ÜBERSICHTSARBEIT
Attenuation correction for PET/MR: Problems, novel
approaches and practical solutions
Axel Martinez-Möller ∗ , Stephan G. Nekolla
Nuklearmedizinische Klinik und Poliklinik der Technischen Universität München, Munich, Germany
Received 6 February 2012; accepted 8 August 2012
Abstract
Attenuation correction in PET is the primary prereq-
uisite for quantification of the radiotracer’s signal.
Absolute quantification is the key to improve diagnostic
performance, to enable comparisons between follow-up
examinations and to perform pharmacokinetic modeling.
A large fraction of the 511 keV annihilation photons from
the positron emitters are scattered by the patient’s body.
Thus, they are discarded or do not even reach the PET
detectors, while others are identified at the wrong loca-
tion after being scattered. To account for these effects and
thus generate quantitative PET images showing the actual
activity distribution, it is necessary to determine an atten-
uation map with the appropriate attenuation coefficients
for 511 keV photons at each voxel. In hybrid PET/CT sys-
tems, this is achieved using the information about the tissue
electron density provided by the CT and adjusting it for the
difference in photon energy. In PET/MR systems, there is
no mechanism to directly measure the attenuation coef-
ficients of the tissue. Determining the attenuation map in
PET/MR is an important challenge involving two problems:
the determination of the patient’s attenuation map and
the determination of the attenuation introduced by addi-
tional hardware components. We describe the approaches
investigated to deal with these problems and, based on
the experience with a fully integrated PET/MR system, we
∗ Corresponding author: Axel Martinez-Möller, Nuklearmedizinische Klinik der Technischen Universität München, Ismaninger Str. 22, D-81675 München.
Tel.: +49-89-41402968; Fax: +49-89-41404950.
E-mail: a.martinez-moller@lrz.tu-muenchen.de (A. Martinez-Möller).
Z. Med. Phys. 22 (2012) 299–310
http://dx.doi.org/10.1016/j.zemedi.2012.08.003
http://journals.elsevier.de/zemedi
Die Schwächungskorrektur bei hybriden
PET/MR-Systemen: Probleme, neue Ansätze
und praktische Lösungen
Zusammenfassung
Eine grundlegende Voraussetzung für die quantitative PET-
Bildgebung liegt in der erfolgreichen Schwächungskor-
rektur. Die Möglichkeit dieser absoluten Quantifizierung
der Aufnahme von Radiopharmazeutika ist einer der
primären Unterschiede zur der konventionellen nuk-
learmedizinischen Bildgebung. Erst damit ergibt sich
z.B. die Vergleichbarkeit serieller Untersuchungen und
die Möglichkeit kinetischer Modellierungen physiologi-
scher Prozesse. Die Schwächungskorrektur ist deswegen
von fundamentaler Bedeutung, da ein großer Anteil der
511 keV-Vernichtungsquanten aus dem Positronenzerfall
im Körper des Patienten gestreut wird. Somit erreicht nur
ein Teil der Quanten die Detektoren oder wird durch die
Streuablenkung falsch zugeordnet. Diese Prozesse kön-
nen korrigiert werden, sofern die räumliche Verteilung
des schwächenden Gewebes und dessen Schwächungsko-
effizienten bekannt sind. Im Fall eines anderen hybriden
Bildgebungsverfahrens – der PET/CT – misst man diese
Verteilung mit Hilfe des integrierten Computertomo-
graphen, wobei in guter Näherung nur Korrekturen für die
300 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310

finally discuss potential solutions and limitations in a close
to routine setting.
unterschiedlichen Energien notwendig sind. In PET/MR-
Systemen stellt sich nun eine neue Herausforderung, da der
Kernspintomographie die Bestimmung der Röntgendichte
des Gewebes nicht zugänglich ist. Allerdings ist das nur
ein Aspekt, wenngleich ein sehr komplexer. Zusätzlich wer-
den die 511-keV-Quanten aber auch durch eine erhebliche
Menge messtechnisch notwendiger Komponenten des MR-
Gerätes geschwächt. Dieser Übersichtsartikel präsentiert
auf der einen Seite publizierte Lösungsvorschläge für diese
Probleme. Darüber hinaus werden aber auch – basierend
auf den ersten Erfahrungen mit einem vollständig inte-
grierten Ganzkörper-PET/MR-System – deren praktische
Umsetzung und Limitationen im routinenahen Betrieb
diskutiert.

Keywords: PET/MR, Attenuation correction Schlüsselwörter: PET/MR, Schwächungskorrektur

Introduction When a 511 keV photon is scattered by the patient’s body,

it will not reach the detector which it would have reached if
Hybrid PET/MR scanners have recently been developed it would not have undergone the interaction (Fig. 1). This is
and many advantages compared to PET/CT and/or standalone known as photon attenuation, and reflects the fact that a line of
MR imaging are foreseen in several imaging scenarios [1–3]. response (LOR, defined as the line between two detectors) will
The manufacturers have proposed PET/MR scanners with two miss some of the photon pairs which were emitted along this
different architectures: sequential, such as the Philips Inge- line because of this interaction taking place for any of both
nuity TF PET/MR [4], and integrated, such as the Siemens photons. As an example, over 90% of the photon pairs for
Biograph mMR [5]. Both architectures share a common LORs going through the center of the abdomen are typically
aspect: the absence of a transmission system (rotating radionu- attenuated and missed by the detectors in their original LOR.
clide source as in standalone PET or CT as in PET/CT) in order
to estimate the attenuation map at 511 keV within the field of
view.
However, PET imaging is severely affected by photon atten-
uation and scatter, and corrections are needed in order to
have quantitative images reflecting the true spatial distribu-
tion of the radiotracer. A prerequisite to these corrections is the
availability of an attenuation map containing the attenuation
coefficients for 511 keV photons at each voxel. The validity
of the PET images depends strongly on the accuracy of the
attenuation map. The determination of the attenuation map is
thus a major challenge for PET/MR [6,7] because MRI does
deliver signals which are in no way related to the attenuation
properties of tissue.
Here we first describe the physical concepts underlying
attenuation and scatter as well as the correction mecha-
nisms usually implemented in PET/CT scanners. Moreover,
we describe the different approaches proposed until now for
PET/MR and their implementation on commercially available
scanners.

Attenuation and scattering of photons

The dominating interaction of PET annihilation photons
with human biological tissue is through Compton scattering, Figure 1. Diagram of a PET detector ring with two annihilation
whereas photoelectric absorption plays only a minor role. events demonstrating the effects of scatter (A) and attenuation (B).
 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310
Compton scattering introduces an additional problem: once
a scattered photon reaches a PET detector, the energy resolu-
tion of the detector might not be good enough to determine
that the energy of the incoming photon was below 511 keV.
If the other photon resulting from the same annihilation is
also detected and accepted (with or without being scattered),
a coincidence event will be incorrectly assigned to the LOR
between both detectors; that is, we will observe photon pairs
along a LOR where they were not generated, an effect some-
what opposite to that of attenuation (Fig. 1). This effect is
often referred as scatter, and the ratio of accepted coinci-
dences which have undergone scatter to the total number of
accepted coincidences as scatter fraction. The scatter fraction
in whole-body 3D imaging (without septa as found in most
modern PET systems) is typically in the range of 30% - 50%.
Attenuation correction
Attenuation correction (AC) is used to account for the back-
to-back photons emitted along an LOR and not detected (at
least not at this LOR) because they undergo Compton scatter-
ing.
As PET is based on the detection of coincident photons,
the attenuation for events occurring within an LOR does not
depend on their location within the LOR, but only on the
total attenuation for this LOR. Accurate correction for photon
attenuation in PET can thus be performed by multiplying the
measured coincidences in each LOR by a factor specific to
each LOR, the so-called attenuation correction factor (ACF),
which for a non-uniform attenuating medium can then be
described as:
1 
ACFLOR =  = e LOR μ(x)dx
−μ(x)dx
e LOR
The linear attenuation coefficient μ for 511 keV photons
varies for biological tissues between below 0.03 cm-1 for the
lung and well above 0.12 cm-1 for cortical bone.
Standalone PET, PET/CT and PET/MR systems use differ-
ent approaches to determine the ACFs. In stand-alone PET,
now nearly obsolete, the ACFs are typically measured with a
rotating radionuclide source (68 Ga/68 Ge or 137 Cs) by deter-
mining the ratio between the number of photons detected with
the imaged object in the scanner and the number of photons
detected in a blank scan.
In PET/CT scanners, the CT data are used instead to
determine the attenuation map containing the attenuation
coefficient μ at each voxel, and the ACFs are computed by
integrating over each LOR. Because CT uses a polyenergetic
beam at lower energies (approx. 60-140 keV for x-rays versus
511 keV in PET) the attenuation coefficients are adjusted by
means of a bilinear transformation [8]. CT-based AC offers
many advantages as compared to transmission measurements
performed with rotating radionuclide sources, including faster
acquisition and reduced noise. However, it also presents new
challenges such as misregistration of the PET and CT data
301
caused by different respiratory states in each acquisition [9,10]
and artifacts introduced by the presence of metal [11] or iodi-
nated contrast agents [12,13]. Furthermore, one must be aware
that CT-based AC results in a higher radiation dose compared
to PET-based transmission scans.
In PET/MR systems, new approaches to provide the atten-
uation map have been developed and will be described later
in this review.
Scatter correction
Scatter correction is used to estimate the number of scat-
tered coincidences which have been accepted by the detectors
and subtract them. Different scatter correction techniques have
been investigated in the past, e.g. [14–21]. Currently most
whole-body scanners implement a single scatter simulation
(SSS) [22–24], which estimates the scatter contribution based
on the emission data and the attenuation map by simulating
exclusively the coincidences where only one of the photons
has been scattered only once.
The SSS algorithm has been shown to be well suited for
whole-body imaging, but its accuracy is compromised by the
scatter arising from annihilations outside the axial field of
view (FoV) as well as by multiple scatter. Watson et al. [25]
further refined the method to match the simulated data to the
measured data by including scale factors derived from the
parts of the sinogram which correspond to the exterior of the
patient, segmented based on the data available in the ACF
sinogram. In this step, it is important that the attenuation map
(and consequently the ACF sinogram) is not transaxially trun-
cated. Otherwise, LORs assigned to the exterior of the patient
(because the attenuation map mistakenly shows air) might
actually pass through true activity causing the scale factors to
be too high; this effect is also sometimes observed in the case
of motion between the emission data and the attenuation map
[26].
Determination of the patient attenuation map
We subdivide the determination of the attenuation map in
PET/MR in two problems: the determination of the patient
attenuation and the determination of the hardware attenuation
produced by other elements present in the FoV (patient bed,
MR coils and occasionally other components such as ECG).
Determination of the patient attenuation map in PET/MR is
the most challenging problem. Three fundamental approaches
have been proposed until now, which we will further refer to as
segmentation-based, template-based and emission-based (see
overview in Table 1).
Segmentation-based approaches. The idea of segment-
ing the attenuation map in different tissue classes was already
proposed by Huang et al. in 1981 [27] in order to reduce
noise propagation from the measured transmission images.
The method was improved with time and widely used on
standalone PET scanners.
302 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310

Table 1
Overview of the potential strengths of each MR-based attenuation correction method for whole-body imaging.
Ability to deal with

Speed Robustness Anatomical variations Bone Metal Limited MR FoV Patient-spec. ␮

Segment +++ +++ +++ + - - -

Atlas - + - ++ + + -
Emission + + +++ ++ ++ ++ ++
Note: reference potential values are provided, the performance and accuracy of each method may strongly vary depending on the particular implementation
and potential developments by future research (e.g. MR sequences allowing direct bone segmentation or faster implementations of the atlas and emission
based methods).

Figure 2. Results of an implementation of the segmentation-based approach using a UTE sequence to identify the skull for brain imaging,
reprinted by permission of the Society of Nuclear Medicine from Catana et al. [34]. (A) Attenuation maps obtained from segmentation of
3-tissue classes based on CT (left) and dual-echo ultrashort echo-time (DUTE) (right). (B) PET images reconstructed using CT-based (left)
and DUTE-based (middle) AC methods and image of the relative changes between the 2 methods (right).
 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310 303

Application of a segmentation-based approach for MR-
based AC was first mentioned in the context of neurological
imaging for standalone PET scanners [28–31], obviating
the need for time-consuming PET transmission images by
spatially registering a segmented T1-weighted MR to the
PET emission data and applying the theoretical attenuation
coefficients. Because cortical bone is hard to visualize in con-
ventional MRI sequences such as T1-weighted images [32],
it was then suggested to use technically rather demanding
ultrashort echo time (UTE) MR acquisitions [33–35], which
facilitate the segmentation of the skull in the context of brain
imaging (Fig. 2).
Application of segmented MR-based AC for whole-body
PET/MR imaging was first proposed by our group [36,37],
using water-weighted and fat-weighted images derived from a
Dixon MR sequence [38] in order to classify each voxel as air,
lungs, fat and soft tissue (Fig. 3). Since bone cannot be robustly
segmented in whole-body MR images, the method ignored the
specific contribution of bone. Steinberg et al. [39] and Schulz
et al. [40] proposed segmentation-based approaches using
instead T1-weighted turbo spin echo sequences, also ignoring
the contribution by bone and allowing a faster acquisition at
the expense of not differentiating between fat (␮=0.086 cm-1 )
and soft tissue (␮=0.096 cm-1 ).
It should be noted that whole-body segmentation
approaches represent an approximation of the true attenu-
ation map and the performance depends on the number of
tissue classes considered [42,43]. The use of predefined atten-
uation coefficients is a limitation which does not reflect the
variability found in patients (e.g. different lung attenuation
coefficients for different respiratory states). Furthermore, cur-
rent segmentation-based approaches for whole-body imaging
ignore the contribution of bone, resulting in an underesti-
mation of the uptake for lesions located in the bone and its
surroundings; different values for the underestimation of the
SUV of bone lesions have been reported by different groups:

Figure 3. Results from an implementation of the segmentation-based approach, reprinted by permission of the Society of Nuclear Medicine
from Martinez-Möller et al. [37]. On the top row, CT, CT-AC-PET and CT-AC-PET/CT images. On the bottom row, MR-based attenuation
map (within the dashed lines), MR-AC-PET and MR-AC-PET/CT images. The MR-derived attenuation map is created by assigning each
voxel of the MR data into air, lung, fat or soft tissue.
304 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310

based on PET/CT data it has been estimated to be 5-15%
[37,40] or up to 23% [41], whereas simulations with the XCAT
phantom have led to a local bias of up to 17% [42] or even 30%
[43]. In addition, the effect of MR contrast agents – while tech-
nically not changing the water/fat separation – might influence
the image segmentation as they result in an increase of MR
signal due to reduced T1 values.
Yet, segmentation-based methods are robust, computation-
ally fast and the associated quantification error for bone
lesions is repeatable and well understood. Therefore, com-
mercially available PET/MR scanners are using this approach
for MR-based AC: the Siemens Biograph mMR uses an adap-
tation of [37] while the Philips Ingenuity TF PET/MR uses an
adaptation of [40].
Template-based approaches. A template-based approach
consists in matching a model to the patient’s anatomy, so that
the known attenuation map from the model can be applied to
the patient data. In early PET systems, such an approach was
sometimes used with simple models, first for brain studies
approximating the outline of the head to an ellipse on each
plane and later refined to automatically identify the body con-
tour from the PET emission data [44–46]. For brain studies,
the approach was further improved, proposing spatial registra-
tion of the SPECT data to a template derived from a digitized
phantom [47] or registration of the PET data to a template
obtained by averaging scans of normal subjects [48].
With the advent of the first PET/MR scanners, several
groups proposed and evaluated the use of a template-based

Figure 4. Implementation and results of a template-based approach for brain imaging, reproduced from Schreibmann et al. [51] with
permission from AAPM. An atlas CT is non-rigidly registered to the acquired MR image, providing a synthetic CT matching the patient’s
anatomy.
 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310 305

approach for PET/MR brain imaging [48–52] (Fig. 4). The
approach was shown to be very successful. However, it
required an accurate non-rigid spatial registration of the
acquired MR images and the template. This is feasible in the
head, but for whole-body imaging non-rigid registration has
been reported to be error-prone [53]. Pure atlas-based meth-
ods fail also in the presence of anatomical abnormalities (e.g.
amputation, lung tumor, etc.). Yet, by combining the template-
based approach with information from the segmented MR and
pattern recognition, the technical feasibility of the method was
also demonstrated for whole-body imaging by Hofmann et al.
[54].
In summary, template-based approaches have shown a great
potential in the context of brain imaging, where the cortical
bone from the skull can thus be robustly taken into account
in the attenuation map. For whole-body imaging, template-
based methods have the potential advantage of taking the bone
into account and partially recovering metal-induced artifacts,
being able to get closer to the real distribution of attenuation
coefficients than segmentation-based methods. However, this
comes at the expense of an increased computational time and
results which depend on the accuracy of a challenging non-
rigid image registration.
Emission-based approaches. The PET emission data can
be used to derive some information about the attenuation
map. In fact, without the influence of scattered radiation,
it would be possible based on the measured PET data to
estimate both the distribution of radiotracer and the atten-
uation map (or at least the part relevant for the image
generation). This approach has been extensively investigated
for standalone PET imaging in order to obviate the need
for a transmission measurement [55–61], but its success is
limited by the crosstalk between emission and attenuation
data.
In PET/MR imaging, robustness is gained by using the
segmented MR data as starting point. Nuyts et al. [62]
described its application in order to recover truncated parts
of the attenuation map (see subsection below) whereas
Salomon et al. [63] reported that, using time-of-flight PET
data and an ideally segmented MR image, it might be
even possible to improve the MR-based attenuation map
and assign proper attenuation values to the bones (Fig. 5).
Defrise et al. [64] have further demonstrated that time-of-
flight PET data can lead to reasonable estimates of the
attenuation map (up to a constant) even without a prior
segmentation.

Figure 5. Results from an implementation of the emission-based technique, reprinted by permission of IEEE from Salomon et al. [63].
From left to right: T1-weighted MR image, segmented label map (each segment gets an attenuation factor), initial attenuation estimate and
estimated attenuation after 12 iterations.
306 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310
Possible artifacts and limitations. The determination of
the patient attenuation map faces two additional challenges:
the limited transaxial field of view in MRI and the occasional
presence of metal.
There is often transaxial truncation in the MR images due
to the limited FoV in MRI, especially for patients acquired
with the arms down (Fig. 6). Increasing the MR bore diameter
would eliminate this limitation but would be extremely costly.
If the MR-based attenuation map is truncated, it can have an
important effect on the attenuation correction, especially for
regions near the arms. Furthermore, implementations of the
scatter correction which sample the outside part of the patient
assuming that all events detected outside are scattered events
[25] might result in a severely biased scatter scaling factor in
the presence of truncation. A simple method to account for
that estimating the patient contour based on the uncorrected
PET images has been reported [65,66]; however, the approach
needs tracer uptake in the truncated regions and tends to create
a convex hull out of the body contour, unable to delineate the
gap between the arms and the chest. To overcome these limi-
tations, a simultaneous emission-transmission reconstruction
using the segmented MR-based attenuation map as starting
point has shown promising results [62,63]. More recently,
a technique to enlarge the MR FoV by optimizing the gra-
dient readout and using distortion correction has also been
investigated [67,68].
Figure 6. Three-dimensional visualization of a stack of transaxial
MR images acquired in overlapping packages. This demonstrates the
impact of the limited FoV in xy-direction which causes truncation in
the arms and occasionally in the hips, breast or abdomen.
Research reports regarding the presence of metal in MR-
based AC are still scarce. Metal artifacts produce a severe
local reduction or even elimination of signal in the MR image,
which might result in underestimation of the attenuation fac-
tor and produce negative artifacts (decreased signal) in the
PET images, as opposed to metal artifacts in PET/CT. In [52],
Hofmann et al. showed that in 2 out of 3 patients with metal
implants a template-based approach could partly alleviate the
problem. Yet, as long as no robust method able to account
for metal artifacts is available in commercial scanners, the
image readers should be aware of this limitation and carefully
analyze the regions surrounding metal implants.
Determination of the hardware attenuation
map
The photon attenuation and scatter within the FoV happens
not only with the patient’s body but also with the hardware
present between the patient and the PET detectors. This is
particularly relevant in PET/MR imaging, where not only the
patient bed is present in the FoV, but typically also coils
to improve the MR signal and occasionally other compo-
nents (positioning aids, patient monitoring devices, cables,
earphones, etc.). These hardware components are invisible in
conventional MR sequences and thus will be absent in the
MR-based attenuation map.
A first important point is whether the attenuation produced
by the components is relevant enough to have an impact on the
data. Indeed, some components such as patient bed and head
coil have a clear impact on the attenuation, while for others
the effect is smaller.
The solution consists in precalculating attenuation models
of the hardware components and adding them at the correct
spatial location within the attenuation map (Fig. 7). Although
the idea is rather basic, its implementation is not. On one hand,
it is challenging to obtain accurate attenuation models because
CT-derived attenuation factors can in general not be easily
translated to 511 keV for non-human tissues; radionuclide
transmission imaging overcomes this problem at the expense
of a limited spatial resolution. On the other hand, the location
should be precisely known (within 1-2 mm) in order to avoid
having local errors above 5% [34,69,70].
The location can be reasonably well-known for rigid and
fixed components such as the patient bed as well as, when
applicable, head and spine coils attached to it. Other flexible
or movable components (e.g. most surface coils, positioning
aids and patient monitoring devices) cannot be currently taken
into account by this method [71–73]. Instead, they should be
redesigned to minimize their impact on the PET data by adapt-
ing the choice of the materials used and its spatial distribution.
In fact, redesigning the components to minimize the attenua-
tion is also convenient for components which can be corrected
with a precalculated attenuation model in order to preserve the
PET image quality. However, this might require a complete
 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310 307

Figure 7. Transmission measurement of an MR head and neck coil acquired using a rotating Germanium source, reproduced from Delso
et al. [70] with permission from IOP Publishing Ltd. (a) Rendering of the head coil. (b) Axial, (c) sagittal and (d) coronal views of the
transmission scan of the head and neck coils.

re-design of currently available and highly sophisticated MR
coils.
Perspectives for attenuation and scatter
correction in PET/MR
The determination of the attenuation map in PET/MR imag-
ing is challenging. The same way PET and MR provide
complementary information about the patient, the investigated
approaches for attenuation correction are not necessarily alter-
natives but also complement each other. As can be seen in
Table 1, the best results can be achieved by combining differ-
ent approaches.
Regarding the hardware attenuation, precalculated attenu-
ation maps for the most attenuating components have to be
included [34,69–72]. This approach is thus complementary to
any approach used for the patient attenuation.
Concerning the photon attenuation produced by the patient,
different approaches appear to be necessary for brain imaging
and whole-body imaging. For brain imaging, identification of
bone is important due to the impact of the skull; segmentation-
based approaches using ultrashort echo time [33,34] and
template-based approaches [48–52] have both proven to be
successful at this task.
For whole-body imaging, a combination of several
approaches seems necessary. The segmentation approach has
been reported to provide good results for separate PET/CT
and MRI acquisitions [37,39,40] as well as for integrated
PET/MR acquisitions in cardiology [74] and oncology [75].
The approach has however three inherent limitations: under-
estimation of tracer uptake for bone lesions; underestimation
in the presence of metal; and errors due to the truncation of
the MR-based attenuation map, especially when the patient is
acquired with the arms down. For the latter issue, the transax-
ial FoV of the MR sequences used for whole-body AC should
be maximized [67,68]; if this is not sufficient, the missing
attenuation data should be recovered with a simultaneous
emission-transmission reconstruction of the region outside
the FoV [62–64], profiting from time-of-flight PET data if
308 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310
available. This can be further combined with a template-based
approach in an attempt to recover the attenuation produced by
the bones and the regions with metal implants [52].
Furthermore, Eiber et al. [76] suggested that the MR
data acquired for attenuation correction purposes in whole-
body imaging might also be useful for the localization and
evaluation of PET positive lesions. Such an approach is rel-
evant in order to minimize acquisition times and allow a
cost-efficient use of PET/MR. In this context, methods deriv-
ing an attenuation map from a clinical MR would also be
advantageous.
A limitation of current MR-based AC approaches for both
brain imaging and whole-body imaging is the variability of
the attenuation factor to assign to each tissue. Whereas soft
and adipose tissue have the same attenuation in all patients,
lung tissue and cortical bone are different for each patient; in
the case of lung tissue, it changes also for different respira-
tory states. None of the current methods can overcome this
limitation at present. It remains to be demonstrated whether
dedicated MR sequences can provide additional informa-
tion to address this issue or if the accuracy of simultaneous
emission-transmission reconstruction can be improved to con-
tribute to assign patient-specific attenuation factors within the
body.
Conclusion
The determination of the attenuation map in PET/MR is
a complex procedure involving a combination of steps to
take into account the attenuation produced by the patient and
by the hardware components. Extensive research has been
done in the last years and, although still with limitations,
currently available PET/MR scanners already offer routine
capable solutions. However, additional trials are needed to fur-
ther improve the accuracy and fully assess the clinical validity
of these complex approaches.
References
[1] Heiss WD. The potential of PET/MR for brain imaging. Eur J Nucl Med
Mol Imaging 2009;36(1):105–12.
[2] Antoch G, Bockisch A. Combined PET/MRI: a new dimension
in whole-body oncology imaging? Eur J Nucl Med Mol Imaging
2009;36(1):113–20.
[3] Nekolla SG, Martinez-Möller A, Saraste A. PET and MRI in cardiac
imaging: from validation studies to integrated applications. Eur J Nucl
Med Mol Imaging 2009;36(1):121–30.
[4] Zaidi H, Ojha N, Morich M, Griesmer J, Hu Z, Maniawski P, et al. Design
and performance evaluation of a whole-body Ingenuity TF PET-MRI
system. Phys Med Biol 2011;56(10):3091–106.
[5] Delso G, Fürst S, Jakoby B, Ladebeck R, Ganter C, Nekolla SG, et al.
Performance measurements of the Siemens mMR integrated whole-
body PET/MR scanner. J Nucl Med 2011;52:1914–22.
[6] Zaidi H, Hasegawa B. Determination of the Attenuation Map in Emis-
sion Tomography. J Nucl Med 2003;44:291–315.
[7] Hofmann M, Pichler B, Schölkopf B, Beyer T. Towards quantitative
PET/MRI: a review of MR-based attenuation correction techniques. Eur
J Nucl Med Mol Imaging 2009;36(Suppl. 1):93–104.
[8] Kinahan PE, Townsend DW, Beyer T, Sashin D. Attenuation correction
for a combined 3D PET/CT scanner. Med Phys 1998;25(10):2046–53.
[9] Beyer T, Antoch G, Blodgett T, Freudenberg LF, Akhurst T, Mueller
S. Dual-modality PET/CT imaging: the effect of respiratory motion
on combined image quality in clinical oncology. Eur J Nucl Med
2003;30:588–96.
[10] Martinez-Möller A, Souvatzoglou M, Navab N, Schwaiger M, Nekolla
SG. Artifacts from misaligned CT in cardiac perfusion PET/CT studies:
frequency, effects, and potential solutions. J Nucl Med 2007;48:188–93.
[11] Goerres GW, Ziegler SI, Burger C, Berthold T, von Schulthess GK, Buck
A. Artifacts at PET and PET/CT Caused by Metallic Hip Prosthetic
Material. Radiology 2003;226:577–84.
[12] Nakamoto Y, Chin BB, Kraitchman DL, Lawler LP, Marshall LT, Wahl
RL. Effects of nonionic intravenous contrast agents at PET/CT imaging:
phantom and canine studies. Radiology 2003;227:817–24.
[13] Yau YY, Chan WS, Tam YM, Vernon P, Wong S, Coel M, et al.
Application of intravenous contrast in PET/CT: does it really introduce
significant attenuation correction error? J Nucl Med 2005;46:283–91.
[14] King PH. Noise identification and removal in positron imaging systems.
IEEE Trans Nucl Sci 1981;28:148–51.
[15] Bergstrom M, Eriksson L, Bohm C, Blomqvist G, Litton J. Correc-
tion for scattered Radiation in a Ring Detector Positron Camera by
Integral Transformation of the Projections. J Comp Assist Tomogr
1983;7(1):42–50.
[16] Grootoonk S, Spinks TJ, Sashin D, Spyrou NM, Jones T. Correction for
scatter in 3D brain PET using a dual energy window method. Phys Med
Biol 1996;41:2757–74.
[17] Jones T, Michel C, Bol A. Correction for Scatter Using a Dual Energy
Window Technique with a Tomograph Operating without Septa. IEEE
Med Imaging Conf 1991:1569–73.
[18] Townsend DW, Geissbuhler A, Defrise M, Hoffman EJ, Spinks TJ,
Bailey DL, et al. Fully three-dimensional reconstruction for a PET
camera with retractable septa. IEEE Trans Med Imaging 1991;10(4):
505–12.
[19] Bendriem B, Trébossen R, Frouin V, Syrota A. A PET scatter correc-
tion using simultaneous acquisitions with low and high lower energy
thresholds. IEEE Med Imaging Conf 1993:1779–83.
[20] Accorsi R, Adam LE, Werner ME, Karp JS. Optimization of a fully
3D single scatter simulation algorithm for 3D PET. Phys Med Biol
2004;49(12):2577–98.
[21] Zaidi H, Koral KF. Scatter modelling and compensation in emission
tomography. Eur J Nucl Med 2004;31(5):761–82.
[22] Ollinger JM. Model-based scatter correction for fully 3D PET. Phys
Med Biol 1996;41:153–76.
[23] Watson CC, Newport D, Casey ME. A single scatter simulation tech-
nique for scatter correction in 3D PET. In: Grangeat P, Amans JL, editors.
Three-Dimensional Image Reconstruction in Radiology and Nuclear
Medicine. Dordrecht: Kluwer Academic Publishers; 1996. p. 255–68.
[24] Watson CC. New, faster, image-based scatter correction for 3D PET.
IEEE Trans Nuc Sci 2000;47(4):1587–94.
[25] Watson CC, Casey ME, Michel C, Bendriem B. Advances in scatter cor-
rection for 3D PET/CT. IEEE Nucl Sci Symp Conf Rec 2004;5:3008–12.
[26] Lodge MA, Mhlanga JC, Cho SY, Wahl RL. Effect of patient arm motion
in whole-body PET/CT. J Nucl Med 2011;52(12):1891–7.
[27] Huang SC, Carson RE, Phelps ME, Hoffmann E, Schelbert HR, Kuhl
DE. A boundary method for attenuation correction in positron emission
tomography. J Nucl Med 1981;22(7):627–37.
[28] Bergström M, Litton J, Eriksson L, Bohm C, Blomqvist G. Determi-
nation of object contour from projections for attenuation correction
in cranial positron emission tomography. J Comput Assist Tomogr
1982;6:365–72.
[29] Le Goff-Rougetet R, Frouin V, Bendriem B. Segmented MR images for
brain attenuation correction in PET. Proc SPIE 1994;2167:725–36.
[30] Zaidi H, Montandon ML, Slosman DO. Magnetic resonance imaging-
guided attenuation and scatter corrections in three-dimensional brain
positron emission tomography. Med Phys 2003;30:937–48.
 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310
[31] Rota Kops E, Qin P, Müller-Veggian M, Herzog H. MRI based attenua-
tion correction for brain PET images. In: Buzug TM, Holz D, Bongartz
J, Kohl-Bareis M, Hartmann U, Weber S, editors. Advances in Medical
Engineering. Berlin Heidelberg: Springer; 2007. p. 93–7.
[32] Wagenknecht G, Rota Kops E, Mantlik F, Fried E, Pilz T, Hautzel H, et al.
Attenuation correction in MR-BrainPET with segmented T1-weighted
MR images of the patient’s head – A comparative study with CT. IEEE
Nucl Sci Symp Conf Rec 2011:2261–6.
[33] Keereman V, Fierens Y, Broux T, De Deene Y, Lonneux M, Van-
denberghe S. MRI-Based Attenuation Correction for PET/MRI Using
Ultrashort Echo Time Sequences. J Nucl Med 2010;51:812–8.
[34] Catana C, van der Kouwe A, Benner T, Vernon P, Wong S, Coel M,
et al. Toward Implementing an MRI-Based PET Attenuation-Correction
Method for Neurologic Studies on the MR-PET Brain Prototype. J Nucl
Med 2010;51:1431–8.
[35] Berker Y, Franke J, Salomon A, Palmowski M, Donker HC, Temur
Y, et al. MRI-Based Attenuation Correction for Hybrid PET/MRI
Systems: A 4-Class Tissue Segmentation Technique Using a Com-
bined Ultrashort-Echo-Time/Dixon MRI Sequence. J Nucl Med
2012;53(5):796–804.
[36] Martínez-Möller A, Souvatzoglou M, Botnar R, Navab N, Ziegler S,
Schwaiger M, et al. An approach for MR-based attenuation correction
for combined MR/PET: effects of ignoring bones (abstract). J Nucl Med
2007;48(Suppl. 2):156P.
[37] Martinez-Möller A, Souvatzoglou M, Delso G. Tissue Classification
as a Potential Approach for Attenuation Correction in Whole-Body
PET/MRI: Evaluation with PET/CT Data. J Nucl Med 2009;50:520–6.
[38] Ma J. Dixon Techniques for Water and Fat Imaging. J Magn Reson
Imaging 2008;28:543–58.
[39] Steinberg J, Jia G, Sammet S, Zhang J, Hall N, Knopp MV. Three-
region MRI-based whole-body attenuation correction for automated
PET reconstruction. Nucl Med Biol 2010;37:227–35.
[40] Schulz V, Torres-Espallardo I, Renisch S, Hu Z, Ojha N, Börnert P, et al.
Automatic, three-segment, MR-based attenuation correction for whole-
body PET/MR data. Eur J Nucl Med Mol Imaging 2011;38:138–52.
[41] Samarin A, Burger C, Wollenweber SD, Crook DW, Burger IA, Schmid
DT, et al. PET/MR imaging of bone lesions – implications for PET
quantification from imperfect attenuation correction. Eur J Nucl Med
Mol Imaging 2012;39(7):1154–60.
[42] Keereman V, Van Holen R, Mollet P, Candenberghe S. The effect of
errors in segmented attenuation maps on PET quantification. Med Phys
2011;38:6010–9.
[43] Akbarzadeh A, Ay MR, Ahmadian A, Riahi Alam N, Zaidi H. Impact of
using different tissue classes on the accuracy of MR-based attenuation
correction in PET-MRI. IEEE Nucl Sci Symp Conf Rec 2011:2524–30.
[44] Michel C, Bol A, DeVolder AG, Goffinet AM. On-line brain attenuation
correction in PET: towards a fully automated data handling in a clinical
environment. Eur J Nucl Med 1989;15:712–8.
[45] Siegel S, Dahlbom M. Implementation and evaluation of a calculated
attenuation correction for PET. IEEE Trans Nucl Sci 1992;39:1117–21.
[46] Weinzapfel BT, Hutchins GD. Automated PET attenuation correction
model for functional brain imaging. J Nucl Med 2001;42:483–91.
[47] Stodilka RZ, Kemp BJ, Prato FS, Kertesz A, Kuhl D, Nicholson RL.
Scatter and AC for brain SPECT using attenuation distributions inferred
from a head atlas. J Nucl Med 2000;41(9):1569–78.
[48] Montandon ML, Zaidi H. Atlas-guided non-uniform attenuation correc-
tion in cerebral 3D PET imaging. NeuroImage 2005;25:278–86.
[49] Rota Kops E, Herzog HR. Template-based attenuation correction of PET
in hybrid MR-PET [abstract]. J Nucl Med 2008;49(Suppl):162P.
[50] Hofmann M, Steinke F, Scheel V, Charpiat G, Farquhar J, Aschoff
P, et al. MRI-based attenuation correction for PET/MRI: a novel
approach combining pattern recognition and atlas registration. J Nucl
Med 2008;49:1875–83.
[51] Schreibmann E, Nye JA, Schuster DM, Martin DR, Votaw J, Fox T. MR-
based attenuation correction for hybrid PET-MR brain imaging systems
using deformable image registration. Med Phys 2010;37:2101–9.
309
[52] Malone IB, Ansorge RE, Williams GB, Nestor PJ, Carpenter TA, Fryer
TD. Attenuation correction methods suitable for brain imaging with a
PET/MRI scanner: a comparison of tissue atlas and template attenuation
map approaches. J Nucl Med 2011;52(7):1142–9.
[53] Beyer T, Weigert M, Quick HH, Pietrzyk U, Vogt F, Palm C, et al.
MR-based attenuation correction for torso-PET/MR imaging: pitfalls in
mapping MR to CT data. Eur J Nucl Med 2008;35:1142–6.
[54] Hofmann M, Bezrukov I, Mantlik F, Aschoff P, Steinke F, Beyer T, et al.
MRI-based attenuation correction for whole-body PET/MRI: quantita-
tive evaluation of segmentation- and atlas-based methods. J Nucl Med
2011;52(9):1392–9.
[55] Censor Y, Gustafson D, Lent A, Tuy H. A new approach to the emission
computerized tomography problem: Simultaneous calculation of atten-
uation and activity coefficients. IEEE Trans Nucl Sci 1979;26:2275–9.
[56] Natterer F, Herzog H. Attenuation Correction in Positron Emission
Tomography. Math Meth Appl Sci 1992;15:321–30.
[57] Welch A. Attenuation correction in PET using consistency information.
IEEE Trans Nucl Sci 1998;45(6):3134–41.
[58] Nuyts J, Dupont P, Stroobants S, Bennick R, Mortelmans L, Suetens
P. Simultaneous maximum a posteriori reconstruction of attenuation
and activity distributions from emission sinograms. IEEE Trans Med
Imaging 1999;18(5):393–403.
[59] Glatting G, Wuchenauer M, Reske SN. Simultaneous iterative recon-
struction for emission and attenuation images in positron emission
tomography. Med Phys 2000;27:2065–71.
[60] Bronnikov AV. Reconstruction of attenuation map using discrete con-
sistency conditions. IEEE Trans Med Imaging 2000;19:451–62.
[61] Panin VY, Kehren F, Hamill J, Michel C. Application of discrete data
consistency conditions for selecting regularization parameters in PET
attenuation map reconstruction. Phys Med Biol 2004;49:2425–36.
[62] Nuyts J, Michel C, Fenchel M, Bal G, Watson C. Completion of a Trun-
cated Attenuation Image from the Attenuated PET Emission Data. IEEE
Nucl Sci Symp Conf Record (NSS/MIC) 2010:2123–7.
[63] Salomon A, Goedicke A, Schweizer B, Aach T, Schulz V. Simultaneous
Reconstruction of Activity and Attenuation for PET/MR. IEEE Trans
Med Imaging 2011;30(3):804–13.
[64] Defrise M, Rezaei A, Nuyts J. Time-of-flight PET data determine
the attenuation sinogram up to a constant. Phys Med Biol 2012;57:
885–99.
[65] Delso G, Martinez-Möller A, Bundschuh R, Nekolla S, Ziegler SI. The
effect of limited MR field of view in MR/PET attenuation correction.
Med Phys 2010;37(6):2804–12.
[66] Tang J, Haagen R, Blaffert T, Renisch S, Blaeser A, Salomon A, et al.
Effect of MR truncation compensation on quantitative PET image recon-
struction for whole-body PET/MR. IEEE Nucl Sci Symp Conf Record
(NSS/MIC) 2011:2506–9.
[67] Blumhagen JO, Ladebeck R, Fenchel M, Kampmeier J, Scheffler K.
MR-based Field-of-View Extension: Compensation of Field Imperfec-
tions. Proc Intl Soc Mag Reson Med 2011;19:2693.
[68] Blumhagen JO, Ladebeck R, Fenchel M, Kampmeier J, Scheffler K.
MR-based Field-of-View Extension: Gradient and B0 Correction Post-
Processing. Proc Intl Soc Mag Reson Med 2011;19:2694.
[69] Zhang B, Pal D, Zhiqiang H, Ojha N, Guo T, Muswick G, et al. Attenua-
tion correction for MR table and coils for a sequential PET/MR system.
IEEE Nucl Sci Symp Conf Record (NSS/MIC) 2009:3303–6.
[70] Delso G, Martinez-Möller A, Bundschuh RA, Ladebeck R, Candidus
Y, Faul D, et al. Evaluation of the attenuation properties of MR equip-
ment for its use in a whole-body PET/MR scanner. Phys Med Biol
2010;55(15):4361–74.
[71] MacDonald LR, Kohlmyer S, Liu C, Lewellen TK, Kinahan PE.
Effects of MR surface coils on PET quantification. Med Phys
2011;38(6):2948–56.
[72] Tellmann L, Quick HH, Bockisch A, Herzog H, Beyer T. The
effect of MR surface coils on PET quantification in whole-body
PET/MR: results from a pseudo-PET/MR phantom study. Med Phys
2011;38(5):2795–805.
310 A. Martinez-Möller, S.G. Nekolla / Z. Med. Phys. 22 (2012) 299–310

[73] Mantlik F, Hofmann M, Werner MK, Sauter A, Kupferschläger J,
Schölkopf B, et al. The effect of patient positioning aids on PET
quantification in PET/MR imaging. Eur J Nucl Med Mol Imaging
2011;38(5):920–9.
[74] Nekolla S, Souvatzoglou M, Schachoff S, Fürst S, Martinez-Möller
A, Ziegler S, et al. MR/PET Attenuation correction using a multi tis-
sue model for quantification of cardiac uptake: Initial comparison to
PET/CT. J Nucl Med 2011;52(Suppl. 1):440.
[75] Drzezga A, Souvatzoglou M, Eiber M, Beer AJ, Fürst S, Martinez-
Möller A, et al. First clinical experience of integrated whole-body
MR/PET. Comparison to PET/CT in patients with oncological diag-
noses. J Nucl Med 2012;53(6):845–55.
[76] Eiber M, Martinez-Möller A, Souvatzoglou M, Holzapfel K, Pickhard
A, Löffelbein D, et al. Value of a Dixon-based MR/PET attenuation
correction sequence for the localization and evaluation of PET-positive
lesions. Eur J Nucl Med Mol Imaging 2011;38(9):1691–701.

Available online at www.sciencedirect.com