Can PET be performed without an attenuation scan?
Received Aug
Can PET be performed without an attenuation scan?
Colin Jones 0 2
Ran Klein 0
0 Reprint requests: Ran Klein, Department of Nuclear Medicine, The Ottawa Hospital , Ottawa, ON , Canada
1 Department of Nuclear Medicine, The Ottawa Hospital , Ottawa, ON , Canada
2 Department of Systems and Computer Engineering, Carleton University , Ottawa, ON , Canada
Positron emission tomography (PET) has emerged as the king of non-invasive molecular imaging, largely due to its ability to quantify tracer concentrations in units of Bq/mL. This ability is heavily reliant on accurate correction of coincidence photon attenuation. In cardiac imaging applications, including myocardial perfusion imaging (MPI), accurate and robust attenuation correction (AC) has been coveted as a means to distinguish between uptake deficits and attenuation artifacts. The notion that PET is able to deliver near perfect AC, along with other advantages, has greatly contributed to the growth of cardiac PET in recent years. From its early days, AC was an integral part of PET. Early instrumentation used weak, rotating transmission sources (e.g., 68Ge/68Ga or 137Cs) to measure attenuation along each line of response. The AC factors were then applied to the measured coincidence data to compensate for attenuation prior to image reconstruction. These transmission scans were lengthy and produced noisy data that propagated into the reconstructed image. With the introduction of hybrid PET-CT, x-ray computed tomography (CT) data could be acquired in a few seconds and manipulated to produce nearly noise-free AC estimates,1 and replaced traditional transmission scans as the de facto AC method in clinical PET. Recently, PET-MR hybrid systems have become available and rely on processing of magnetic resonance images (MR) to derive AC data. MR-based AC is a complex problem requiring further research and development,2 but has been
-
demonstrated to produce equivalent clinical finding
compared to PET-CT.3,4 While the nuclear imaging
community welcomed these hybrid technologies due to
the shorter scan times and the ability to co-register
anatomical and functional information, it quickly became
clear that certain challenges lay ahead with regards to AC.
Both CT and MR approaches suffer from
misalignment problems due to patient motion between
acquisition on attenuation and emission data.
Rigidbody motion (e.g., of the head) can be effectively
corrected using manual or automated registration methods.5
However, compensation for non-rigid patient motion is
much more challenging and remains the topic of much
research. In the context of nuclear cardiology, respiratory
and cardiac motions further compound the challenge.
Both CT and MR data are captured split second, resulting
in mid-breath and mid-cardiac cycle images that do not
fully correspond to PET data that is acquired over many
breathing and cardiac cycles, even when emission data
are gated (respiratory or cardiac). Breath-hold regiments
during CT or MR data acquisition and cardiac triggering
are common strategies to minimize motion artifacts
within the attenuation image. Nevertheless cardiac
imaging always requires verification and often correction
to ensure optimal AC registration. Registration is
performed with regards to the heart, often resulting in
misregistration of other structures in the image (e.g., liver,
lung, skeletal muscle and bone).
Respiratory or cardiac motion related AC
misregistration can be partially accounted for using 4D-CT or
-MR techniques; however, 4D techniques typically assume
regular periodic motion such as uniform breath cycles and
cardiac motion.6 Other, more practical limitations of these
technologies have also been acknowledged. CT-AC can be
a significant source of radiation to the patient compared to
the tracer dose (e.g., *0.5 mSv CT-AC dose vs *1 mSv
from a 82Rb scan7) especially if 4D-CT-AC is used to
account for respiratory motion.16 PET-MR systems are
currently prohibitively expensive, can increase the overall
study length, and have significant counter indications
(e.g., non-MR compatible implanted devices and
claustrophobia).
ESTIMATING ATTENUATION FROM THE
EMISSION DATA
Another approach to correct for attenuation, which
was proposed over three decades ago,8 is to determine
the attenuation information from the emission scan
directly. Nuyts et al. explored this approach in 1999
using two algorithms: maximum likelihood (ML), and
maximum a posteriori (MAP) iterative reconstructions
using decreasing ordered subsets (OS). The results of
both algorithms showed promise in comparison to no
AC or standard AC, with MAP outperforming ML.9
Despite promising results, simultaneous estimation of
attenuation and activity suffers from cross-talk,
propagating errors between both images.10
Time of flight (TOF) information can be used to help
localize updates during iterative reconstructions and reduce
the cross-talk problem. With increasing prevalence of TOF
capabilities in moder (...truncated)