Towards integration of PET / MR hybrid imaging into radiation therapy treatment planning

PURPOSE
Multimodality imaging has become an important adjunct of state-of-the-art radiation therapy (RT) treatment planning. Recently, simultaneous PET/MR hybrid imaging has become clinically available and may also contribute to target volume delineation and biological individualization in RT planning. For integration of PET/MR hybrid imaging into RT treatment planning, compatible dedicated RT devices are required for accurate patient positioning. In this study, prototype RT positioning devices intended for PET/MR hybrid imaging are introduced and tested toward PET/MR compatibility and image quality.


METHODS
A prototype flat RT table overlay and two radiofrequency (RF) coil holders that each fix one flexible body matrix RF coil for RT head/neck imaging have been evaluated within this study. MR image quality with the RT head setup was compared to the actual PET/MR setup with a dedicated head RF coil. PET photon attenuation and CT-based attenuation correction (AC) of the hardware components has been quantitatively evaluated by phantom scans. Clinical application of the new RT setup in PET/MR imaging was evaluated in anin vivo study.


RESULTS
The RT table overlay and RF coil holders are fully PET/MR compatible. MR phantom and volunteer imaging with the RT head setup revealed high image quality, comparable to images acquired with the dedicated PET/MR head RF coil, albeit with 25% reduced SNR. Repositioning accuracy of the RF coil holders was below 1 mm. PET photon attenuation of the RT table overlay was calculated to be 3.8% and 13.8% for the RF coil holders. With CT-based AC of the devices, the underestimation error was reduced to 0.6% and 0.8%, respectively. Comparable results were found within the patient study.


CONCLUSIONS
The newly designed RT devices for hybrid PET/MR imaging are PET and MR compatible. The mechanically rigid design and the reproducible positioning allow for straightforward CT-based AC. The systematic evaluation within this study provides the technical basis for the clinical integration of PET/MR hybrid imaging into RT treatment planning.


INTRODUCTION
For years computed tomography (CT) has been the basis for radiation therapy (RT) treatment planning and is still the main imaging modality for precise dose planning and target volume definition.It provides highly accurate three-dimensional (3D) anatomical information of the patient as well as lin-ear attenuation coefficients (LACs) of tissues and a direct relation to the electron density needed for dose calculation.However, state-of-the-art RT planning is increasingly based on additional imaging modalities.Magnetic resonance (MR) imaging for instance has become an important part in RT treatment planning for more precise delineation of the target volume 1,2 due to its excellent soft tissue contrast and the potential use of functional imaging parameters. 3Some studies even suggested RT planning based on only MR images without the additional use of CT. 4,5 Furthermore, positron emission tomography (PET) or PET/CT is increasingly used as an additional imaging modality improving the accuracy of the target volume delineation. 6,7  recent years, the new hybrid imaging modality PET/MR has become available for head/neck 8,9 and whole-body imaging 10 enabling simultaneous acquisition of both MR and PET data.2][13] A major challenge in PET/MR imaging is PET attenuation correction (AC) (Ref.14) a precondition to provide correct quantitative values with PET.In contrast to PET/CT, where the CT image provides information about the photon attenuation and can directly be used for AC of PET data, MR images do not provide such a direct relation to photon attenuation.Different proposals have been made including MR-based image segmentation, 15,16 atlasbased AC, 17 and pseudo-CTs using ultrashort echo time MR sequences. 18,19 urthermore, hardware devices, such as the patient table or radiofrequency (RF) coils have to be considered in PET/MR AC.For stationary and rigid devices, CTbased 3D attenuation maps (μ-maps) are stored in the system and used for PET AC if the object is located in the PET field-of-view (FOV).For flexible RF coils, this method is not applicable, since they vary in position and geometry and are inherently not visible in MR.AC methods using UTE sequences 20 or MR markers 21 have been suggested and evaluated.Despite all these challenges, PET/MR might play an important role in RT treatment planning 22 and a first intensity-modulated radiotherapy (IMRT) treatment planning with 68 Ga-DOTATOC-PET/MR, but without dedicated RT equipment has been evaluated. 23ntegration of PET/MR into RT treatment planning requires reproducible patient positioning devices including a flat tabletop and patient positioning aids that have to be MR and PET compatible.Current PET/CT tabletops for RT treatment planning are mostly based on carbon fibers that provide low photon attenuation but are otherwise not MR-compatible since carbon fibers are electrically conducting and potentially will produce MR image artifacts and signal voids due to eddy currents being generated in the carbon surface.Recently, developed MR tabletops for RT treatment planning are thus manufactured from glass fibers that, however, highly reduce the PET signal and might produce PET image artifacts and potentially increase PET quantification errors.Another consideration for PET/MR integration of RT planning is that local RF coils have to be used for high quality MR imaging without touching and deforming the surface of the patient.Thus, specifically designed RF coil holders are usually used in MRbased RT treatment planning, which then also have to be optimized toward reduced PET photon attenuation in the context of PET/MR imaging.
In this work, a flat whole-body RT table overlay and RF coil holders for head/neck imaging are introduced that are suitable for patient positioning in hybrid PET/MR imaging.All prototype components have been systematically evaluated toward MR and PET compatibility.MR image quality has been tested and compared with the current PET/MR imaging setup.PET photon attenuation has been evaluated quantitatively by phantom scans and clinical usage of the new RT setup in PET/MR imaging has also been tested with an initial in vivo study on two patients.

2.A. PET/MR hybrid system
Hybrid PET/MR imaging was performed on an integrated PET/MR whole-body hybrid system (Biograph mMR, Siemens AG Healthcare Sector, Erlangen, Germany), which enables simultaneous acquisition of PET and MR data.The hybrid system consists of an active shielded 3.0 T wholebody MR system with a maximum gradient strength of 45 mT/m and a maximum slew rate of 200 mT/m/ms.The integrated PET detector covering an axial field-of-view of 25.8 cm is located in its isocenter and is comprised of 8 detector rings, each with 56 lutetium oxyorthosilicate (LSO) scintillator crystal blocks each read out by a MR compatible array of avalanche photodiodes (APD).The hybrid PET/MR system is equipped with a full set of phased-array RF receiver coils that cover the patient's body and that connect to 32 independent RF receiver channels (Tim, Total imaging matrix, Siemens AG Healthcare Sector, Erlangen, Germany). 13ET AC of the hybrid system is performed in two steps.Attenuation maps of MR-visible objects ("human" μ-map) are based on a Dixon-VIBE sequence. 15Rigid and stationary RF coils and the patient table are included into the "hardware component" μ-map based on 3D CT images.Flexible RF coils are designed toward improved PET transparency and are currently not considered in PET AC of the standard PET/MR protocol.

2.B. RT equipment for the PET/MR system
Integration of PET/MR into radiation treatment planning requires dedicated equipment that has to be PET and MR compatible.The following equipment has been designed and evaluated within this study (Qfix, Avondale, PA) suitable for the Biograph mMR system: RT table overlay (198 cm length, 51 cm width, 15 mm thickness): The flat tabletop consists of plastic sandwich structure with a foam core and is equipped with a Varian Exact Style Indexing system.The RT overlay is placed on top of the PET/MR systems spine array RF coil (without the flexible foam cushion).The RT overlay fits into the fixation system of the head/neck RF coil of the PET/MR system table and two support blocks are affixed at the foot-end to guarantee rigidity and stability on the patient table [Fig.1(a)].
RF coil holders: Due to the presence of the RT overlay, the actual head/neck RF coil cannot be mounted and is moreover too small for head imaging of patients with RT head masks.Thus, two thin coil holders (about 3 mm thickness) were designed, which each fix one 6-channel flexible body matrix RF coil into a C-shape [Fig.1(b)].One RF coil holder on each side slides into a track system integrated at the head part of the 072505-3 RT table overlay to wrap around the patient's head as shown in Fig. 1(c).The two coil holders are clipped together at the top to fix and stabilize the setup.The cables of the body matrix RF coils point toward the systems bore and are thus not located in the PET FOV and do not have to be considered in PET AC.

2.C. Hardware component AC
CT acquisitions of all RT devices were performed on a stand-alone CT scanner (Somatom Definition FLASH, Siemens AG Healthcare Sector, Forchheim, Germany) to generate "hardware component" μ-maps for PET AC.All CT scans were acquired with identical scan parameters of 140 kV tube voltage, 400 mAs tube current, and the CT patient table was segmented out of the CT data.
To reduce noise around the RT overlay, a mask was generated by thresholding and closing operators, which only contains the material of the RT overlay.All residual background voxels of the CT image were set to −1000 Hounsfield units (HU).Subsequently, the CT image was scaled to LACs at 511 keV using the bilinear scaling approach 24 and a 5 mm Gaussian filter was applied.An adapted scaling for high attenuating materials was not necessary, since the RT overlay only consists of low LACs. 25he CT acquisition of the RF coil holders including the RF coils was performed with the RT table overlay placed on a flat tabletop to guarantee the same geometry as in the PET/MR hybrid system.Before scaling the CT image to LAC at 511 keV with an adapted conversion, 25 the RT table overlay and the flat tabletop were segmented out.The noise of the μmap including CT streak artifacts of the high attenuating elements was reduced by thresholding.All voxels below a LAC of 0.02 cm −1 were set to zero.The μ-map was also smoothed with a 5 mm Gaussian filter.

2.D. Repositioning accuracy of the RF head coil setup
A requirement for automatic CT-based AC of hardware devices in PET/MR hybrid imaging is an accurate repositioning of the specific component to ensure an identical position between μ-map and actual position.Since the RT table overlay is rigid and its position is fixed on the PET/MR systems patient table, validation of the coregistration of RT table overlay and its μ-map position has to be performed only once.
The attachment of the RF coil holders, which are mounted and removed between patient scans potentially has more variability and was thus examined toward repositioning accuracy.Two 68 Ge rod sources with a diameter of 8 mm also used for routine PET/MR alignment quality control measurements were attached to the outer surface of each RF coil holder.The RF coil holders and the flexible RF coils, respectively, were repeatedly mounted and removed five times.A 6 min PET acquisition was performed between each cycle and the absolute position and alignment along the active rod sources was evaluated by comparing the nonattenuation-corrected (NAC) PET images of all scans.

2.E. MR compatibility measurements
To ensure MR compatibility of both devices, disturbance of the MR shim and a potential signal detection from the RF components have been tested using a field sequence with 0.25 ppm/line and a gradient echo sequence with a very short echo time (TE = 2.21 ms), respectively.In addition, MR signal-to-noise ratio (SNR) measurements of the new RT head/neck imaging setup were performed (two 6channel body matrix RF coils) and compared with the standard PET/MR setup using a dedicated 16-channel head/neck RF coil.SNR-scaled maps of the central transaxial slice of a homogeneous cylinder phantom (7.3 l, distilled water with NiSO 4 ) were calculated by using a modified dual image acquisition and subtraction method.Image parameters of the gradient echo sequence were as follows: TR = 300 ms, TE = 20 ms, flip angle = 60 • , slice thickness = 3 mm, image matrix = 256 × 256, in-slice pixel size = 1.17 × 1.17 mm 2 .

2.F. MR volunteer measurements
The RT setup for head/neck imaging was evaluated with MR measurements on two healthy volunteers.Images were again compared to the standard PET/MR setup with the dedicated 16-channel head/neck RF coil.A T1-weighted 3D magnetization-prepared rapid gradient echo (MPRAGE) sequence was acquired with TR = 1900 ms, TE = 2.44 ms, and flip angle = 9 • .Image matrix was 512 × 512 with 192 slices, a pixel size of 0.49 × 0.49 mm 2 and 1 mm slice thickness.
Additionally, the MR Dixon-VIBE AC sequence was acquired to compare the MR-based μ-map of the volunteer's head between both RF coil setups.

2.G. PET compatibility measurements
Evaluation of the RT table overlay was performed with a cylindrical uniform phantom filled with water (4.5 l) and 200 MBq 18 F-fluoride.The active PET phantom was fixed to the PET/MR patient table such that difference measurements with/out RT table overlay in place was possible without moving the phantom [Fig.1(d)].The phantom was scanned with the RT table overlay (RT scan, 10 min) and without the RT table overlay (reference scan, 11 min) in place.The spine array RF coil was not mounted during both scans to avoid its effect on PET data acquisition.
The RF coil holders including the mounted flexible RF coils were quantitatively evaluated with a homogenous phantom bottle (1.9 l) filled with distilled water, NiSO 4 , and 54 MBq 18 F-fluoride.The phantom was scanned with the RF coil holders including the RF coils (RT scan, 10 min) as shown in Fig. 1(c) and without the RF coil holders (reference scan, 11 min).The RF coil holders were removed without moving the phantom bottle, ensuring identical positioning in both scans.
Longer scan duration of the reference scan of both setups was set to compensate for the decreased 18 F activity.For both setups, the MR-based "human" μ-map that includes the phantom fluid was replaced by a CT-based μ-map of the phantom to additionally correct for the plastic housing.Both "human" and "hardware component" μ-map and the overall combined μ-map are shown in Fig. 2, exemplarily for the RT head setup.
Photon attenuation of both hardware components was evaluated by comparing regions-of-interest (ROIs) between the reference scan and the RT scan.Tables I and II give an overview about the labels of the reconstructed PET images and about the hardware component AC details.For both setups, one large ROI (18 cm for the RT table and 15 cm for the RF coil holders) and five small ROIs (8 and 5 cm, respectively) were drawn inside the phantom.
All PET images have been reconstructed iteratively with a 3D ordered-subset expectation maximization (OSEM) algorithm with 3 iterations and 21 subsets smoothed with a 4 mm Gaussian filter.Image parameters of the PET images were: 127 image planes, 172 × 172 image matrix at 4.17 × 4.17 × 2.03 mm 3 ., pixel bandwidth = 965 Hz/pixel.PET data of the patients were also reconstructed iteratively with identical image parameters as for the phantom scans.

3.A. Repositioning accuracy of the RF head coil setup
The active rod sources were clearly visible in all PET NAC images as shown in Fig. 4(a) superimposed with the "hardware component" μ-map.The x-and y-positions of all four rod-sources were calculated at two different z-positions (overall 8 measurement points for each of n = 5 repetitions).The standard deviation in x-direction is calculated to be 1.0 and 0.4 mm in y-direction with a maximal error of 1.9 and 0.8 mm, respectively.Line profiles for one point are provided exemplarily in Fig. 4(b) for the x-direction and Fig. 4(c) for the y-direction.The axial z-direction was not evaluated since both RF coil holders slide into the track of the RT table overlay that fixes the z-direction.

3.B. MR compatibility measurements
MR measurements show that both devices, RT table overlay and RF coil holders, are MR compatible.No visible or measurable MR signal was detected from the hardware components, and the MR shim was not disturbed.
The SNR maps of both RF coil setups (16-channel dedicated PET/MR head RF coil and 2× 6-channel body matrix RF coil with coil holders) reveal increased SNR in the center of the phantom bottle compared to the peripheral parts of the bottle phantom [Figs.5(a) and 5(b)].Overall SNR of the MR image acquired with the 16-channel head/neck RF coil is about 25% higher than with the RT head setup.The overall signal homogeneity of both RF coil setups is comparable.For improved quantitative assessment and comparison of the SNR differences across the phantom, horizontal and vertical line profiles were drawn and plotted in Figs.5(c) and 5(d), respectively.

3.C. MR volunteer scans
T1-weighted MPRAGE images of the volunteer scan are shown in Fig. 6 for the actual PET/MR setup with the 16-channel head/neck RF coil (a) and the RT setup (b) in direct comparison.A zoomed section of the transaxial image (dashed square) is shown on the right for both setups.In direct visual comparison both images appear comparable providing high image quality with no distortion and no essential differences.The SNR is slightly lower for the scan acquired with the RT head RF coil setup (see zoomed section of Fig. 6).
The "human" μ-map based on the Dixon-VIBE sequence did not show any considerable differences or artifacts for the head region between both setups and thus, standard MR-based AC with the RT setup is feasible within this region.

3.D. PET compatibility measurements
Mean values (Bq/ml) of the large ROI analysis along the z-direction of the activity-filled phantom (slice 20-110) are plotted in Fig. 7(a) for the RT table overlay and in Fig. 7(c The mean count rate attenuation of the RT table overlay was 3.8%.If the RT table overlay is considered in AC the un-derestimation of the activity concentration is largely corrected and amounts less than 0.6%.Attenuation of the RF coil holders including the RF coils is around 13.8%, while this deviation is reduced to 0.8% if the RF coil holders and mounted RF coils are considered in AC.Error range, mean deviation, and standard deviation of all ROIs are listed in Tables IV and V for the RT overlay and the RF coil holders, respectively.Additional to the large ROIs, values are also listed for the small ROIs (see gray ROIs in Fig. 7).

3.E. Patient scan
The PET only image of patient #1 is shown in Fig. 8(a).No active lesions are observable in the head region of this patient.Thus, the activity distribution (kBq/ml) across the brain volume was compared between the PET images with/out AC of the RF coil holders.Difference images displayed in percent where the scan without the RF coil holders was used as a reference are shown in Fig. 8  mean/maximum of the standardized uptake values (SUV) were calculated and listed in Table VI.

DISCUSSION
In this work, a prototype RT table overlay and associated RF coil holders for head/neck imaging in the context of RT planning are introduced, which are suitable for hybrid PET/MR imaging.Both devices have been systematically tested toward MR and PET compatibility.MR image quality of the proposed RT head/neck setup with two 6-channel body matrix RF coils was compared with the standard PET/MR setup employing a 16-channel head/neck RF coil by phantom scans and volunteer scans.Repositioning accuracy, necessary for appropriate "hardware component" AC, was validated and AC methods of all hardware components have been tested with appropriate phantom scans.Furthermore, clinical application of the new devices with an integrated PET/MR system has been tested with an initial in vivo study on two patients.
Two RT devices have been tested in this context, a flat tabletop that is placed on top of the spine array RF coil on the Biograph mMR integrated PET/MR system and two RF coil holders that allow for head/neck imaging.The RT table overlay consists of a plastic sandwich composition with a foam core and is optimized toward reduced and homogeneous photon attenuation when used in conjunction with PET imaging.The composite sandwich construction of the RT table overlay provides a lightweight and mechanically stiff alternative to standard carbon fiber RT overlays that are not MR compatible in general.The RT table overlay has a fixed and defined position on the patient table and it provides a rigid geometry, which is a precondition for CT-based AC of this hardware component.The RF coil holders are designed to fix the body matrix RF coils into a C-shape and to be placed at the head end of the RT table overlay to wrap around the patient's head.
Both devices have been tested toward MR compatibility and image quality.No disturbance of the MR shim was observable and no signal was detected from the devices.Both RF coil setups provided high and diagnostic MR image quality with high signal homogeneity.The SNR of the suggested RT setup for head/neck imaging (two 6-channel body matrix RF coils) was decreased by about 25% when compared to the standard PET/MR setup with a dedicated 16-channel head/neck RF coil.While this effect is not considered a major limitation for RT treatment planning, it is attributable in most part to the larger distance of the RF coils from the scanned object in the RT setup.This larger distance on the other hand allows a fixation of the patient with an individualized thermoplastic head mask to guarantee accurate patient positioning of the patient.The RF coil holders are designed such that the patient mask and an additional mask holder frame has enough space to ensure a contact free setup between RF coils and patient.Volunteer scans demonstrated high image quality with the RF coil setup revealing slightly lower SNR compared to the standard head/neck RF coil as validated with the quantitative phantom SNR comparison.MR-based μ-maps of the volunteers/patients acquired with a Dixon-VIBE sequence did not show any artifacts and were usable for MR-based PET/MR AC.
For the MR head acquisitions within this study, image distortion was not observed with the RT head setup using two RT coil holders and two body matrix RF coils.However, when it comes to whole-body imaging, distortion of the MR image and truncations due to the limited MR FOV have to be considered for potential RT planning.Several approaches have been introduced recently that compensate for distortions and truncations in simultaneous PET/MR scanning. 26,27 r appropriate PET image quality and quantification, AC of the RT hardware components is essential.For easy and automatic AC of the devices, they should provide a rigid geometry and have a fixed position on the PET/MR systems patient table.Flexible RF surface coils are currently not considered in the routine PET/MR protocol, since their position and geometry during an examination is not known a priori.Different proposals have been made for position determination by using ultrashort echo time sequences 20 or MR markers. 21It has been shown that the tolerable shift of the flexible RF coil relative to its AC map is in the order of the PET resolution of 4 mm. 25The accuracy of multiple repositionings of the RF coil holders, tested with active 68 Ge rod sources, was calculated to be below 2 mm in x-direction and below 1 mm in y-direction.The smaller deviation in y-direction is due to the fact that the RF coil holders are placed directly on the patient table restricting the displacement in y-direction.Considering that the distance between the RF coil and the patient is larger than during a routine PET/MR scan where the RF coil is placed directly on top of the patient, the quantitative effect of this small misregistration might even be smaller.The small deviation of multiple repositionings demonstrates that robust CT-based AC of the proposed RT devices is feasible without using markers.It has to be assured though, that in routine clinical use the setup is mounted carefully to avoid obvious misplacements of the hardware components.
Overall photon attenuation of the RT table overlay was calculated to be 3.8% evaluated with a homogenous phantom.Compared to MR-only tabletops consisting of glass fibers this value is relatively low.This is due to the fact that the RT overlay is made of a lightweight plastic sandwich composite with a foam core and is optimized toward low and homogeneous photon attenuation.The same is true for the RF coil holders, which have a thickness of around 3 mm.The overall attenuation of the RT head/neck setup is calculated to be around 13.8%, which is in the range of the global attenuation of a standard MR head/neck RF coil that was measured to be 17%. 28The high attenuation has been expected, since each of the flexible body matrix RF coil has an overall attenuation of about 6% (Refs.20, 25, and 29) and altogether two RF coil holders with two body matrix RF coils are used for the RT setup.With CT-based μ-maps the error can be reduced to less than 1%, as shown within this work.
To use PET/MR data as a single imaging modality to base RT treatment planning and dose calculation on, dedicated algorithms that allow a transformation of MR values into pseudo-CT values 30 as well as a dedicated system for external-internal reference point definition are required.The definition of such isocenter point can be realized using a special laser system for RT simulation, which could be installed for combined PET/MR systems.However, in most clinical settings an additional planning CT for RT treatment planning and dose calculation will be required.In this case, robust and accurate algorithms for multimodal image registration are required. 31his study only includes the main hardware components used for RT planning with integrated PET/MR.As mentioned above, for more accurate patient positioning, individual thermoplastic head masks have to be used in combination with a fixation system.While a mask fixation system can be included to the "hardware component" μ-map of the RT overlay, thermoplastic head masks might be disregarded in PET AC since they are very thin, do not significantly affect PET quantification and can thus be neglected in the AC process.The additional integration of dedicated body RF coil holders will extend this concept to a whole-body imaging application.Overall, an integration of PET/MR into RT treatment planning with dedicated equipment is feasible and is expected to simplify the imaging and planning workflow, since instead of combining separate PET/CT and MR image information, only PET/MR imaging is necessary additional to the planning CT.

CONCLUSION
A RT table overlay and associated RF coil holders have been designed to integrate PET/MR hybrid imaging directly into RT treatment planning.All proposed RT hardware components are compatible with MR, PET, and combined PET/MR imaging and provide low attenuation of PET signals.Due to a mechanically rigid design and reproducible positioning of the RT hardware components on the patient table of the PET/MR systems, all components can be con-sidered during CT-based AC.Quantitative evaluation on phantoms and patients demonstrated a high repositioning accuracy and exact PET quantification following AC of the hardware components.The development of the proposed RT table overlay and RF coil holders and the systematic evaluation in this study provides the technical basis for integration of PET/MR hybrid imaging into RT treatment planning.

FIG. 1 .
FIG. 1.(a) RT table overlay placed on top of the spine array RF coil on the PET/MR hybrid system.(b) RF coil holder that is designed to fix the flexible 6-channel body matrix RF coil for RT head/neck imaging.(c) Mounted RT head/neck imaging setup including the RT table overlay, the RT coil holders, and the body matrix RF coils.(d) Setup used for the PET evaluation of the RT table overlay.
) for the RF coil holders.The percentage difference to the reference scan is shown in (b) and (d), respectively.

FIG. 4 .
FIG. 4. (a) Fusion of the PET NAC image (rod sources) and the "hardware component" μ-map.Line profiles through the line sources are plotted for each of the five repeated measurements in x-direction (b) and y-direction (c).

FIG. 5 .
FIG.5.Transaxial SNR maps of the 16-channel head/neck RF coil(a) and the RT setup with two 6-channel body matrix RF coils (b).Line profiles are plotted in horizontal (c) and vertical (d) direction for both setups.Note that the increased RF signal in the center of the phantom results from an inhomogeneous RF excitation profile inside the large bottle phantom.This is assumed not to affect the relative differences in the SNR measurements.

FIG. 6 .
FIG.6.MPRAGE images of the volunteer scans for the actual PET/MR setup with the 16-channel head/neck RF coil (a) and the proposed RT setup with two 6-channel body matrix RF coils (b).A zoomed section (dashed square) of the transaxial view is shown on the right for both setups.

FIG. 7 .
FIG. 7. (a) Mean activity concentration (Bq/ml) of the 18 cm ROI (white) for the RT table overlay setup.The values are plotted for the reference scan without the RT table overlay (RT_reference) and for the RT scan with AC (RT_AC) and without AC of the overlay (RT_noAC).(b) Percentage difference of scan with the RT table overlay (with/out AC) in relation to the reference scan.(c) Mean activity concentration inside of the 10 cm ROI (white) for the RT head setup.The values are plotted for the reference scan without the RF coil holders (CH_reference) and for the RT scan with AC (CH_AC) and without AC of the coil holders (CH_noAC).(d) Percentage difference of the RF coil holders (with/out AC) in relation to the references scan.

FIG. 8 .
FIG. 8. (a) PET-only image of the activity distribution [k Bq/ml] in the brain of patient #1.Difference images in percent of the activity concentration without (b) and with (c) AC of the RF coil holders.The reference scan was acquired without the RF coil holders.
FIG. 9. (a) Sagittal MIP of the PET-only image of patient #2.All five active lesions in the head/neck region were evaluated and corresponding SUVs are listed in Table VI.(b) Transversal PET/MR fusion image of the active lesion #4.

TABLE I .
PET setups for the phantom measurements with the RT table overlay.The label "RT" indicates that the evaluation is performed with the RT table overlay only."noAC" is related to the AC of the RT device.Both patients underwent a clinically indicated PET/CT examination before the additional acquisition on the PET/MR hybrid system was performed.Both patients provided written consent.No additional radiotracer was injected for the PET/MR examination.A PET/MR head respective head/neck examination (one bed position only) of two patients was performed with the RT table overlay in place, serving as standard of reference.Since the attenuation of the RT table overlay was expected to be small from the PET compatibility measurements and its position is fixed it was assumed that AC of the RT table will correct for its attenuation and should have only minor effects on the evaluation.Furthermore, the removal of the RT table overlay would have required the patient to move.A second PET/MR scan was then performed subsequently without moving the patients with the RF coil holders and RF coils mounted to evaluate their effect on PET images.Patient details and imaging parameters are listed in TableIII.Identical "human" μ-maps based on a Dixon-VIBE sequence were used for AC of both patients, respectively.The first patient was evaluated toward global activity concentration in the brain, the second patient toward SUVs of five active lesions in the head/neck region.Combined μ-maps of the patient and the hardware components are shown in Fig.3(a) for patient #1 and Fig. 3(b) for patient #2, respectively.Scan parameters of the coronal Dixon-VIBE sequence were: 192 × 126 matrix, 128 slices, and 2.6 × 2.6 × 2.6 mm 3 resolution, TR = 3.6 ms, TE = 2.46 ms, flip angle = 10

TABLE II .
PET setups for the phantom measurements with the RF coil holders overlay.The RT table overlay was present in all measurements and considered in AC.The label "CH" stands for coil holders."noAC" here is related to the AC of the two coil holders.

TABLE III .
Patient information and imaging parameters for the PET/MR scan.

TABLE IV .
Range, mean value, and standard deviation in percent for all z-slices of the ROIs inside the phantom bottle for the PET scan with the RT table overlay.Both are compared to the reference scan without the RT table overlay.Note that the 8 cm ROI evaluation includes the measured values of all five ROIs.

TABLE V .
Range, mean value, and standard deviation in percent for all z-slices of the ROIs inside the phantom bottle for the PET scan with the RF coil holders.Both are compared to the reference scan without the coil holders.Note that the 5 cm ROI evaluation includes all measured values of all five ROIs.

TABLE VI .
SUVs (mean and maximum) of the active lesions of patient #2 for the reference scan without RF coil holders (REF) and for the scan with the RF coil holders with AC (AC) and without AC (noAC).The percentage difference of the mean value to the reference scan is given in parentheses.