Artifacts and Diagnostic Pitfalls in Positron Emission Tomography-Magnetic Resonance Imaging Claudia Martinez-Rios, MD,* Raymond F. Muzic Jr, PhD,*,‡ Frank P. DiFilippo, PhD,§ Lingzhi Hu, PhD,║ Christian Rubbert, MD,¶ and Karin A. Herrmann, MD, PhD*,†

Introduction

P

ositron emission tomography-magnetic resonance (PETMR) is a new hybrid imaging modality that has recently been introduced in clinical practice for oncologic imaging and is increasingly being used in various clinical indications.1-5 PET-MR unifies the complementary capabilities of magnetic resonance imaging (MRI) and PET in a single imaging modality. Excellent anatomical and morphologic information with high soft tissue resolution and contrast from MRI and the best possible molecular, functional, and physiological information from PET are complementary and have the potential to provide maximum diagnostic information in a single procedure.1,6,7 However, the incremental diagnostic value of PET-MR over the current standard imaging procedures in various clinical scenarios is still under investigation. Initial results indicate that PET-MR is particularly promising in oncologic diseases2,5,8 and is at least equivalent to PET-computed tomography (PETCT) in lesion detection.9,10 Besides the hope for improved clinical and diagnostic performance, there are further expected benefits of PET-MR over current conventional imaging techniques: PET-MR harbors the potential to improve the patients’ workflow and save the patient and the administrative organization the *Department of Radiology, Case Western Reserve University, Cleveland, OH. †Department of Radiology, University Hospitals Case Medical Center, Cleveland, OH. ‡Case Center for Imaging Research, Case Western Reserve University, Cleveland, OH. §Department of Nuclear Medicine, Cleveland Clinic, Imaging Institute, Cleveland, OH. ║Philips Healthcare, Cleveland, OH. ¶Institute of Diagnostic and Interventional Radiology, University Hospitals, Düsseldorf, Germany. Address reprint requests to Karin A. Herrmann, MD, PhD, Department of Radiology, University Hospitals Case Medical Center, Case Western Reserve University, 11100 Euclid Ave, Cleveland, OH 44106. E-mail: [email protected]

http://dx.doi.org/10.1053/j.ro.2014.07.004 0037-198X/& 2014 Elsevier Inc. All rights reserved.

implications of numerous appointments for multiple imaging procedures.11 As such “one-stop-shop” imaging modality, PET-MR may be more time efficient than conventional imaging techniques, such as CT or even PET-CT.1,7,12 Furthermore, there is the potential to decrease the overall radiation exposure to the patient from diagnostic imaging; an issue that gains particular relevance in the world of pediatric patients and young adult patients with a need for repetitive imaging follow-up.8,13 Although, both PET and MRI have been well-established individual standalone imaging modalities for decades, the marriage of the 2 components in 1 device came along with significant hardware and software adjustments. Inherent to such adjustments is the risk and likelihood for new imaging effects and artifacts that need to be addressed to guarantee appropriate, reliable, and reproducible diagnostic interpretation. The fact that today’s commercially available models of PET-MR devices operate with significant vendor-dependent technical differences adds complexity to these systems.14,15 Differences in the technical approach to PET-MRI increases the necessity to identify, improve, or if not possible at least, describe technology-related imaging artifacts and effects and bring these effects to the attention of the ultimate user of this new technology. The present review first aims to provide an overview over the most common artifacts and pitfalls encountered with PET-MR. It describes their imaging characteristics and discusses the technical background and potential ways of mitigating these issues. It focuses on the potential clinical implications of these artifacts to increase the awareness of these challenges and helps avoid interpretation errors in the clinical use. Secondly, this review addresses the challenges in workflow and in setting up appropriate functional and practical imaging protocols for the use of this complex combined hybrid imaging modality. It will provide suggestions and examples of practical approaches. 1

2

PET-MR: Technical Background Differences in Technical Design and Vendor-Specific Modifications Current commercially available PET-MR devices include 3 essentially different technical approaches: (option a) the “simultaneous” approach with the PET detector ring inserted into the MRI system referred to as the “fully integrated simultaneous system”1,12; (option b) the “sequential” approach with the PET scanner in the same room connected to the MRI system via a rotating examination table as a “tandem arrangement” referred to as the “sequentially integrated system”; and (option c) the “sequential” approach with a traditional PET-CT system in vicinity to a MRI system sharing a shuttle table that connects to both physically independent systems, referred to as the “trimodality system PET/CT-MR.” Each approach has individual advantages but at the same time shares some technical challenges, including MR attenuation correction, artifacts, and imaging workflow.15 The fully integrated PET-MR system (option a) allows simultaneous data acquisition of PET and MRI at the same bed position.12,16,17 However, it has undergone the most extensive hardware changes from the corresponding separate PET and MRI systems in that the photomultipliers tubes were replaced by avalanche photodiodes (APDs) and shielding had to be adjusted to allow for the PET electronics to operate in the presence of quickly changing gradient fields.12 Although the use of APDs has the advantage of insensitivity to magnetic field, the compromise is that the resolution is not sufficient to support time-of-flight PET. Another disadvantage of the fully integrated system is that the entire procedure may be prolonged compared with PET-CT, if during whole-body imaging, multiple dedicated MR sequences are performed per each bed position exceeding the time slot per bed position in PET. In that case, PET acquisition of the head occurs at a considerably different time point than that of the pelvis and vice versa. However, the major advantage of this fully integrated system is that the misregistration from patient motion can be minimized and physiological effects and functional parameters between the 2 components can be exactly matched in time. In the sequential approach (option b), the data acquisition of the PET and the MRI occurs at different time points but 1 immediately after the other with the table rotating between both the systems.18 The patient remains in the exact same position during the entire procedure with the coil system in place. Although patient movement between the data acquisitions for PET and MRI may occur, it can be controlled and minimized with appropriate patient instruction and training to the technologists. Image fusion is performed based on the known relative positions of the PET and MR isocenters. Advantages of this system are less technical modifications to the individual components of the system, the uninterrupted workflow within each imaging step for PET and MRI, and the ability to operate the devices, specifically the MRI, independently. As (fast) photomultiplier tubes (and not APDs) are used, the system provides time of flight.19 Both of these options however require that attenuation correction is provided based on MR data sets.20-26 Also, the

C. Martinez-Rios et al 511-keV photons of PET interact in MRI-related electronic and nonelectronic hardware components such as coils and must be taken into account in attenuation and scatter correction of the PET data. In the sequential solution, a coil identification scan is performed before any imaging starts to locate any external devices, as archived attenuation templates are inserted into the corresponding locations during PET reconstruction. The trimodality system PET-CT þ MRI (option c) operates the PET-CT and the MRI independently, in separate rooms. The patient is transported between the PET-CT and MRI using a shuttle that docks to each patient bed and loads or unloads the patient without disturbing his or her position.27,28 This design offers the advantages of simplicity and robustness of low-dose CT for the PET image reconstruction and avoids technical complexities and performance degradation of operating a PET scanner in the presence of magnetic fields that are associated with MRI. However, the physical separation of the 2 systems with a moving shuttle table that needs to be disconnected from 1 and reconnected to the second modality appears to bear the highest risk of patient changing position between scans and of failed coregistration. From an operational standpoint however, the potential for independent use of the PET-CT and MR components may enable the user to be more time efficient in workflow and patient throughput. Awareness of these technical variations is essential to understand imaging effects and artifacts that may occur in PET-MRI with respect to the respective system used.11,14 Challenges of MR Attenuation Correction Compared With PET-CT Both the fully integrated PET-MR system and the sequentially integrated PET-MR system share the challenge to perform attenuation correction based on MRI.2,23,29,30 To understand technically induced artifacts in PET-MRI, knowledge of basic technical components and functionality is required. Attenuation of the 511-keV gamma rays of PET occurs owing to their interactions with inner-shell electrons in the case of photoelectric effect and outer-shell electrons in the case of Compton scatter. The probability of interaction depends on both the atomic number and the electron density of the material. As such, different tissues in the human body attenuate the gamma rays differently. Air, with its low density, and bone (or metal) with its higher density and atomic number, represent the 2 extremes of photon attenuation. Photon attenuation affects the PET image quality and quantification accuracy.31 To quantify PET registration and standardize this quantification, these variations in the human tissues need to be taken into consideration.31,32 Attenuation correction is a method to account for these differences in photon attenuation so that tracer uptake in PET can be quantified in a standardized fashion with the so-called standardized uptake values (SUV), which is simply the radioactivity concentration in the tissue normalized by injected dose and patient’s body size.31,33 In PET-CT, information on tissue-specific photon attenuation is derived from CT, which provides information on probability of interaction of photons with the electrons.31,34,35 A complication is that the interaction probability depends on the energy of the photons (x-rays in the case of CT and gamma

Artifacts and diagnostic pitfalls in PET-MRI rays in the case of PET). The methods for accounting for this energy dependence are mature. The most common approach is to use a bilinear relationship to convert the CT Hounsfield Units values to linear attenuation coefficients for 511-keV photons.36 This is done voxel-by-voxel without the need to segment tissue. The resulting image is referred to as attenuation map (Fig. 1). Determining photon attenuation from MR data is complicated and not fully resolved. There is no simple relation between MRI signal intensity and attenuation coefficients. The MRI signal depends on protons and their local environment, whereas photon attenuation in PET depends on electron density and atomic number. Because MRI does not measure these properties, MR-based attenuation correction must determine the linear attenuation coefficients by inference.25,26,37,38 An approach has been to use an atlas or database of MRI39-43 and corresponding CT images. Most commonly, this has been done for the head. MR data from a patient are compared to a database, and the CT scan matching the MR anatomy the best is then used for attenuation correction.44-47 Variations on this are also possible with the best-match data being stretched or warped to match the image of the current patient. A fundamental limitation to this approach is that anatomy, particularly of patients that may have pathologies and surgical history, can vary widely. An alternative approach, which addresses this pitfall and is applicable to body imaging, is to create a personalized attenuation map from MR data.43 A simple method implementation uses a fast, 3-dimensional T1-weighted gradientecho MR sequence and relies on an expected range of shapes to identify 3 “tissue classes”: air outside the body, lungs, and soft tissue on a regional level.48 An alternative is to use a Dixon sequence49,50 and use the information it provides to identify fat as an additional tissue class. The regions are then assigned

3 linear attenuation coefficients representative of such tissue types. Occasionally the methods will be compromised when a patient has an amputation, implant, or another deviation from normal anatomy.48,50-52 The general idea is to collect an informative MR data set from which the type of tissue that is present can be determined either regionally or voxel wise. For PET-MRI, air and bone represent a distinct challenge.44,53-55 Using conventional acquisition sequences, both have little MR signal and appear as “black” and are hard to differentiate. This is especially problematic as bone significantly attenuates the PET gamma rays, whereas air does not. Because of this, methods to differentiate bone and air have been pursued. The most popular approach uses ultra-short echo time (UTE) sequences.56-58 Although air has few protons and would never have much signal, bone does produce a signal provided it is collected at a time that is short (ie, echo time is r1 ms) compared with its T2* relaxation time. Thus, using UTE, it is possible to include bone as an additional class.50,57 Following the acquisition of the MR attenuation correction (MRAC) sequence, various tissue segmentation algorithms are applied using either model-based or threshold-based algorithm based on MR images to generate the MR-based PET attenuation map. Depending on the underlying MRAC sequence, 3-class, 4-class, or 5-class tissue segmentation models are used in both the fully integrated system and in the sequentially integrated system.50-52,59 The trimodality PET/ CT þ MRI still uses CT attenuation correction. In the early times of PET-MR, atlas-based CT attenuation was used for the purpose of quantification40,41; however, this technique is not widely applied given the advances and successes in developing MR-based attenuation. The quality, stability, and reliability of the MRAC sequences and the respective tissue segmentation models have significant

Figure 1 CT- and MR-based attenuation maps. Coronal presentation of whole-body PET-CT (A) with CT-based attenuation map CTAC (B) shown compared with nonenhanced 3D T1w GRE whole-body MRI in coronal presentation, with a 3-segment PET-MR attenuation map MRAC (D). Low-dose CT (A) is acquired for attenuation correction in PET-CT, which includes attenuation correction for the bone. The corresponding whole-body MR attenuation map (D) accounts for lung, soft tissues, and air but not bone. The curvilinear delineation of the periphery of the patient owing to truncation artifacts can be noted. 3D, 3-dimensional; CTAC, CT attenuation correction; GRE, gradient-echo. (Color version of figure is available online.)

C. Martinez-Rios et al

4 effect on the accuracy of quantification in PET-MR. Errors in both components can result in specific artifacts, which is discussed in detail in the following sections of the article.

Artifacts and Pitfalls in PET-MR Artifacts in PET-MRI may be classified into 3 major types and may relate to the following: 1) Technical design and hardware of the PET-MR system (vendor specific) 2) Effects of MR physics affecting MR attenuation correction 3) Physiological or medical conditions of the patient or the procedure

Artifacts Related to the Technical Design and Hardware Artifacts of PET-MRI inherent to the technology can result from the specific design of these new hybrid systems, including the magnet, the PET technology, the bore size, and the presence of new hardware such as radio frequency (RF) coils and electronics within the field of view (FOV).2,11,29,60 In addition, MR attenuation correction methods and segmentation models can be the source of artifacts, including the inherent limitations of MRI. In the fully integrated system where PET is inserted into an MRI, new hardware and software modifications were required to minimize interference of the 2-system components.2,5,11,14 MRI-compatible, magnetic field–insensitive photo detectors (APDs) were introduced in place of the conventional photomultiplier tubes. Specific shielding was established to avoid electromagnetic interference with the PET on the MRI side.16 Furthermore, it was necessary for both the fully integrated system and the sequentially integrated system to develop a method to account for the presence of additional MR-related electronics and hardware in the FOV during the PET.2,12,61 MR coils, for example, and other hardware produce attenuation and scatter in PET and require accurate corrections. For this purpose, the exact type and location of the hardware needs to be determined before each examination. Currently, there is no reliable solution to visualize the position of RF coils and other hardware in an MR image, therefore fixed positioning of these hardware are required for a PET-MR study. The fully integrated system (Siemens mMR) uses both fixed coils and special low-attenuation surface coils. The fixed coils are at known positions, and CT-measured attenuation is used. The surface coils are not included in the attenuation map; however, their effect is minimal.24 The use of coils other than those specially designed for mMR (see webside Siemens Medical systems) is not recommended. In the sequential system, a so-called “coil identification scan” is obtained before all diagnostic image acquisitions. Depending on the type of scans and coil configuration, corresponding preacquired coil attenuation templates are inserted into the attenuation map to account for the attenuation of these coils.

In general, there are 2 main approaches to acquire an attenuation map of MR hardware. A CT scan of the MR hardware can be used to image the attenuation of x-rays, which can be further converted into the attenuation of 511-keV via bilinear conversion. A more direct measurement deploys a PET scan with transmission source, and the benefit of this approach is its direct measurement of attenuation at 511-keV level.

Artifacts Related to MR Physics With Effect on MR Attenuation Correction A large number of artifacts in PET-MR originate from the fact that MRI instead of CT is used for attenuation correction and scatter correction thereafter. This includes that limitations inherent to MRI and specific to the nature of MRI physics translate into the process of attenuation correction and result in degradation of PET data. Limitations inherent to MRI are well known since its use in medical imaging. Appropriate image quality in MRI depends on a wide spectrum of factors related to the equipment, the imaging procedure, and the actual subject to be examined. Some of these factors include magnetic field strength, scanner design including bore size and length of the magnet, magnetic field homogeneity, and software. The factors that are more related to the imaged subject are susceptibility artifacts and signal voids from metal or air within or adjacent to the human body, aliasing artifacts from reduced FOV, the paucity of protons in a given tissue, and motion artifacts from organs. All of these are well known to negatively affect image quality in MRI and can have an effect on diagnostic confidence. In PET-MR, these effects are also present with the same implications. In addition, however, these effects risk to negatively interfere with the process of tissue segmentation and MRAC. This will in addition result in image quality compromise or artifactual image information in PET. To avoid diagnostic misinterpretation of these effects, familiarity with these artifacts, their appearance, and their sources is pivotal. Truncation Artifacts Truncation describes an imaging effect where there is lack of registration of body parts outside a given FOV. This effect results in a more or less abrupt cutoff of the body structures in the periphery of an image, typically laterally, given the larger transverse than anterior-posterior diameter of a human body. Truncation is seen in both CT and MRI and is also a wellknown effect in PET-CT.32,60,62,63 In PET-CT, truncation produces streak artifacts at the boundaries of the CT image, which alter the attenuation coefficients in the area of the streaks.64 This phenomenon results in an underestimation of radioactivity concentration along lines that pass through the truncated area. Generally, the degree of underestimation is greatest in or near the truncated area. Lesions with increased tracer uptake may not be represented adequately in the PET images. Some PET-CT scanners generate a separate nondiagnostic CT image with extended FOV to improve PET quantitative accuracy. Truncation artifacts in PET-MR occur because the physical FOV in MRI is smaller than the FOV in PET. These imaging artifacts are frequently seen in large patients and when patients

Artifacts and diagnostic pitfalls in PET-MRI are scanned with their arms down. Truncation may also occur when patients are positioned away from the in-plane center. In PET-MRI, truncation artifacts are mostly owing to the off resonance at the edge of MR FOV caused by shimming irons and a limited bore size in the MR portion of the PET-MR scanner. The bore size in the integrated PET-MR system is 60 cm, in the sequentially integrated system 60 cm for the MRI scanner and 70 cm for the PET scanner, whereas the physical FOV of MR is usually even more limited because of field inhomogeneity.29,60,62,63 Truncation will cause an incomplete encompassing of the patient in the transverse planes with an incomplete attenuation map and, as a result, artifacts in the PET image. Both truncation and magnetic field nonuniformities at the edge of the transverse MR images translate into sinusoidal borders in the reformatted coronal MR images65,66 (Fig. 2). When the patient’s dimensions extend beyond the physical FOV, the extended part is “truncated,” meaning not represented in the reconstructed images. This leads to a lack of attenuation correction values for the corresponding region in the PET scan emission data. The result is a defect in the PET attenuationcorrected images, which can account for 15%-50% error in attenuation correction and subsequent underestimation of the SUV if it is not compensated for.32,64 The effect may result in false-negative findings in areas, which are not correctly attenuation corrected. This error can be decreased to o10% if truncation compensation is performed.63,67 Different algorithms have been proposed to correct truncation artifacts for both PET-CT and PET-MRI.60,62,68 An approach that is already known from PET-CT and was published early on by Delso et al63 for PET-MR is to use the nonattenuation-corrected emission data of the PET images,

5 determine the true contour of the body including the areas that were subject to truncation, and create a compensated attenuation map from this data. The compensated map is then fused with the truncated attenuation map (Fig. 3) and the missing fields are assigned the appropriate attenuation coefficients. The MR data are used within the FOV, and the PET-derived attenuation information is added outside the MR FOV.68 This method is used in both the fully integrated system and the sequential system. Transaxial slice-by-slice truncation compensation algorithms and phantom data may be part of this process among other postprocessing steps to create the corrected attenuation map. Software for this process is typically provided by the vendor as part of the product. A similar artifact is caused by magnetic field heterogeneity. It refers to the image distortion, which occurs in the periphery of a magnetic field due to nonhomogeneous magnetic field. These image distortions again are at risk to entail errors in tissue segmentation and attenuation correction.65 Wraparound or Aliasing Artifact The wraparound or aliasing artifact, also known as fold over artifact, is another common artifact in MRI. It occurs if the desired FOV is smaller than the body part to be imaged along the phase-encoding directions. Any structure located beyond the defined FOV will project into or “fold into” the reconstructed MR image. The risk is poor image quality in the anatomical reference image and potentially errors in image interpretation and diagnosis, if these artifacts are interpreted as pathologic lesions. If aliasing artifacts occur in an atMR sequence in PET-MR, they risk interfering with the segmentation algorithm. The result may be missegmentation and erroneous assignment of

Figure 2 Truncation artifacts and their correction. Truncation occurs when a given object is incompletely included in a given field of view. (A) coronal image of the same patient with CT (color coded red) and whole-body (WB) MRI (underlying anatomical image) digitally fused. The true body contour of the patient (CT) exceeds the field of view in MRI (arrows). The MRAC map (B) generated from the WB-MRI shows sinusoidal or curvilinear delineation of the body contour alongside the lateral borders, which translates into areas of inaccurate attenuation and thus linear irregularities in the PET image (arrow in (C)). The attenuated PET image acquired with PET-MR demonstrates no abnormality (D). These areas of tissue subject to truncation can be corrected for using the information from the MR-NAC image (Fig. 3). NAC, nonattenuation-corrected. (Color version of figure is available online.)

C. Martinez-Rios et al

6

Figure 3 Method for correction of truncation artifacts. Unenhanced 3D T1-weighted MRAC (A) acquired for attenuation correction demonstrates curvilinear delineation of the contour of the arm due to truncation. From this image, a noncompensated map is generated (red) (B). The nonattenuation-corrected (NAC) PET data include image information outside the truncated MR field of view (green areas in (C)) and determine the true contour of the body. The discrepant areas between the truncated WB-MRI and the NAC PET are added to the truncated attenuation map (green areas in (D)) to compensate for the attenuation correction error in this region. Reconstruction with the corrected map will help to compensate for these attenuation errors (D). 3D, 3-dimensional; WB, whole-body. (Color version of figure is available online.)

attenuation coefficients. If the same MRAC sequence is used also for anatomical reference, these artifacts may mimic pathology and lead to diagnostic misinterpretation. Solutions to correct for aliasing artifacts include changing of phaseencoding direction or k-space oversampling, which could potentially result in increased acquisition time.

Pulsation Artifact Pulsation artifacts are owing to modulation of the k-space data in phase-encoding plane in MRI caused by different factors including pulsatile flow, cardiac motion, and vascular pulsation or systemic pulsation. This artifact and its clinical implications are not specific to PET-MR but are a wellknown phenomenon in MRI. Specific to PET-MR, however, is the fact that when a pulsation artifact occurs, usually in the chest region in the MRAC, it may affect tissue segmentation and therefore result in artifact in PET image as well as incorrect quantification of tracer uptake within the tissue in this area. If the same MRAC sequence is also used for anatomical reference, there is again a risk to interpret these artifacts as pathology and false-positive findings (Fig. 4). The solution to correct pulsation artifact is to optimize basic data acquisition of the MRAC sequence and the segmentation algorithm.69 This can be done by choosing appropriate flow suppression techniques and changing the frequency or phase direction of the image acquisition. To avoid misinterpretation of this falsepositive lesion as pathologic abnormality, comparison with other imaging sequences of the same area needs to be taken into consideration.

Artifacts Related to Human Physiology or Medical Conditions Sometimes, human body physiology and medical conditions are at the origin of imaging artifacts. A large number of these challenges are already known from PET-CT and are now similarly encountered in PET-MR. A considerable challenge in MRI and thus also in PET-MR is the phenomenon of motion in the human body. As the goal is to coregister the anatomical and functional information from MRI with the metabolic information from PET, any motion carries the risk of falsified registration and errors in alignment, segmentation, and quantification. Motion encountered in the human body can be voluntary such as change of position of the body parts during the examination or breathing. It can be involuntary and related to many physiological processes within the human body such as cardiac contraction, vascular pulsation, bowel peristalsis, and muscle fasciculation. A possible advantage of simultaneous PET-MR is the ability to perform MR-derived motion correction of PET data. Initial reports are encouraging, and hopefully this approach will be available soon in clinical PET-MR systems.70 Respiratory Motion Respiratory motion during scanning causes the most prevalent artifact in PET-CT imaging.32,65 Whole-body PET data in both PET-CT and PET-MR is typically acquired stepwise in several bed positions in craniocaudal direction or reversely. Per each bed position, PET data acquisition time is typically selected in the range between 1 and 4 minutes depending on the patient body

Artifacts and diagnostic pitfalls in PET-MRI

Figure 4 Pulsation artifact in the WB-MRI acquired for attenuation correction. Axial unenhanced 3D T1-weighted GRE (A), PETMR (B), and fused PET-MR images (C) are demonstrated at the level of the aortic arch. A focus of increased tracer uptake in the left lung correlates with a soft tissue mass on T1w imaging (black and white arrows) and is consistent with a primary lung malignancy in this patient. Likewise, there is increased FDG uptake in 2 foci along the aortic arch corresponding to lymphadenopathy. A pseudolesion is seen in the right anterior lung field of the same level, which however corresponds to a pulsation artifact from the ascending aorta (yellow arrow). 3D, 3-dimensional; GRE, gradient-echo; T1w, T1-weighted; WB, whole-body. (Color version of figure is available online.)

7 habitus, amount of radioactivity injected, elapsed time since tracer injection, scanner characteristics (eg, time of flight and sensitivity), and image quality preferences of the interpreting physician. During data acquisition in each bed position, the patient is instructed to breathe normally. As PET data acquisition occurs over several minutes, the reconstructed PET image reflects the position of the diaphragm averaged over multiple breathing cycles. In both CT and MRI, however, images suffer from motion in that they are blurred, lacking sharpness of details, and exhibiting disturbing double contouring. To avoid this effect in MRI, imaging in breath hold is preferred with 1 breath hold being between 15 and 20 seconds. When imaged with breath hold, however, the position of the diaphragm depends on the depth of inspiration and may vary by several centimeters in craniocaudal extension. This fact can cause significant misalignment of structures close to the diaphragm within the lung recess or the liver dome or the spleen. The result is a crescent-shaped photopenic zone in the lungs or in adjacent organs parallel to and following the contour of the diaphragm (Fig. 5). Respiratory motion artifacts are related to this discrepancy in position of the diaphragm between the CT or MRI and the PET. If the MRAC sequence is acquired in free-breathing mode, the position of the diaphragm between PET and MRI may be relatively close and misalignment can be minimized. However, the images suffer from respiratory artifacts in MRI, and these artifacts increase the risk to cause errors in segmentation when creating the attenuation map. As a result, lesions in the lung, liver or spleen or other organs close to the diaphragm may be falsely aligned and measure inaccurate SUV. To achieve appropriate image quality in PET-MR and to avoid misalignment, the acquisition of the MRAC sequence and the sequence for anatomical reference is recommended in breath hold. Recent studies have shown that end-expiratory breath hold or shallow free-breathing MR results in more accurate alignment of thoracic structures than MRAC sequence acquired in end-inspiratory phase.15 Breath hold in midinspiration with intermediate positioning of the diaphragm is another alternative to end-inspiratory breath hold or randomly free-breathing acquisition.15 These breathing instructions are particularly relevant for examinations in the sequentially designed scanner. For the scanner design with simultaneous acquisition of PET and MRI data, a number of techniques with respiratory gating and MR-based respiratory motion correction have recently been proposed to improve the alignment between both the images in the chest region.71 Bone Artifacts In current 3-class and 4-class segmentation models for attenuation correction, bone is not taken into account. Therefore, accurate quantification of tracer activity suffers particularly within cortical bone and in the immediate vicinity to bone, especially in the pelvis.35,66,72 The lack of correction for photon attenuation, resulting from cortical bone results in underestimation of lesions in or adjacent to bone by 11.2% ⫾ 5.4% (range: 1.5%-30.8%) and 3.2% ⫾ 1.7%

8

C. Martinez-Rios et al applied mainly in the skull and brain, whereas whole-body applications are still in development and under investigation.73 Bowel Motility Bowel motility is again an issue for MRI as motion degrades the image quality and decreases diagnostic confidence. Multiple sequence types are available, which are relatively robust against motion artifacts; above all half Fourier acquired single-shot turbo spin-echo sequences. They deliver a relatively stable image quality even for bowel structures with a very high soft tissue contrast, which is superior to that of low-dose noncontrast CT. Although they are not used for attenuation correction, they still improve the anatomical morphologic image information. Bowel motility itself has no relevant effect on the MR-based attenuation maps. Air-filled bowel loops, by contrast, may undergo tissue segmentation in PET-MR and be assigned respective attenuation coefficients of air (Fig. 6). However, in our experience, it is relatively inconsistent and may or may not happen. Also, the presence of a contrast agent in the stomach (see below) may cause segmentation of this specific area. How this is relevant or misleading for the clinical interpretation has not been systematically, yet. The issue with bowel structures in PET-MRI is more that of physiological bowel activity, which sporadically results in increased tracer metabolism in PET. This phenomenon is well known from PET-CT and is physiological. Correct diagnostic

Figure 5 Artifacts secondary to respiratory motion. 3D T1-weigthed GRE sequence (not shown) for attenuation correction—if acquired in freebreathing mode—is subject to artifacts related to motion from respiration and therefore inaccurate for anatomical localization. If acquired in breath hold, the position of the diaphragm is important with regard to the average position of the diaphragm during PET acquisition to obtain accurate coregistration. In inspiration and breath hold, the position of the diaphragm is likely to be different and lower than the average position in PET. This results in crescent-shaped areas of misregistration (red arrows) paralleling the contours of the diaphragm as seen on sagittal and coronal views of PETMR (A and B). This effect is also see in PET-CT, unless CT is acquired in free-breathing mode where the diaphragm is in similar position than during PET (C and D). A solution to avoid this phenomenon is acquiring the MRAC sequence in midinspiration or in end exspiration. 3D, 3-dimensional; GRE, gradient-echo. (Color version of figure is available online.)

(range: 0.2%-4%).53 With newly developed sequences such as UTE and others, this issue may be overcome in the future; however, to date, these sequences have been successfully

Figure 6 Segmentation of air within bowel loops. Three segment-based attenuation correction models sometimes also account for air in the bowel based on the 3D T1-weighted MRAC sequence. Whether this affects accurate quantification in the abdomen has not been shown in the literature. However, air in the bowel may interfere with the correct segmentation of the lungs if close to diaphragmatic structures (eg, gastric fundus or colon flexure). This example of a coronal (A) and sagittal (B) whole-body MRAC map demonstrates correct delineation of the lungs, soft tissues, and air. 3D, 3-dimensional.

Artifacts and diagnostic pitfalls in PET-MRI interpretation is easy when this increased tracer uptake is longitudinal or linear; however, when more focal, the distinction from true lesions in the bowel or adjacent to the bowel is difficult with PET. A particular challenge for PET are lesions associated with the peritoneal surface of the bowel such as peritoneal implants secondary to malignancies. In these conditions, initial experience shows that PET-MR is actually better than PET-CT owing to its superior soft tissue contrast. Abnormal lesions close to bowel are better identified on MRI, which improves the interpretation of tracer uptake in the bowel. However, this initial experience still lacks confirmation in larger populations.

Artifacts Due to Specific Medical Conditions Other typical artifacts inherent to MRI are owing to increased susceptibility, which most of the times result from the presence of air or metallic or ferromagnetic objects within the human body. These substances destroy or significantly perturb the local magnetic field in their immediate vicinity and thereby cause local distortion of the MR signal and image.74 The corresponding effect in the image is a signal void blooming pattern and other irregular artificial patterns. Most of the time and depending on the size, shape, position and type of material, the signal void exceeds the true spatial extent of the object itself. Different sequences are variably sensitive or robust against these artifacts. A large variety of metallic objects implanted in the human body cause these susceptibility artifacts and signal voids: metallic port-catheter systems (mPCS), dental restoration material, sternal wires after sternotomy, joint replacements and prostheses, spinal fusion hardware, and many others. As for PET-MR, there are 2 major undesirable effects that result from susceptibility artifacts of such kind: (1) the lack of diagnostic information and inaccurate attenuation correction locally in the area of the signal void itself and (2) the interference with the tissue segmentation algorithm of the whole body and errors in the resulting attenuation maps. In an MR-based attenuation map, the areas of signal distortion and complete signal void from metallic susceptibility are typically assigned the attenuation coefficient of air. This misclassification in the attenuation map eventually causes quantification errors and results in a photopenic zone in the attenuation-corrected PET image (Fig. 7) with a considerable underestimation of the SUV with 2-deoxy-2-[18F]fluoro-D-glucose (FDG).75,76 Depending on the anatomical location of the signal void, particularly when close to the lung and chest, the model-based segmentation algorithm may be perturbed. If an area of signal void is located close to the chest wall, it may disrupt the outer contour of the lungs and chest. As the shape and volume of the lungs are integrated variables of the tissue segmentation algorithm, any interference with the pattern recognition and growing algorithm will perturb the correct function of the segmentation algorithm. As a consequence, an entire organ, an anatomical region, a or body part (mostly the lung) may be

9 assigned an incorrect attenuation coefficient (Figs. 8 and 9) and therefore be subject to incorrect quantification. Besides these effects on the attenuation correction in PETMR, metallic implants cause high-photon absorption and attenuate 511-keV photons. The fact that there is no radioactivity in the region of the implant itself causes the “cold” area in the reconstructed PET image of PET-MR. In addition, because it is difficult to assign correction LCA to these implanted devices, PET quantitation accuracy is usually significantly compromised in the organs with implants. When analyzing this effect in PET-CT the streaking artifact originating from metal actually is attributed a higher attenuation coefficient, which by contrast results in focal areas of overestimation of the FDG activity in the region of and immediately next to the metallic implant (Fig. 10). It is well known that this increases the risk of misdiagnosis with regard to potential tumor recurrence in the immediate vicinity to an implant. Software for metallic artifact reduction is available to minimize this effect in PET-CT.77 In the following paragraphs, some of the diagnostic challenges in PET-MR are discussed in more detail and with regard to the type of metallic implants, its location, and type of artifact. Dental Restoration Hardware Dental hardware is well known to produce artifacts in PET-CT. The beam hardening from the metal induces attenuation correction errors as seen with other metal implants in the body. The result in PET-CT is a higher attenuation value and a pseudoactivity, which can result in false-positive lesions as mentioned earlier. The mapping of Hounsfield units in CT to linear attenuation coefficient at 511 keV was designed for tissue-type materials. In the presence of metal, it does not provide reliable attenuation coefficients. With PET-MR and MR-based attenuation correction, the signal voids from susceptibility induced by metal will result in underestimation of the tissue attenuation coefficient and eventually in underestimation of FDG activity in this area. The effect of dental implants on PET-MR attenuation correction maps is extremely variable and unpredictable. Depending on the material, susceptibility may cause larger or smaller signal voids. As the shape and asymmetry are arbitrary, these effects are difficult to be accounted for with pattern recognition or growing algorithms. Furthermore, different sequence types are variably sensitive to metallic susceptibility (Fig. 7). The head with the presence of the nasal sinuses, the nasal cavity, the facial bones, and the oral cavity are extremely complex anatomical regions, which represent a considerable challenge for the segmentation algorithm. The diagnostic effect of this is mainly in areas where air, soft tissue bone, and metal are in close anatomical vicinity to each other. As a consequence, the effect of this effect on tumors of the head and neck and the skull base and lymphadenopathy is significant. This area is still a work in progress. Initial clinical experience with PET-MR is reported in 32 patients and 66 lesions. Results indicate that for head and neck cancers, there is a high correlation of SUVmean and SUVmax in pathologic lesions with the absolute quantification of PET-MR generally being slightly

C. Martinez-Rios et al

10

Figure 7 Artifacts from dental implants. The extent and shape of susceptibility artifacts in 3D T1-weighted GRE sequence derived from metal secondary to dental implants are dependent on the type of material and are oftentimes unpredictable. They may be minor or major. Axial image at the level of the mandible shows a significant signal void (asterix), on unenhanced 3D T1weighted acquired for attenuation correction (A), MRAC map (B), and contrast-enhanced Dixon (water) images derived from a metallic artifact from a dental restoration. It can be noted that this did not affect the correct segmentation of the lungs. 3D, 3-dimensional; GRE, gradient-echo; T1w, T1-weighted. (Color version of figure is available online.)

lower than with PET-CT.46,78 This article addresses head and neck tumors in general without specific focus on the regions with critical attenuation correction. Therefore, further studies will be needed to address specifically the effect of these artifacts on clinical practice in various areas of the head and neck. Sternal Wires Susceptibility artifacts and signal voids from sternal wires secondary to sternotomy may present a specific problem for the model-based segmentation algorithm. These typically round and focal artifacts in the midline of the anterior chest wall interrupt the body surface contour and mimic continuity between the outside air space and the lung field. This causes the model-based algorithm to fail and to attribute erroneous tissue attenuation factors (Fig. 9). This may apply to 1 lung or both the lungs. Errors in SUV are the consequence. In these cases, it is helpful to refer to the source images of MRI or MRAC

sequence, and the respective attenuation map to identify the signal voids in this region. Metallic Port-Catheter Systems A similar phenomenon may occur with mPCS. These are mostly implanted within the subcutaneous tissue of the upper chest wall. The port chamber produces a characteristic metallic susceptibility artifact, which is typically round or oval. Depending on the material, the resulting signal void and artifact pattern may be limited to the skin level and may not affect any further anatomical structures of the chest. In this case, it is of minor diagnostic relevance. A certain diagnostic challenge, however, may arise in individual cases, if the radiotracer is injected through the port-catheter system. A residual amount of radioactive tracer left in the port chamber or catheter or at the catheter tip will cause a focus of activity, which may be misleading. The fact that in the corrected PET image of PETMR, the port chamber is not visible, and in the MR image, the

Artifacts and diagnostic pitfalls in PET-MRI

11 contour may cause the segmentation algorithm to fail and will cause errors in the attenuation correction. In extreme cases, this involves the entirety of the lung field on 1 side (Fig. 9). Joint Replacements Joint replacement is of the most frequent sources of metalinduced susceptibility artifacts, resulting in larger areas of signal void and respective photopenic areas in PET (Fig. 10). If located in the pelvis or the knee, they are less likely to negatively affect the creation of the attenuation map. However, segmentation errors still have been observed, especially when bilateral hip joint replacements are present. The resulting signal voids may be symmetric in these cases and resemble the shape of the lungs, which puts the algorithm again at risk to fail. Even if the signal void is registered as air instead of lung tissue, the resultant can be significant photopenic regions in the PET image, which extends beyond the true dimensions of the implant and can obscure relevant clinical information. To avoid errors and misinterpretation resulting from these artifacts, it is mandatory to be aware of these effects. One solution to the problem is to always refer to the nonattenuation-corrected images and to monitor the quality of the attenuation map.79 It is desirable that in the future, similar standards of practice apply for PET-MR as for PET-CT where the guidelines require archiving both the attenuationcorrected and nonattenuation-corrected images. For some of these metallic artifacts, a semiautomated software tool has been proposed in the literature to correct for the PET quantification bias using a manual correction (or fill-in algorithm) before MRAC restoring quantitative activity distribution.80 In a recent article by Schramm et al,81 a newly developed automatic method for compensation of metalimplant–induced segmentation errors in MR-based attenuation maps was presented to correct for these errors. This method is not generally implemented in current PET-MR systems as a product, yet. It emphasizes, however, the needs for correction of segmentation failure owing to the significant effect on tracer quantification.

Figure 8 Artifact from metallic port-catheter systems. Susceptibility from metal in port reservoir cause signal voids in the level of the skin or subcutaneous tissue. Axial MR attenuation map (A), unenhanced 3D T1-weighted (B), and fused PET-MR image (C) at the level of the upper chest demonstrate a loss of emission data on the MR map (A) and a signal void on the MR images (red caliper). 3D, 3-dimensional. (Color version of figure is available online.)

catheters with radiopaque marking may not be visible, these misinterpretations can occur. In these cases, the signal void within the respective source MR image will be key to providing the diagnostic hint. The fact, however, that this does not always occur requires active surveillance of the quality of the attenuation map to account for these errors (Fig. 9). If the signal void of the port-catheter system however is large enough, it is at risk to mimic a direct “communication” between the air space outside the subject and the lung fields within the chest cavity. This disruption of the outer body

Artifacts Related to the Procedure Misalignment Due to Voluntary Motion of Body Parts Voluntary or involuntary patient motion is a common source of artifacts and is well known from PET-CT. Besides respiration and bowel motion, which are rather involuntary and were already addressed, patients may voluntarily move body parts during the image acquisition, resulting in misalignment of images between PET and the anatomical reference, CT or MRI. Except in fully integrated simultaneous systems, each part of the examination, PET and MRI is usually acquired within different time frames; therefore, the registration between both the types of images is prone to misalignment. In addition, the likelihood for this motion or change in position to occur in PET-MR depends on the scanner design. Certain authors have claimed that there is better alignment and less misregistration with fully integrated systems.15 However, clear instructions to the patient oftentimes help to avoid these issues even in systems with sequential design.

C. Martinez-Rios et al

12

Figure 9 Artifact from metallic sternal wires with segmentation error. Sternal wires and metallic port-catheter systems (mPCS) can result in signal void, which may disrupt the outer body profile and contour in the MR image. Axial unenhanced 3D T1-weighted (A) GRE sequence demonstrates a signal void from metallic wires (A) at the level of the sternum, which results in erroneous segmentation in the MR attenuation map (B). The disrupted contour is misleading the segmentation algorithm. As a result, the right hemithorax is “considered” to be in continuity with the outside air and is attributed an attenuation factor of air instead of lung that falsifies the quantitative measurements within the lung. The opposite lung is not “recognized” as symmetric lung cavity and is assigned m-values of soft tissue and air in a spotted pattern (appearance of the “spotted lung”) (B). 3D, 3-dimensional; GRE, gradient-echo.

Misalignment of the Bladder An organ that is usually subject to misalignment is the urinary bladder. Given the continuous physiological

secretion of urine into the urinary bladder and the desired hydration to eliminate the PET-tracer, the bladder may significantly change in volume during the image

Figure 10 Artifact from metallic implants and orthopedic hardware. Metallic streak artifact in CT (first column) results in focal lack of emission data in the region of the implant and adjacent areas of overestimation of the tracer activity on the PET images immediately next to the implant (red arrows). This is due to a higher attenuation coefficient originating from the metallic implant. The overestimation of tracer uptake may lead to misinterpretation as tumor recurrence or inflammation in certain clinical conditions. On MRI (axial, sagittal, and coronal 3D T1-weighted sequences in second, third, and fourth columns, respectively), the respective signal void and lack of emission data are also noted represented by resultant photopenic regions on the PET. However, no overestimation of tracer uptake observed in PET-MRI. 3D, 3-dimensional. (Color version of figure is available online.)

Artifacts and diagnostic pitfalls in PET-MRI

13

Figure 11 Example of error in lung segmentation. On coronal unenhanced 3D T1-weighted GRE image (A) and corresponding MR attenuation map (B), the diaphragm is in different positions (white arrows). This is even better demonstrated when fusing the images (white arrows in (C)).The error occurred because the air pocket in the gastric fundus (not shown) was segmented as lung. Corresponding PET-MRI (D) demonstrates errors in assignment of m-values in these areas, affecting the corresponding PET images with artifactual photopenic regions (black arrows in (C) and (D)). 3D, 3-dimensional; GRE, gradient-echo. (Color version of figure is available online.)

acquisition in PET-MR. This again may be less of a concern, when PET and MRI sequences of the pelvis are acquired simultaneously. In sequentially designed systems, the workflow should ideally be optimized such that the timing delay between the PET component and the MRI component is minimal. Sequential PET-MR design offers the advantage that MRI can be performed during the uptake phase of 18F-FDG before starting PET imaging. This saves the patient and personnel the additional hour of waiting for the distribution of the tracer. If MRI is performed during the uptake phase, however, the volume of the bladder may change considerably. As a solution to this problem, a fast MR location sequence for image fusion with PET (eg, single-shot turbo spin-echo sequences) may be obtained just before completion of the MRI protocol to ensure most accurate registration with subsequent PET. A disadvantage of this approach is that the PET scan will have intense activity in the bladder, which may complicate interpretation of nearby regions. If scheduling permits the PET study to be performed before the MR study (to give the patient the opportunity to void immediately before the PET study), then PET image quality will be improved, and radiation dose to the bladder wall will be minimized. Artifacts Due to Presence of Contrast Agents Specifically in a setting where Dixon is used for attenuation correction and Dixon water-fat segmentation is applied, another artifact is worth mentioning. This artifact is related to the presence of contrast agents within hollow or solid organs or structures creating missegmentation. The presence of contrast agents typically increases the T1 relaxation time and could potentially result in higher attenuation of 511-keV photons than water. Special positive contrast offered by MR contrast, if not accounted for during system design, could potentially lead to segmentation failure and inaccurate tissue attenuation coefficient assignment. This may happen if the stomach is filled with a contrast agent from, for example, superparamagnetic iron oxide particles, or the liver or spleen

are enhancing after intravenous administration of gadolinium contrast agent. This causes signal inversion when using the standard Dixon water-fat segmentation.65 A similar phenomenon may be seen, if the liver or the spleen demonstrates abnormally decreased signal intensity (eg, in iron storage and hemosiderosis). When the signal in the liver is abnormally low, 3-class segmentation algorithms may account the liver for air and erroneously extend the lung fields inferiorly (Fig. 11). Currently, there is no automated algorithm in place to detect these individual abnormalities. Manual correction can be performed but is cumbersome and requires systematic check of the quality of attenuation map generation.

Pitfalls in PET-MR Compared With PET-CT Besides the described artifacts that represent false image information, we would like to complement our review with some pitfalls in PET-MRI, which have clinical effect but are other than technical errors. They are not related to distorted image information but to missing information. Calcifications Calcifications in an anatomical structure or pathologic lesion can be a helpful diagnostic clue toward a specific diagnosis. In lung nodules, for example, calcifications can be a sign of benign etiology. As MRI is inferior to CT in depicting calcifications, the lack of this key imaging feature can render a diagnosis more difficult. Radiologists need to be aware of this information gap. However, severe calcifications in CT or PET-CT can create substantial deterioration of the image quality, which is not the case in MRI or PET-MR. Non Visualization of Radiopaque Marker, Catheters, and Tubings Most surgical and interventional devices, catheters, tubes, markers, and drains consist of or are labeled with radiopaque material to guarantee their visibility with x-ray-based imaging techniques. This radiopaque markings or metallic

C. Martinez-Rios et al

14

Figure 12 Pitfalls in PET-MR by missing radiopaque markers. Axial and coronal PET-CT (first and fourth columns) and axial and coronal PET-MR (second and third columns) in a patient with tracer injection via the port-a-cath system is demonstrated. A small amount of residual tracer causes a focus of hyperactivity in the mediastinum, which can be misinterpreted as pathology. The radiopaque catheter is visualized on CT (white arrow) but not on MRI. (Color version of figure is available online.)

substance is meant to help identifying the position of invasively placed supportive material and devices within the human body. External monitoring equipment (such as an electrocardiographic box) is another example. Metallic place holders and fiducial markers help to guide therapy targeted to a specific area. In both cases, x-ray-based imaging is used to visualize the respective target. In MRI, this function is not available. Fiducial markers placed for radiotherapy planning are rarely visible with MRI. Indwelling material may be missed, as they may be not well visualized on MR source images. As a consequence, cutaneous focal FDG activity related to these devices and their pathways may be misinterpreted based on lacking knowledge of the indwelling catheter (Fig. 12). This is of particular concern, for example, for patients under sedation, who require a fair amount of monitoring equipment. Also, thick padding and blankets do not cause visible artifacts but do affect the PET SUV in PET-MR. This is not an issue in PET-CT, but it can be significant in PET-MR.

Conclusion Artifacts are well known in hybrid imaging, and most artifacts in PET-MR are related to the fact that attenuation correction is

based on MRI. Limitations of MRI due to the nature of MR physics risk translating into the process of attenuation correction and cause image distortion and errors in segmentation and quantification. Many of these can be accounted for with specific compensation mechanisms, which are in part already technically implemented. However, specifically the issue of metal artifacts needs to be further addressed, and technical solutions for automated or semiautomated correction are desired. In general, however, it is essential to be aware of these artifacts and to develop an understanding of the technical background and underlying algorithms to understand the effect on clinical imaging and to avoid misinterpretation. In clinical routine, it is recommended to always refer to and analyze the quality of the MR attenuation map, which is generated from the MRAC sequence to verify correct segmentation. It is also valuable to view the uncorrected PET images alongside the attenuation-corrected PET images for better identification of artifacts.

References 1. Torigian DA, Zaidi H, Kwee TC, et al: PET/MR imaging: Technical aspects and potential clinical applications. Radiology 267:26-44, 2013

Artifacts and diagnostic pitfalls in PET-MRI 2. Quick HH, von Gall C, Zeilinger M, et al: Integrated whole-body PET/MR hybrid imaging: Clinical experience. Invest Radiol 48:280-289, 2013 3. Vargas MI, Becker M, Garibotto V, et al: Approaches for the optimization of MR protocols in clinical hybrid PET/MRI studies. MAGMA 26:57-69, 2012 4. Schwenzer NF, Stegger L, Bisdas S, et al: Simultaneous PET/MR imaging in a human brain PET/MR system in 50 patients—Current state of image quality. Eur J Radiol 81(11):3472-3478, 2012 5. Drzezga A, Souvatzoglou M, Eiber M, et al: First clinical experience with integrated whole-body PET/MR: Comparison to PET/CT in patients with oncologic diagnoses. J Nucl Med 53:845-855, 2012 6. Schlemmer HP, Pichler BJ, Krieg R, et al: An integrated MR/PET system: Prospective applications. Abdom Imaging 34:668-674, 2009 7. Herrmann KA, Gaeta MC, et al: PET/MRI: Applications in clinical imaging. Curr Radiol Rep 1:161-176, 2013 8. 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 43:860-875, 2013 9. Rakheja R, DeMello L, Chandarana H, et al: Comparison of the accuracy of PET/CT and PET/MRI spatial registration of multiple metastatic lesions. Am J Roentgenol 201:1120-1123, 2013 10. Rakheja R, Chandarana H, DeMello L, et al: Correlation between standardized uptake value and apparent diffusion coefficient of neoplastic lesions evaluated with whole-body simultaneous hybrid PET/MRI. Am J Roentgenol 201:1115-1119, 2013 11. Martinez-Moller A, Eiber M, Nekolla SG, et al: Workflow and scan protocol considerations for integrated whole-body PET/MRI in oncology. J Nucl Med 53:1415-1426, 2012 12. Schlemmer HP, Pichler BJ, Schmand M, et al: Simultaneous MR/PET imaging of the human brain: Feasibility study. Radiology 248:1028-1035, 2008 13. Purz S, Sabri O, Viehweger A, et al: Potential pediatric applications of PET/ MR. J Nucl Med 55(suppl 2):32S-39S, 2014 14. Herzog H, Van Den Hoff J: Combined PET/MR systems: An overview and comparison of currently available options. Q J Nucl Med Mol Imaging 56:247-267, 2012 15. Brendle CB, Schmidt H, Fleischer S, et al: Simultaneously acquired MR/ PET images compared with sequential MR/PET and PET/CT: Alignment quality. Radiology 268:190-199, 2013 16. Judenhofer MS, Wehrl HF, Newport DF, et al: Simultaneous PET-MRI: A new approach for functional and morphological imaging. Nat Med 14:459-465, 2008 17. Zaidi H, Mawlawi O, Orton CG: Point/counterpoint. Simultaneous PET/ MR will replace PET/CT as the molecular multimodality imaging platform of choice. Med Phys 34:1525-1528, 2007 18. Kalemis A, Delattre BM, Heinzer S: Sequential whole-body PET/MR scanner: Concept, clinical use, and optimisation after two years in the clinic. The manufacturer’s perspective. MAGMA 26:5-23, 2013 19. Mollet P, Keereman V, Bini J, et al: Improvement of attenuation correction in time-of-flight PET/MR imaging with a positron-emitting source. J Nucl Med 55:329-336, 2014 20. Kartmann R, Paulus DH, Braun H, et al: Integrated PET/MR imaging: Automatic attenuation correction of flexible RF coils. Med Phys 40:082301, 2013 21. Delso G, Furst S, Jakoby B, et al: Performance measurements of the Siemens mMR integrated whole-body PET/MR scanner. J Nucl Med 52:1914-1922, 2011 22. Visvikis D, Monnier F, Bert J, et al: PET/MR attenuation correction: Where have we come from and where are we going? Eur J Nucl Med Mol Imaging 41(6):1172-1175, 2014 23. Quick HH: Integrated PET/MR. J Magn Reson Imaging 39:243-258, 2014 24. Paulus DH, Tellmann L, Quick HH: Towards improved hardware component attenuation correction in PET/MR hybrid imaging. Phys Med Biol 58:8021-8040, 2013 25. Bezrukov I, Mantlik F, Schmidt H, et al: MR-based PET attenuation correction for PET/MR imaging. Semin Nucl Med 43:45-59, 2013 26. Keereman V, Mollet P, Berker Y, et al: Challenges and current methods for attenuation correction in PET/MR. MAGMA 26:81-98, 2013

15 27. Reiner CS, Stolzmann P, Husmann L, et al: Protocol requirements and diagnostic value of PET/MR imaging for liver metastasis detection. Eur J Nucl Med Mol Imaging 41(4):649-658, 2014 28. Stolzmann P, Veit-Haibach P, Chuck N, et al: Detection rate, location, and size of pulmonary nodules in trimodality PET/CT-MR: Comparison of low-dose CT and Dixon-based MR imaging. Invest Radiol 48:241-246, 2013 29. Quick HH: Integrated PET/MR. J Magn Reson Imaging 39(2):243-258, 2014 30. Schramm G, Langner J, Hofheinz F, et al: Quantitative accuracy of attenuation correction in the Philips Ingenuity TF whole-body PET/MR system: A direct comparison with transmission-based attenuation correction. MAGMA 26:115-126, 2013 31. Visvikis D, Costa DC, Croasdale I, et al: CT-based attenuation correction in the calculation of semi-quantitative indices of [18F]FDG uptake in PET. Eur J Nucl Med Mol Imaging 30:344-353, 2003 32. Sureshbabu W, Mawlawi O: PET/CT imaging artifacts. J Nucl Med Technol 33:156-161 2005; [quiz 163-4] 33. Buchbender C, Hartung-Knemeyer V, Forsting M, et al: Positron emission tomography (PET) attenuation correction artefacts in PET/CT and PET/ MRI. Br J Radiol 86:20120570, 2013 34. Martinez-Moller A, Nekolla SG: Attenuation correction for PET/MR: Problems, novel approaches and practical solutions. Z Med Phys 22 (4):299-310, 2014 35. Kershah S, Partovi S, Traughber BJ, et al: Comparison of standardized uptake values in normal structures between PET/CT and PET/MRI in an oncology patient population. Mol Imaging Biol 15:776-785, 2013 36. Kinahan PE, Hasegawa BH, Beyer T: X-ray-based attenuation correction for positron emission tomography/computed tomography scanners. Semin Nucl Med 33:166-179, 2003 37. Al-Nabhani KZ, Syed R, Michopoulou S, et al: Qualitative and quantitative comparison of PET/CT and PET/MR imaging in clinical practice. J Nucl Med 55:88-94, 2014 38. Izquierdo-Garcia D, Sawiak SJ, Knesaurek K, et al: Comparison of MRbased attenuation correction and CT-based attenuation correction of whole-body PET/MR imaging. Eur J Nucl Med Mol Imaging 41 (8):1574-1584, 2014 39. Schreibmann E, Nye JA, Schuster DM, et al: MR-based attenuation correction for hybrid PET-MR brain imaging systems using deformable image registration. Med Phys 37:2101-2109, 2010 40. 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 49:1875-1883, 2008 41. Hofmann M, Bezrukov I, Mantlik F, et al: MRI-based attenuation correction for whole-body PET/MRI: Quantitative evaluation of segmentation- and atlas-based methods. J Nucl Med 52:1392-1399, 2011 42. Hofmann M, Pichler B, Scholkopf B, et al: Towards quantitative PET/MRI: A review of MR-based attenuation correction techniques. Eur J Nucl Med Mol Imaging 36(suppl 1):S93-S104, 2009 43. Beyer T, Weigert M, Quick HH, et al: MR-based attenuation correction for torso-PET/MR imaging: Pitfalls in mapping MR to CT data. Eur J Nucl Med Mol Imaging 35:1142-1146, 2008 44. Andersen FL, Ladefoged CN, Beyer T, et al: Combined PET/MR imaging in neurology: MR-based attenuation correction implies a strong spatial bias when ignoring bone. Neuroimage 84:206-216, 2014 45. Varoquaux A, Rager O, Poncet A, et al: Detection and quantification of focal uptake in head and neck tumours: (18)F-FDG PET/MR versus PET/ CT. Eur J Nucl Med Mol Imaging 41:462-475, 2014 46. Garibotto V, Heinzer S, Vulliemoz S, et al: Clinical applications of hybrid PET/MRI in neuroimaging. Clin Nucl Med 38:e13-e18, 2013 47. Larsson A, Johansson A, Axelsson J, et al: Evaluation of an attenuation correction method for PET/MR imaging of the head based on substitute CT images. MAGMA 26(1):127-136, 2013 48. Schulz V, Torres-Espallardo I, Renisch S, et al: Automatic, three-segment, MR-based attenuation correction for whole-body PET/MR data. Eur J Nucl Med Mol Imaging 38:138-152, 2011 49. Dixon WT: Simple proton spectroscopic imaging. Radiology 153:189-194, 1984

16 50. 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 53:796-804, 2012 51. Boellaard R, Hofman MB, Hoekstra OS, et al: Accurate PET/MR quantification using time of flight MLAA image reconstruction. Mol Imaging Biol 16:469-477, 2014 52. Eiber M, Souvatzoglou M, Pickhard A, et al: Simulation of a MR-PET protocol for staging of head-and-neck cancer including Dixon MR for attenuation correction. Eur J Radiol 81:2658-2665, 2012 53. Bezrukov I, Schmidt H, Mantlik F, et al: MR-based attenuation correction methods for improved PET quantification in lesions within bone and susceptibility artifact regions. J Nucl Med 54:1768-1774, 2013 54. Varoquaux A, Rager O, Poncet A, et al: Detection and quantification of focal uptake in head and neck tumours: F-FDG PET/MR versus PET/CT. Eur J Nucl Med Mol Imaging 41(3):462-475, 2014 55. Kim JH, Lee JS, Song IC, et al: Comparison of segmentation-based attenuation correction methods for PET/MRI: Evaluation of bone and liver standardized uptake value with oncologic PET/CT data. J Nucl Med 53:1878-1882, 2012 56. Keereman V, Mollet P, Berker Y, et al: Challenges and current methods for attenuation correction in PET/MR. MAGMA 26(1):81-96, 2013 57. Keereman V, Fierens Y, Broux T, et al: MRI-based attenuation correction for PET/MRI using ultrashort echo time sequences. J Nucl Med 51:812-818, 2010 58. Keereman V, Fierens Y, Broux T, et al: MRI-based attenuation correction for PET-MRI using ultrashort echo time sequences. J Nucl Med 51 (5):812-818, 2010 59. Kohan AA, Kolthammer JA, Vercher-Conejero JL, et al: N staging of lung cancer patients with PET/MRI using a three-segment model attenuation correction algorithm: Initial experience. Eur Radiol 23:3161-3169, 2013 60. Blumhagen JO, Braun H, Ladebeck R, et al: Field of view extension and truncation correction for MR-based human attenuation correction in simultaneous MR/PET imaging. Med Phys 41:022303, 2014 61. Kalemis A, Delattre BM, Heinzer S: Sequential whole-body PET/MR scanner: Concept, clinical use, and optimisation after two years in the clinic. The manufacturer’s perspective. MAGMA 26:5-23, 2013 62. Schramm G, Langner J, Hofheinz F, et al: Influence and compensation of truncation artifacts in MR-based attenuation correction in PET/MR. IEEE Trans Med Imaging 32:2056-2063, 2013 63. Delso G, Martinez-Moller A, Bundschuh RA, et al: The effect of limited MR field of view in MR/PET attenuation correction. Med Phys 37:2804-2812, 2010 64. Beyer T, Bockisch A, Kuhl H, et al: Whole-body 18F-FDG PET/CT in the presence of truncation artifacts. J Nucl Med 47:91-99, 2006 65. Keller SH, Hansen AE, Holm S, et al: Image distortions in clinical PET/MR imaging. In: Ignasi Carrio PR (ed): PET/MRI. Methodology and Clinical Applications. Berlin Heidelberg, Springer-Verlag, 21-42, 2014 66. Keller SH, Holm S, Hansen AE, et al: Image artifacts from MR-based attenuation correction in clinical, whole-body PET/MRI. MAGMA 26:173-181, 2013

C. Martinez-Rios et al 67. Ohnesorge B, Flohr T, Schwarz K, et al: Efficient correction for CT image artifacts caused by objects extending outside the scan field of view. Med Phys 27:39-46, 2000 68. Nuyts J BG, Kehren F, Fenchel M, et al: Completion of a truncated attenuation image from the attenuated PET emission data. IEEE Trans Med Imaging 32(2):237-246, 2013 69. Lavdas E, Mavroidis P, Hatzigeorgiou V, et al: Elimination of motion and pulsation artifacts using BLADE sequences in knee MR imaging. Magn Reson Imaging 30:1099-1110, 2012 70. Chun SY, Reese TG, Ouyang J, et al: MRI-based nonrigid motion correction in simultaneous PET/MRI. J Nucl Med 53:1284-1291, 2012 71. Grimm R, Furst S, Dregely I, et al: Self-gated radial MRI for respiratory motion compensation on hybrid PET/MR systems. Med Image Comput Comput Assist Interv 16:17-24, 2013 72. Samarin A, Burger C, Wollenweber SD, et al: PET/MR imaging of bone lesions—Implications for PET quantification from imperfect attenuation correction. Eur J Nucl Med Mol Imaging 39:1154-1160, 2012 73. Lingzhi Hu CS, Eggers Holger, Hu Zhiqiang, et al: Whole-body UTEmDixon: A potential one-scan solution for PET/MR attenuation correction and localization. Presented at the Joint Annual Meeting ISMRM-ESMRMB 2014. Milan, Italy, 2014 74. Hargreaves BA, Worters PW, Pauly KB, et al: Metal-induced artifacts in MRI. Am J Roentgenol 197:547-555, 2011 75. Kamel EM, Burger C, Buck A, et al: Impact of metallic dental implants on CT-based attenuation correction in a combined PET/CT scanner. Eur Radiol 13:724-728, 2003 76. Goerres GW, Hany TF, Kamel E, et al: Head and neck imaging with PET and PET/CT: Artefacts from dental metallic implants. Eur J Nucl Med Mol Imaging 29:367-370, 2002 77. Hamill JJ, Brunken RC, Bybel B, et al: A knowledge-based method for reducing attenuation artefacts caused by cardiac appliances in myocardial PET/CT. Phys Med Biol 51:2901-2918, 2006 78. Partovi S et al: Qualitative and Quantitative performance of 18F-FDGPET/MR and 18F-FDG-PET/CT in patients with head and neck cancers. AJNR 2014 79. FDG-PET/CT Technical Committee: FDG-PET/CT as an imaging biomarker measuring response to cancer therapy, Quantitative Imaging Biomarkers Alliance, Version 1.05, Publicly Reviewed Version. QIBA. Available at: RSNA.ORG/QIBA. RSNA, 2013 80. Ladefoged CN, Andersen FL, Keller SH, et al: PET/MR imaging of the pelvis in the presence of endoprostheses: Reducing image artifacts and increasing accuracy through inpainting. Eur J Nucl Med Mol Imaging 40:594-601, 2013 81. Schramm G, Maus J, Hofheinz F, et al: Evaluation and automatic correction of metal-implant-induced artifacts in MR-based attenuation correction in whole-body PET/MR imaging. Phys Med Biol 59:2713-2726, 2014