV) of 19.2 cm. The diameter of the FOV is 31.4 cm. The detector modules are separated by gaps of 2.5 mm. The front end of the detector module is a 12 × 12 matrix of 2.5 × 2.5 × 20 mm3 individual LSO crystals coupled via a light guide to a 3 by 3 array of Hamamatsu S8664–55 APDs, each with an area of 5 × 5 mm2. In addition, each detector module contains a high voltage board supplying 500 V to the APDs, a board with a custom 10-channel charge-sensitive preamplifier ASIC, a mapping and pulse-shaping board and ASIC output driver board. To avoid interference with the MR radiofrequency field, each single cassette is copper shielded. There are 32, 10 m long cables connecting the cassettes to the filter plate, and thereafter to a modified version of the QuickSilver data acquisition electronics (24) developed by Siemens for small animal PET. To reduce radiation from outside the FOV, a ring of polymer containing lead is placed around the patient at the front edge of the BrainPET. The coincidence window of the BrainPET is 12 ns and the energy window ranges from 420 to 580

Fig. 2 Plots of tangential (■), radial (◆) and axial (Δ) image resolution expressed as full width at half maximum (FWHM) measured with line and point sources in air at different radial offsets.


keV. Since the APDs are very temperature dependent, the temperature of the BrainPET is stabilized with cooled air.

Data acquisition The data acquisition systems of the two parts of the 3TMR-BrainPET are completely independent of each other. MR sequences are defined, initiated and controlled with a standard SYNGO console. This console also serves to display the reconstructed MR images. The PET acquisition is defined and activated by dedicated commands within the command window of a PC. All PET data reported here were recorded in list mode. Each entry of the listmode data consists of 48 bits and represents a prompt coincidence, a delayed coincidence, a time mark (at intervals of 200 μs), a block counter value or a gating signal. The listmode stream can be divided into predefined time frames of appropriate duration from several minutes down to a few seconds as demonstrated below (씰Fig. 8).

Processing and reconstruction of PET data Prior to reconstruction and the preceding correction procedures, the listmode data are rebinned and histogrammed into a prompt sinogram and a delayed crystal map (DCMAP) for each time frame. The DCMAP is derived from the delayed coincidences measured by the delayed window technique (14) and contains the number of counts for each crystal. A “smooth” random sinogram is calculated from the DCMAP by the variance reduction approach (6) in order to reduce the statistical noise of the originally measured delayed coincidences. There are 1399 sinograms of prompts and “smooth” randoms subdivided into one transaxial and 14 oblique segments. The span is 9 and the maximum ring difference is 67. The transaxial segment consists of 153 sinograms corresponding to 153 reconstructed image planes. Each sinogram has 256 bins and 192 projections. Finally, the measured data are scaled to the frame length to obtain counts per seconds. © Schattauer 2011


Image reconstruction is performed using the Siemens OSEM3D software, which implements the ordered subset – expectation maximisation (OS-EM) algorithm (17). Input data are the prompts and the random sinograms as well as the normalisation data for crystal efficiencies, the attenuation map and the scatter sinogram. The scatter correction is calculated using a single scatter simulation (SSS) (38). Since the 3TMR-BrainPET does not offer a facility for measured attenuation correction, a method was developed by our group which allows the derivation of attenuation correction factors using the individual MR image of the subject examined (30). This method is based on a template of an MPRAGE image averaged from 8 scans of different subjects, which were spatially normalized to the head shape of one subject and a corresponding template of an attenuation map obtained from transmission scans of these subjects recorded with the HR+. Using SPM (1, 8) the deformation matrix between the MR template and the individual MPRAGE image is determined and then applied to the HR+-based attenuation template. The resulting individualized attenuation map is completed by an attenuation map of the head coil obtained from a transmission measurement in the HR+ scanner. Currently, four reconstruction schemes are available. The unweighted OSEM which reconstructs from true sinograms with all corrections performed prior to the reconstruction. However, this approach disregards the assumption that the data obey Poisson statistics. Therefore, two versions of weighted OSEM are available where corrections for attenuation (AW-OSEM) and additionally normalisation (NAW-OSEM) are moved into the reconstruction algorithm (22). Finally, a scheme for fully 3D ordinary Poisson reconstruction can be selected, where all corrections, including that for scatter are performed within the reconstruction process (OP-OSEM). This preserves the Poisson statistics of the measurement and is the standard image reconstruction used in our BrainPET. The OP-OSEM reconstruction with 16 subsets and 6 iterations used in the present work requires about 3 min per frame. The reconstructed image has 256 × 256 × 153 isotropic voxels of (1.25 mm)3. The resulting voxel intensities have to be multi© Schattauer 2011

Fig. 3 Transaxial, coronal and sagittal images (from left to right) of the Iida brain phantom filled with [18F]-solution and measured in the HR+ PET (top row) and the BrainPET (middle row). The bottom row displays CT images of the phantom with clear delineation of the “bone” (white) and the “grey matter” (dark grey).

Fig. 4 Transaxial (top) and sagittal (bottom) BrainPET images of a tumour patient. The left column shows images without attenuation correction, whereas the middle column presents the images with attenuation correction and the right column those after post-filtering with a 3 mm 3D Gaussian kernel.

plied with a calibration factor to achieve activity concentrations (e. g. in kBq/cc) instead of counts. Furthermore, since the reconstructed images are affected by noise they are filtered slightly with a 3D Gaussian kernel (filter width 3 mm). With regard to the scanner resolution of about 3 mm, this kernel can be regarded as a matched filter.

Measurements with the HR+ As mentioned, the BrainPET measurements reported here were performed as a

comparison with diagnostic measurements carried out with HR+ PET-only scanner, the performance of which has been described by Brix and coworkers (4). The major differences relevant here in comparison to the BrainPET are the shorter axial FOV of 15.5 cm and the reconstructed image resolution of about 6 mm. The HR+ data were recorded with interplane septa retracted in 3D mode. After FORE-rebinning, images were iteratively reconstructed with OSEM2D (16 subsets, 6 iterations). Random coincidences were determined by the delayed window technique and directly Nuklearmedizin 2/2011

77

78


ing the two PET images with the CT images: in contrast to the HR+, the grey matter compartment visualized in the BrainPET images is an excellent agreement with the grey matter compartment shown in the CT image.

Patient studies

Fig. 5 Transaxial, coronal and sagittal images (from left to right) of a patient with a glioblastoma. The images shown in the top row were recorded with the HR+ PET and summed from 20 to 40 min after injection (p. i.) of about 200 MBq [18F]-fluoro-ethyl-tyrosine (FET). The images shown in the middle row were recorded with the BrainPET from 70 to 100 min p.i.. The bottom row displays the MP-RAGE images acquired for 10 min during the BrainPET measurement superimposed with the BrainPET images.

Fig. 6 MR sequences during the FET-BrainPET of the patient shown in Figure 5 (from left to right): T1-weighted MP-RAGE without and with contrast medium and T2-weighted TIRM.

subtracted from the prompt coincidences. For measured attenuation correction a transmission scan of 10 min was acquired. The reconstruction yielded 63 images per frame with 128 × 128 voxels of 2 × 2 × 2.45 mm3.

Results Phantom measurements The results of the line source measurements are shown in 씰Figure 2. In the centre of the FOV, the resolution is 3 mm in Nuklearmedizin 2/2011

the tangential, radial and axial directions. At radial offsets of up to 75 mm the tangential resolution stays at about 3 mm, whereas the radial resolution decreases to 4.5 mm. The axial resolution decreases to 3.5 mm at a radial offset of 50 mm. The point source sensitivity of the BrainPET is approximately 6%, whereas it is 4.3% measured with the HR+. 씰Figure 3 compares images of the Iida brain phantom measured in the BrainPET and HR+ PET. The details of the grey matter compartment filled with [18F]-solution are much better delineated by the BrainPET. This can also be seen by compar-

When reconstructing BrainPET images, the attenuation correction is a decisive step in the processing chain. Without attenuation correction one observes not only the well-known underestimation of tracer uptake in the central brain, but a disturbing pattern of vertical artefacts especially in coronal and sagittal views (씰Fig. 4). These are partly related to the gaps between the detector cassettes and partly to neglected attenuation caused by the head coils. The artefacts disappear in most cases after attenuation correction. In some cases they are still present although faint – a fact which needs further investigation. After reconstruction the attenuation and scatter corrected images are relatively noisy owing to the low count rate of the tiny crystals. Image quality can be improved effectively by filtering the data with a 3D Gaussian kernel of 3 mm. The first comparative studies in brain tumour patients aimed to examine the image quality of the BrainPET and the feasibility of simultaneous MR imaging. After injection of 200 MBq FET, a patient with a glioblastoma was scanned for 50 min in the HR+ scanner. Quickly thereafter, the patient underwent an additional scan for 30 min in the BrainPET. 씰Figure 5 presents the images of both PET scanners. When comparing the FET distribution between the two PET scans one must take into account that FET uptake is not stable and may change differently in the normal cortex and in the tumour over time. The tumour to brain ratio (TBR) obtained in the HR+ image (20 to 40 min p. i.) was 2.25 while the BrainPET scan yielded a TBR of 1.77 using the same ROIs. The lower TBR may be explained by the decreasing FET-uptake in the tumour while the radioactivity in the normal cortex remained constant. Different MR sequences were used during the BrainPET acquisition, namely a © Schattauer 2011


T1-weighted MP-RAGE sequence both with and without gadolinium as the contrast medium and a T2-weighted TIRM sequence (씰Fig. 6). The MR images have an excellent quality without artefacts but a slight inhomogeneity along the axial direction can be observed in the coronal and sagittal MPRAGE image. This is caused by the fact that the transmit body coil is switched off after insertion of the BrainPET and the outer head coil is used as a transmitter resulting in a decreased homogeneity of the transmit in comparison to that obtainable from the body coil. After the successful realization of simultaneous PET and anatomical MRI, the acquisition protocol was expanded by adding functional MRI (fMRI) to measure brain activation. A patient with an astrocytoma, WHO grade II-III, was first measured for 50 min in the HR+ scanner after injection of FET as described above. Directly thereafter, a list mode recording of 35 min was started with the BrainPET and five different MR sequences including EPI were acquired simultaneous to the PET measurement. During the EPI sequences of 12 min duration, the patient performed a motor task of right-handed, left-handed and twohanded finger-tapping (4 min each). Within the 4 min period, finger-tapping of 30 sec alternated with pauses of 30 seconds. 씰Figure 7 compares sagittal HR+ images summed from 20 to 40 min p. i. with the BrainPET images summed over 30 min. Furthermore, the different activations located at the motor cortex are displayed as projections on the anatomical MP-RAGE images. Finally, the chronological order of measurements in the HR+ scanner and the BrainPET was changed in some patients. The patients were first studied in the BrainPET after injection of FET (50 min listmode study) and a dynamic image sequence was reconstructed. After a short interval, an HR+ measurement of 30 min plus a transmission scan of 10 min was added. 씰Figure 8a displays the time-activity curves of the FET kinetics in normal cortex, tumour and carotid artery generated with the data from the BrainPET insert in a patient with a astrocytoma, WHO grade II. The frame length during the blood peak was 5 s. 씰Figure 8b shows transaxial © Schattauer 2011

R

Fig. 7 Astrocytoma WHO grade II-III: a motor task was performed and cerebral activation measured by fMRI for 12 min during the BrainPET acquisition. Top row (from left to right): Transaxial and sagittal BrainPET image (70 to 105 min p. i.) and sagittal HR+ image (20 to 40 min p. i.). Bottom row (from left to right): Transaxial MP-RAGE images with superimposed centres of neuroactivation at the motor cortex during left-handed, right-handed and bi-manual finger-tapping.

images across the carotid artery and the tumour at different time points. Another study investigated the performance of the hybrid system by imaging the density of central benzodiazepine receptors in a patient with partial seizures. The patient was first scanned in the HR+ tomograph from 10 to 25 min after injection of 512 MBq [11C]-flumazenil (25) and thereafter in the 3TMR-BrainPET for another 20 min. 씰Figure 9 presents transversal and sagittal images recorded in the two tomographs together with the MP-RAGE images obtained during the BrainPET measurement. The sagittal BrainPET image in particular documents the superior resolution of the new BrainPET. While the cortical gyri are blurred in HR+ PET images the BrainPET provides a considerably improved delineation of single gyri. The right temporal lobe showed a decreased flumazenil binding indicating the focus responsible for the seizures. Further, a study with the dopamine transporter ligand [18F]-FP-CIT was performed in a patient with a preliminary diagnosis of Parkinson’s disease (19). 60 min after injection of 200 MBq [18F]-FP-CIT the diagnostic scan with the HR+ PET was started for 40 min. In this case a transmission scan of 10 min was recorded after the injection of the tracer at the end of the emission scan. Thereafter, with an interval of 23 min the patient volunteered for an ad-

ditional examination in the BrainPET, where a listmode study was recorded for 30 min together with an MP-RAGE sequence. The PET images (씰Fig. 10) revealed a heavily reduced bilateral uptake of [18F]-FP-CIT in the putamen confirming the diagnosis of Parkinson’s disease. Once again, this example underlines the better image resolution of the BrainPET. The ratio between the maximum uptake in the caudate nucleus relative to the whole brain activity was 7.9 for the HR+ scan and 10.2 for the BrainPET scan. The higher ratio for the BrainPET may be related partly to the decreased partial volume effect and partly to different rates of tracer washout in caudate nucleus and brain. The comparison of BrainPET and MRI demonstrates the decreased transmitter receptor metabolism in the putamen without visible structural changes.

Discussion The simultaneous PET/MRI measurements described here demonstrate the suitability of the present prototype of the 3TMR-BrainPET for high-resolution PET imaging of the human brain. Compared to the first feasibility report (31), considerable progress in image quality of the BrainPET has been achieved. The first images obtained with our BrainPET showed signifiNuklearmedizin 2/2011

79

80


a)

b)

Fig. 8 Time-activity curves (a) of the FET kinetics in normal cortex, tumour and carotid artery of a brain tumour patient measured in the BrainPET for 50 min after injection of FET. b) Image (left) at the level of the carotid arteries acquired from 15 to 75 seconds p. i. and two images at the level of the tumour recorded from 10 to 20 min p. i. (middle) and 40 to 50 min p. i. (right) indicating the slowly increasing FET uptake in the astrocytoma II.

cant vertical artefacts (씰Fig. 4); these have been eliminated almost completely by the continuous progress of the correction procedures. This progress is related to ● an improved accuracy of the attenuation map of the head coils, ● a regular update of the normalisation procedure, and ● the application of 3D scatter correction. The most prominent feature of the BrainPET compared to the HR+ PET is the improved spatial resolution which is primarily based on the smaller crystal size of the BrainPET. To a minor extent the different reconstruction programs, i. e. OSEM3D for the BrainPET and OSEM2D after FORE-rebinning may contribute to the better resolution of the BrainPET. The Nuklearmedizin 2/2011

qualitative impression of the high resolution BrainPET images is confirmed by the results of the line and point source measurements. At the centre of the FOV the image resolution expressed as FWHM is 3 mm in all directions. The tangential resolution is about 3 mm for radial offsets of up to 75 mm, whereas the radial FWHM increases to 4.5 mm due to the fact that the BrainPET component has no depth-of-interaction capability. The axial resolution has a FWHM of 3.5 mm at a radius of 5 cm. These values for tangential and axial resolution are slightly inferior to those of the dedicated HRRT brain tomograph (7) which has a central resolution of 2.6 to 3 mm. This can be explained by the HRRT’s smaller crystal width of 2.1 mm compared to 2.5 mm for the BrainPET. An important

difference between the PET systems is the decreasing radial resolution with increasing radial offset. In contrast to the BrainPET, the HRRT has double-layer crystals so that the depth-of-interaction effect is considerably reduced. Considering the imaging characteristics of the BrainPET it must be noted that the resolution at the cortical surface of the human brain is not isotropic. Nevertheless, the BrainPET delivers excellent images of the cortex as documented by 씰Figures 3 and 9. 씰Figure 9 presents an excellent agreement of the cortical structures visualized by the uptake of [11C]-flumazenil in BrainPET images with the grey matter displayed by the MP-RAGE images. This agreement may open the possibility to perform an MRbased partial volume correction of [11C]-flumazenil PET images. The model applied for such a correction requires, however, an accurate knowledge of how the intensity changes in the MR-data are related to the cortical [11C]-flumazenil uptake. The primary images resulting from OSEM3D reconstruction appeared quite noisy. Therefore, they were filtered with a 3D Gaussian kernel with 3 mm filter width. However, the noise reduction is related to a loss of resolution. The iterative reconstruction methods used here do not apply an advanced modelling system. If this type of modelling is applied, high resolution images with little noise are expected which may need no postfiltering. An examination of this issue is, however, beyond the scope of the present paper. Concerning the comparisons of the PET scans obtained with the BrainPET to those with the HR+ PET one has to consider the radiotracer decay and differing kinetic changes in different tissues between the two scans. Furthermore, the BrainPET’s (point source) sensitivity with a value of 6% is better than that of the HR+ with 4.3%. This difference compensates for the disadvantage of the small crystals of the BrainPET to some extent. The experiment with the Iida brain phantom allows a direct comparison between the image resolution of the BrainPET and the HR+, which is an older, obsolete tomograph with an image resolution similar to the PET component of modern PET/CT machines. Here, differ© Schattauer 2011


ing kinetic changes, as found in human data, play no role. The statistics recorded are very similar with an interval of just 20 min between the scans. Thus, one can conclude that the primary reason for the better image quality obtained with the BrainPET compared to the HR+ PET is the better resolution. When analysing tracer kinetic data, the high resolution of the BrainPET, an optimal 3 mm, offers the possibility to obtain an image derived-input function (IDIF) from the arterial blood in the carotid artery as has been suggested and examined for the HRRT (23). For this application the decrease of radial resolution does not play a role, as the carotid artery is located quite close to the z-axis. For the measurement of the IDIF, head motion can pose a problem but simultaneous MRI may deliver information on head motion which can then be used to correct the PET data, either the line-of-response data or at the image level. Although the results of 3TMR-BrainPET are encouraging, further improvements of this new hybrid imaging device are needed to avoid the minor image artefacts described. Furthermore, the present status with separate platforms for PET and MRI and a command line driven PET processing delivering images in raw format does not allow for the efficient workflow necessary for both research and clinical studies with higher patient throughput. For brain imaging the necessity of hybrid PET/MRI remains a matter of discussion (11, 28). With respect to anatomical co-registration, the rigid structures of the head allow a very reliable overlay of PET and MR images using dedicated software; the data of both modalities may be acquired sequentially. The development of PET/MRI, however, aims not only at perfect anatomical co-registration and time saving in the clinical imaging process: PET/MRI allows the simultaneous acquisition of various molecular and functional parameters that are variable over time and cannot be measured sequentially (12).

During a dynamic PET of, e. g., 60 min to examine brain metabolism or neuroreceptor occupancy, versatile MRI with different sequences may be performed. In this way, © Schattauer 2011

Fig. 9 Patient with temporal lobe epilepsy investigated with the benzodiazepine receptor ligand [11C]-flumazenil (from left to right): transaxial (upper row) and sagittal (lower row) HR+ images (10 to 25 min p. i.), BrainPET images (30 to 50 min p. i.), BrainPET images fused with the simultaneous MP-RAGE images, simultaneous MP-RAGE images. The arrow points at an area of low uptake of [11C]-flumazenil at the right frontal temporal lobe what may indicate the epileptic focus.

Fig. 10 Patient with Parkinson’s disease investigated with [18F]-FPCIT for dopamine transporter function top row: transaxial, coronal and sagittal images (from left to right) recorded with the BrainPET from 120 to 150 min p. i. middle row: MPRAGE images acquired for 10 min during the BrainPET measurement bottom row: HR+ PET recorded from 60 to 100min p.i..

various studies including functional MRI (fMRI) altered by mental challenge, MRbased perfusion studies with arterial spin labelling (ASL) or perfusion weighted imaging (PWI) may be recorded simultaneously along with PET. Another example could be the observation of tumour physiology (perfusion, cell density, macromolecules) during therapy with positron emitting radionuclides.

Conclusion This report demonstrates the feasibility of simultaneous high resolution PET/MRI imaging of the brain using the newly developed 3TMR-BrainPET. The possibility to combine PET with a large variety of simultaneously acquired MRI methods opens new horizons for bimodal multi-parametric imaging of the human brain. Nuklearmedizin 2/2011

81

82


Acknowledgments We appreciate the continuous and valuable support and advice of L. Byars, C. Michel and M. Schmand (Siemens Medical Solutions, Knoxville). We thank K. Frey, S. Schaden and E. Theelen for excellent assistance with the PET and PET/MRI studies. We thank our colleagues at the Institute of Neurosciene and Medicine – 5 (Radiochemistry) for reliably supplying us with all the necessary radiotracers. Conflict of interest The authors declare, that there is no conflict of interest.

References 1. Ashburner J, Friston KJ. Nonlinear spatial normalization using basic functions. Human Brain Mapping 1999; 7: 253–266. 2. Beyer T, Townsend DW, Brun T et al. A combined PET/CT scanner for clinical oncology. J Nucl Med 2000; 41: 1369–1379. 3. Bolard G, Prior JO, Modolo L et al. Performance comparison of two commercial BGO-based PET/ CT scanners using NEMA NU 2–2001. Med Phys 2007; 34: 2708–2717. 4. Brix G, Zaers J, Adam LE et al. Performance evaluation of a whole-body PET scanner using the NEMA protocol. J Nucl Med 1997; 38: 1614–1623. 5. Buck AK, Nekolla S, Ziegler S et al. SPECT/CT. J Nucl Med 2008; 49: 1305–1319. 6. Byars LG, Sibomana M, Burbar Z et al. Variance reduction on randoms from delayed coincidence histograms for the HRRT. IEEE Nucl Sci Symp Conf Record 2005; 2622–2626. 7. De Jong HWAM, van Velden FHP, Kloet RW et al. Performance evaluation of the ECAT HRRT: an LSO-LYSO double layer high resolution, high sensitivity scanner. Phys Med Biol 2007; 52: 1505–1526. 8. Friston KJ, Frith CD, Liddle PF, Frackowiak RS. Comparing functional (PET) images: the assessment of significant change. J Cereb Blood Flow Metab 1991; 11: 690–699. 9. Grazioso R, Zhang N, Corbeil J et al. APD-based PET detector for simultaneous PET/MR imaging. Nucl Instr Methods Phys Res A 2006; 569: 301–305.


10. Hasegawa BH, Gingold EL, Reilly SM et al. Description of a simultaneous emission-transmission CT system. Proc SPIE 1990; 1231: 50–60. 11. Heiss WD. The potential of PET/MR for brain imaging. Eur J Nucl Med Mol Imaging 2009; 36 (Suppl S1): S105–S112. 12. Herzog H, Pietrzyk U, Shah NJ, Ziemons K. The Current State, challenges and perspectives of MRPET. NeuroImage 2010; 49: 2072–2082. 13. Herzog H, Tellmann L, Marx B et al. First performance tests of the 3TMR-BrainPET. J Nucl Med 2009; 50 (Suppl. 2): 310P. 14. Hoffman EJ, Huang SC, Phelps ME, Kuhl DE. Quantitation in positron emission computed tomography: 4. Effect of accidental coincidences. J Comput Assist Tomogr 1981; 5: 391–400. 15. Hoffman EJ, Phelps ME, Huang SC. Performance evaluation of a positron tomograph designed for brain imaging. J Nucl Med 1983; 24: 245–257. 16. Holte S, Eriksson L, Dahlbom M. A preliminary evaluation of the Scanditronix PC2048–15B brain scanner. Eur J Nucl Med 1989; 15: 719–721. 17. Hudson HM, Larkin RS. Accelerated image reconstruction using ordered subsets of projection data. IEEE Trans Med Imag 1994; 13: 601–609. 18. Iida H, Zeniya T, Imabayashi E et al. Three-dimensional realistic brain phantoms containing detailed grey matter and bone structures for nuclear medicine imaging. J Nucl Med 2009; 50 (Suppl 2): 123P. 19. Kazumata K, Dhawan V, Chaly T et al. Dopamine transporter imaging with fluorine-18-FPCIT and PET. J Nucl Med 1998; 39: 1521–1530. 20. Langen KJ, Hamacher K, Weckesser M. O-(2-[18F]fluoroethyl)-L-tyrosine: uptake mechanisms and clinical applications. Nucl Med Biol 2006; 33: 287–294. 21. Mazoyer B, Trebossen R, Deutch R et al. Physical characteristics of the ECAT 953B/31: a new high resolution brain positron tomograph. IEEE Trans Med Imaging 1991; 10: 499–504. 22. Michel C, Liu X, Sanabria S et al. Weighted schemes applied to 3D-OSEM reconstruction in PET. IEEE Nucl Sci Symp Conf Record 1999; 3: 1152–1157. 23. Mourik JE, van Velden FH, Lubberink M et al. Image derived input functions for dynamic High Resolution Research Tomograph PET brain studies. Neuroimage 2008; 43: 676–686. 24. Newport DF, Siegel SB, Swann BK et al. QuickSilver: A flexible, extensible, and high-speed architecture for multi-modality. IEEE Nucl Sci Symp Conf Record 2006; 2333–2334. 25. Niimura K, Muzik O, Chugani DC et al. [11C]flumazenil PET: activity images versus parametric images for the detection of neocortical epileptic foci. J Nucl Med 1999; 40: 1985–1991.

26. Pauleit D, Floeth F, Hamacher K et al. O-(2-[18F]fluoroethyl)-L-tyrosine PET combined with magnetic resonance imaging improves the diagnostic assessment of cerebral gliomas. Brain 2005; 128: 678–687. 27. Phelps ME, Hoffman EJ, Huang SC, Kuhl DE. ECAT: A new computerized tomographic imaging system for positron-emitting radiopharmaceuticals. J Nucl Med 1978; 19: 635–647. 28. Pichler B, Kolb A, Nagele T, Schlemmer HP. PET/ MRI: paving the way for the next generation of clinical multimodality imaging applications. J Nucl Med 2010; 51: 333–336. 29. Rota Kops E, Herzog H, Schmid A et al. Performance characteristics of an eight-ring whole body PET scanner. J Comput Assist Tomogr 1990; 14: 437–445. 30. Rota Kops E, Herzog H. Template based attenuation correction for PET in MR-PET scanners. IEEE NSSMIC Record 2008; 3786–3789. 31. Schlemmer HP, Pichler B, Schmand M et al. Simultaneous MR/PET imaging of the human brain: feasibility study. Radiology 2008; 248: 1028–1035. 32. Schmand M, Burbar Z, Corbeil J et al. BrainPET: First human tomograph for simultaneous (functional) PET and MR imaging. J Nucl Med 2007; 48: 310P. 33. Spinks TJ, Jones T, Gilardi MC, Heather JD. Physical performance of the latest generation of commercial positron scanner. IEEE Trans Nucl Sci 1988; 35: 721–725. 34. Surti S, Kuhn A, Werner ME et al. Performance of Philips Gemini TF PET/CT scanner with special consideration for its time-of-flight imaging capabilities. J Nucl Med 2007; 48: 471–480. 35. Thielemans K, Mustafovic S, Tsoumpas C. STIR: Software for tomographic image reconstruction release 2. IEEE NSS-MIC record 2006; 2174–2176. 36. Townsend DW, Beyer T, Blodgett TM. PET/CT scanners: a hardware approach to image fusion. Semin Nucl Med 2003; 33: 193–204. 37. Trébossen R, Comtat C, Brulon V et al. Comparison of two commercial whole body PET systems based on LSO and BGO crystals respectively for brain imaging. Med Phys 2009; 36: 1399–1409. 38. Watson CC. New, faster, image-based scatter correction for 3D PET. IEEE Trans Nucl Sci 2000; 47: 1587–1594. 39. Wester HJ, Herz M, Weber W et al. Synthesis and radiopharmacology of O-(2-[18F]fluoroethyl)L-tyrosine for tumor imaging. J Nucl Med 1999; 40: 205–212. 40. Wienhard K, Schmand M, Casey ME et al. The ECAT HRRT: performance and first clinical application of the new high resolution research tomography. IEEE Trans Nucl Sci 2002; 49: 104–110.

© Schattauer 2011