Background

The need for attenuation correction in positron emission tomography (PET) has been understood from the inception of the development of PET imaging. Attenuation correction for the γ-rays emitted from the positron annihilation is needed due to the high detection of scattered γ-rays and the loss of γ-rays due to attenuation from tissue and other material along the detection line of response (LOR). In PET imaging, 30–50% of the detected γ-rays in the main photopeak in the detection energy window consist of scattered photons, and the magnitude of attenuation correction is more pronounced in PET imaging (typical by a factor of ~20 compared to ~6 for [99 m-Tc] single-photon emission imaging) due to the need to image two-coincidental γ-ray photons traveling two separate antiparallel distances through the patient at a time over an extended imaging geometry [1,2,3]. Uncorrected attenuation losses from PET lead to increased image noise, distortions in the image (such as from areas of high uptake or filling, e.g., urinary bladder), image artifacts, misrepresentation of uptake intensity (as seen along the skin surface due to a lack of attenuation effects compared to deeper in the body), and loss of quantitative accuracy. To account for the attenuation losses in the PET image, a means to develop an attenuation map, or μ-map, is necessary. The μ-map, once calculated, represents a spatial distribution of linear attenuation coefficients of the patient imaged and is applied to the raw PET image data at the time of PET reconstruction. The μ-map serves to provide a mapping of attenuating material/tissue (based on their linear attenuation coefficients of that material/tissue) in the body and to compute an estimate of the scatter probability of the photons. To this end, various different approaches have been researched to provide attenuation correction for the PET image. Transmission imaging has largely been the most popular approach since the earliest days of emission tomography development [1].

In PET, transmission imaging is primarily tasked with correcting the degrading effects of annihilation photon attenuation from emission imaging, but it also provides anatomical co-localization of the emission and transmission data [2, 3]. From its inception, PET imaging focused largely on imaging the brain [4]. By the mid 1970s, with the introduction of the first ring geometry PET scanner, imaging with 2-deoxy-2-[18-F] fluoro-d-glucose ([18-F]-FDG) [5], and the introduction of a mechanized patient table that could insert the patient deeper into the PET scanner, PET imaging began to be developed for the whole body and not just the brain. Previously, attenuation correction for brain-only PET studies was largely a mathematical correction based on edge detection of the cranium [6]. The skull thickness and density were fairly consistent with patient size and only introduced small bias in the reconstructed PET image due to this variability. However, as PET studies began to cover the whole body, a need for more robust attenuation correction became essential due to larger-scale variability in patient body habitus [7].

Early in the evolution of PET, a [68Ge/68-Ga] rod or a [137-Cs] point source was used to acquire transmission scans through the patient to provide attenuation correction for the emitted positron radiation within the body [1,2,3, 8, 9]. The [68-Ge/68-Ga] source was the most commonly used source for acquisition of transmission data for attenuation correction [10]. For attenuation correction, initially, the [68-Ge/68-Ga] radionuclide was employed as a ring-shaped, rod-shaped, or point source-shaped transmission source, where the point source and rod were rotated near the surface of the scanner bore around the patient [11]. Three transmission scan methods using a radionuclide attenuation correction device were used over the years (Table 1). The first entailed performing the transmission scan before the patient was injected with a radiotracer. This method eliminated the interference between the activity from the [68-Ge/68-Ga] transmission source and injected radiotracer, but was time consuming, because the patient had to remain on the scanner table during the radiotracer uptake (in the case of [18-F]-FDG, 45–60 min) and often suffered from transmission and emission co-registration errors from the patient moving during the uptake phase. The second and third methods entailed scanning the patient after the radiotracer was injected. The post-uptake transmission scan would either be acquired simultaneously during the emission scan, or after the emission scan was concluded. To overcome detector confusion from detecting the emission and transmission scan photons simultaneously [11], the scanner would track the location of the rod source position to know which detector block(s) were being exposed by the transmission rod. Two major limitations of these methods were (1) the count-rate performance of the detector and (2) length of time it took to acquire the transmission scan. First, the detector had to be able to handle the emission and transmission activities simultaneously. Second, due to the count-rate limitation of the PET detectors, transmission scanning with [68-Ge/68-Ga] rods took 2–10 min per bed position [1,2,3, 12].

Table 1 Summary of transmission scan methods using a positron radionuclide (e.g., [68-Ge/68-Ga]) single-photon radionuclide (e.g., [137-Cs]) and CT for attenuation-corrected PET images
Full size table

Many of the transmission imaging limitations introduced by the [68-Ge/68-Ga] rod were overcome with the use of single-photon γ-ray point source, such as [137-Cs]. The [137-Cs] source was shielded on the side closest to the detector ring and rotated around the patient. Because the PET detectors on the opposite side of the gantry were the only ones accumulating transmission signal, and since there were not any random events to account for when counting single-photon events, the [137-Cs] could be a stronger transmission source, increasing the count rate by five to tenfold, thus decreasing noise in the transmission data and decreasing the time required to perform the transmission scan [9, 13]. Still, the use of [137-Cs] came with its limitations, namely: a 5–15% bias was introduced in the attenuation coefficient calculation due to the need to scale from a [137-Cs] photon energy of 662 keV to the positron annihilation energy 511 keV [13]. In addition, a minor increase in the fraction of cross-plane scatter photons came from the higher flux density from the [137-Cs] source, and since no co-linearity gate could be applied to the single γ-ray photon to reject the additional scatter.

In the early 1990s, low-cost PET configurations filled only a small subsection of the gantry with detector modules, and the PET gantry was rotated to produce the full PET image. Within the space between the PET detector modules, it was envisioned to place computed tomography (CT) scan equipment and use CT to acquire the transmission scan [14]. By 2000, a refined working model of a fully diagnostic hybrid PET/CT scanner was introduced [15]. By 2006, PET only scanners using radionuclide transmission sources were largely unobtainable for purchase, and were rapidly replaced by PET/CT. Though CT-based attenuation correction did not offer any improvement in the actual attenuation correction of the radionuclide image, as was shown by O’Conner et al. [16], CT did offer the advantage of high X-ray flux scanning obtained on dedicated X-ray detectors. Using CT for transmission scanning solved both the count-rate limitation from using PET detector modules, and the long transmission scan times of the radionuclide transmission scanning devices, since CT scanning is faster (~10 to 20 s total scan coverage) compared to the several minute per bed position radionuclide transmission scanning. In addition, since PET imaging has limited spatial resolution and limited visualization of anatomical landmarks [17], when combining PET and CT into one imaging modality, the CT component provides fast [10, 18] co-registration for anatomical localization in addition to attenuation correction for PET imaging [8].

Though PET/CT imaging has synergistic advantages over either component individually, and is the standard for PET imaging today, the use of CT has introduced several common artifacts, namely: (1) internal motion (major contributing factors: cardiac, respiratory, and bowel) caused misalignment between PET and CT; (2) truncation; (3) metallic implants; and (4) contrast medium [19]. Motion artifacts are the most prevalent artifact in PET/CT. (1) In general, the CT portion of the PET/CT exam is acquired with a form of breath hold (inspiration or expiration), thus obtained quickly without respiratory motion, whereas the PET portion of the PET/CT exam is acquired with free breathing, resulting in emission image anatomy averaged over many respiratory and cardiac cycles. The acquisition scheme mismatch often places CT anatomy, close to the lungs and heart, at different locations in the image than the PET imaged anatomy, thus leading to under or over attenuation correction of the radionuclide uptake (Fig. 1a). The most common example of this artifact is the lack of perceived radionuclide uptake in the shape of a curvilinear (or downward facing “banana”) at the dome of the liver. (2) Truncation artifacts arise from the limited maximum field of view (FOV) of the CT scanner. If any portion of the patient anatomy resides outside the CT transaxial scan FOV (SFOV) during the CT scan, then the ability for the CT exam to provide attenuation correction for that anatomy is limited, and thus, the PET image reconstruction may be compromised (Fig. 1b). Some scanner manufacturers provide software corrections to mathematically account for truncation artifacts from attenuation losses in anatomy directly outside of the CT transaxial SFOV for PET attenuation correction. Such corrections have shown to improve attenuation reconstruction in PET to better than ~5% as compared to the uncorrected SUV calculation of PET imaging without corresponding CT image outside the CT transaxial SFOV [20]. (3) Metallic implants, such as hip prosthesis, create CT artifacts in the form of streaking, which is due to a lack of signal due to photon starvation from lack of penetration of the metallic device(s). The streaking artifacts lead to high CT numbers around the device, which in turn may produce an overestimation of PET uptake. The lack of signal near a metallic device would have the opposite effect leading to an underestimation of uptake near the device (Fig. 1c). (4) Certain types of CT exams may require the injection of intravenous (IV) or ingestion of oral contrast agents. The purpose of contrast material is to provide visual contrast in the image, and help delineate anatomical structures that would otherwise have similar CT numbers and, therefore, similar grayscale appearance in the reconstructed image. Contrast material by nature is highly absorbent of photons and results in high CT numbers in the reconstructed image. When unaccounted for, these high CT numbers lead to erroneous μ-map correction factors and over estimation of radionuclide uptake; however, it should be noted that the reconstruction software is used to mathematically account for high CT numbers from IV and oral contrast, and thus, contrast has been shown to be safe to use without adversely affecting PET attenuation correction or standard uptake value (SUV) quantitation [21,22,23,24].

Fig. 1
figure 1

a Internal motion caused artifacts generally manifest as a misalignment of CT anatomy and PET uptake in the heart (left panel) and at the dome of the liver (right panel). b Truncation artifacts occur when a portion of the patient’s anatomy resides outside the CT transaxial SFOV (left panel–right arm clipped by CT SFOV causing sever CT artifacts). The ability for the CT to provide attenuation correction for that anatomy is limited (central panel—time of flight (TOF) attenuation correction [AC] and right panel—non-AC [NAC]). c Metallic implants, such as left femur prosthesis, create CT artifacts in the form of streaking. The streaking artifacts lead to high CT numbers around the device (left panel—yellow arrow), which in turn may produce an overestimation of PET uptake (central panel—TOF AC yellow arrow compared to right panel—NAC yellow arrow). The lack of signal near a metallic device (left panel–red arrow) leads to an underestimation of uptake near the device (central panel—TOF AC red arrow compared to right panel—NAC red arrow)

Full size image

Finally, with the combination of PET/CT imaging, so too is the radiation dose to the patient from both emission and transmission imaging components combined. CT X-ray flux density is on the order of 10 times [4] higher than isotope imaging sources, such as [68-Ge/68-Ga] rods [15]. A study by Wu et al. calculated an effective dose for a typical adult, and demonstrated that [68-Ge/68-Ga] transmission rods contributed an additional 0.2–0.3 mSv to the radionuclide PET study dose [10]. As a comparison, effective dose calculated for CT attenuation ranged from 7 to 43 mSv as reported in the past literature [10, 25,26,27,28,29,30]; the wide ranging effective dose levels varied largely on acquisition parameter selection, i.e., from dose reduced to fully diagnostic in quality, and demonstrates that no consensus has been achieved on how to acquire CT for attenuation correction. Various efforts to reduce CT radiation dose have been investigated, such as implementing automatic tube current modulation [31, 32], optimizing scan acquisition parameters (e.g., tube current, rotation time, and tube potential [18, 25, 33,34,35,36]), noise suppression methods to allow the reduction of radiation output (i.e., tube current and tube potential) parameters while maintaining a defined reconstructed image quality (e.g., sinogram smoothing [37] or using iterative reconstruction [26, 38, 39]), and adding additional filtration [33]. Ultimately, CT acquisition parameters should be optimized based on the intended use of the CT for attenuation correction purposes.

Introduction

The type of radiation dose optimization used for CT-based attenuation correction will largely depend on how the CT is utilized in conjunction with the PET image. Three broad categories have been suggested for the purpose of CT-based attenuation correction on the ImageWisely website [40]: (1) diagnosis; (2) anatomic localization of PET images; and (3) attenuation correction only for PET images; each approach to optimize the CT scan will provide the necessary attenuation correction information for full fidelity PET image formation. The three broad categories described on the ImageWisely website provide an opportunity for CT radiation dose adjustment depending on the purpose of the PET/CT study.

As was just described in Section I, the radionuclide transmission scan of the PET study was used for attenuation purposes only and did not provide a sufficient image quality to produce an image for more than gross anatomical localization [41] as can be seen in (Fig. 2a). In similar fashion, the acquisition of a CT exam for simple attenuation correction (CTAC) purposes of a PET/CTAC study should produce the lowest level of radiation dose possible for patient dose optimization. However, if anatomical co-localization of the PET and CT images along with attenuation correction is desired, an escalation in CT radiation dose, beyond CTAC level, would be required. In this scenario, the CT exam used for attenuation correction is not considered diagnostic but will provide the needed attenuation correction and CT image quality for anatomical localization (CTAL) for the PET/CTAL study [42]. Finally, when a diagnostic CT exam is ordered in conjunction with a PET/CT study, it may be possible, for patient dose optimization, to incorporate or replace the attenuation correction CT of the PET/CT study with the diagnostic CT exam and remove the redundant CT scan that would have been the CTAL or CTAC. Thus, the CT component of the study will be used for both attenuation correction and diagnostic purposes (CTD) and would be acquired using technique factors appropriate to produce a diagnostic CT exam in conjunction with the PET/CTD study [30, 42].

Fig. 2
figure 2

a Typical image quality of a radionuclide transmission scan of the chest and abdomen for attenuation correction purposes only. The images provided were acquired on a Siemens ECAT EXACT system using a [68-Ge/68-Ga] rotating rod source with approximately 2 min per bed position transmission scan time. b CTAL exam modified to be acquired using low-dose (0.8 mSv) acquisition technique factors (100 kVp, 10 mAs, and 1.0 mm in-plane pixel resolution) with the 100% iterative reconstruction software (ASiR; GE Healthcare). The images provided were acquired on a GE Discovery 690 system. c Typical diagnostic neck-chest-abdomen-pelvis CT exam using dose-reduced (2.4 mSv) diagnostic acquisition parameters (100 kVp, 38 mAs, and 0.6 mm in-plane pixel resolution) with IV and oral contrast administered. Images from b and c were acquired on the same day from the same patient, where image c was acquired first. A residual amount of contrast can be seen in the right kidney of image b due to excretion of recently administered IV contrast from image c

Full size image

Strategies for CT attenuation correction optimization

CTAC for PET attenuation correction only

It has been demonstrated that when the CTAC is used for attenuation correction only, such as how the radionuclide sources were used in PET only devices that preceded the development of PET/CT, the radiation dose is lower, by approximately an order of magnitude, than typical low-dose CTAL exams [33, 37]. The primary limitation for implementation of the very low-dose CTAC only exams is the bias introduced into the PET reconstruction due to very noisy CT images. The noisy CT images are due to photon starvation at the detector due to low CT output from low tube current time product (mAs) and tube potential (kV) acquisition parameters and the presence of electronic noise in the CT electronic data acquisition systems (DAS). CT detectors are built to efficiently operate in the presence of high X-ray flux [43] due to the fact that the CT DAS does not energy discriminate and, therefore, was designed to integrate the inherently high flux density, polyenergetic spectrum of X-rays [43]. Low CT output leading to under correction of CT-based attenuation is primarily a limitation in adult PET/CT imaging for CT radiation dose reduction; moreover, for pediatric CT imaging, it was shown, in one study [35], that radiation levels could be reduced by 100 fold and only introduce a noise bias in the PET reconstruction of 2%. The traditional radionuclide transmission attenuation correction was not bias-limited, because the PET detector DAS was designed for temporal and energy discrimination of detected photons and was thus better equipped to reject bias inducing scatter, single, and random events [43,44,45]; in addition, the radionuclides were of relatively lower photon count rate (even with the higher count rate of the single-photon γ-ray [137-Cs] source) compared to CT and the high penetrability of the [137-Cs] 611 keV and the [68-Ge/68-Ga] 511 keV γ-ray sources [33, 43]. To overcome the bias introduced into the PET reconstruction process from high-noise contaminated CTAC exams, sinogram smoothing [43], sparse sampling CT reconstruction with iterative reconstruction (such as model-based iterative reconstruction) [37], and added beam filtration [33] have been investigated and demonstrate good noise control for PET reconstruction bias on the order of only 5%.

With proposed noise limitation corrections for CT, Xia et al. [33] reported low-dose CT to vary between 0.14 and 0.42 mGy. Similarly, Fahey et al. [35] reported CTAC dose levels for pediatric exams between 0.3 and 0.42 mGy. Assuming a traditional eyes to thighs PET study, it is reasonable to assume comparable effective dose values, i.e., 0.14–0.42 mSv based on an effective dose weighting factor (w T) approximately equal to one. These low-dose CTAC estimated patient dose values are on the order of [68-Ge/68-Ga] transmission rod contributed radiation dose of 0.2–0.3 mSv [10].

Practical scenarios for application of PET/CTAC rely on the need for PET image data only for clinical diagnosis and no additional information provided from a visual interpretation of the CT. An example of this approach may be for the evaluation of the uptake and biodistribution of relatively long lived tracers, such as [89-Zr]-antibodies, that may require multiple imaging sessions over several days to determine the kinetics and optimal imaging times. A diagnostic PET/CTD or anatomical localization PET/CTAL could be obtained, followed by serial PET/CTAC for attenuation only in subsequent sessions to minimize dose exposure from the CT component of the examination. Another example would be for application of respiratory gated PET/CT. Respiratory gating supplies information about internal organ/tumor motion during a patient’s breathing cycle. When tracking the respiratory motion of the patient using CT, the CT acquires a series of images corresponding to the PET axial position covering the diaphragm, over a period of time as to capture the full breathing cycle of the patient [46]. The CT scanner is either prospectively triggered to acquire an image at a particular phase of the respiratory cycle using a tracking device attached to the patient’s body (either a passive system such as an infrared tracking marker placed on the skin surface [46, 47] or a pressure sensitive device placed around the patient’s chest [47, 48]), or by acquiring a series of images covering the full respiratory cycle and retrospectively sorting the images based on the respiratory cycle trace measured using a tracking device. In one study [49], adult anthropomorphic phantom organ dose levels were measured based on typical CT acquisition parameters for respiratory gating; a mean effective dose of 33.1 mSv was calculated based on absolute organ dose values measured in the lungs, thyroids, breasts, and colon. If low-dose CTAC parameters were adopted as proposed by Fahey et al. [35] or Xia et al. [33] (i.e., 0.14–0.42 mGy) and compared to Matsuzaki et al. (i.e., 33.1 mGy), a potential radiation dose savings of 98% (0.42/33.1; mSv) to 99% (0.14/33.1; mSv) may potentially be possible for CT respiratory gating.

CTAL for anatomic localization of CT and PET images

A more commonly found scenario for clinical prescription of a PET/CT examination is the acquisition of PET/CTAL with a sufficient CT image quality to provide anatomical localization between the CT and PET images, (Fig. 2b). In this common scenario, the radiation dose and other acquisition technique factors from the CT are elevated or enhanced to improve the quality of the image, but not to a level that would render the CT examination of a quality such to replace a diagnostic CT. Two items of consideration affecting the level of radiation output used for attenuation correction were used to define CT attenuation correction for anatomical localization or CTAL: first, the use of low tube current time product and tube potential technique parameters that do not bias the PET reconstruction and lead to accurate PET SUV quantification and second, a sufficient radiation output to limit quantum noise superimposition on the CT image [50].

First, studies have demonstrated CTAC radiation dose reductions by approximately 90%, from the standard CTD baseline dose levels, without introducing negative bias into pediatric PET reconstruction [25, 26, 35]. Substantial dose reduction can be applied for pediatric sized patients. For adult sized patients, the primary limitation to CTAL dose reduction was under correction of the CTAL-based attenuation correction from low kVp and low mAs scan techniques [18, 25, 26, 50]; in these studies, the CTAL acquisition technique was limited to 100–120 kVp and 35–65 mAs and produced dose reductions to 35–62% from a CTD dose.

The second investigation was that of the CTAL image quality and suitability for use of anatomical co-localization of the CT and PET image. At the reduced CTAL dose levels (~90% for pediatric sized patients and 35–62% for medium-to-large sized adults), the appearance of the CT image is severely degraded and limits the ability for co-registered anatomical localization of [18-F]-FDG uptake in organs [50]. To preserve CT image quality, multiple different noise suppression and image quality preservation methods have been proposed. Xia et al. demonstrated with added filtration (0.5 mm Cu) and CT sinogram smoothing; at low-dose radiation output levels, CT and PET image quality is preserved [33, 43]. The use of iterative reconstruction for CT has been well documented to lower traditional CT radiation dose levels and maintains or improves diagnostic CT image quality [51,52,53,54,55,56]. Brady and Shulkin [26], Yang et al. [57], and Palmer and Fahey [58] investigated the use of adaptive statistical iterative reconstruction (ASiR; GE Healthcare), Rui et al. investigated the use of Model-based Iterative Reconstruction (VEO; GE, Healthcare) [37], and Tsa et al. investigated the use of Sinogram Affirmed Iterative Recon (SAFIRE, Siemens Medical) [59] to minimize noise infiltration in low-dose CTAL examinations for co-registered anatomical localization. Each study demonstrated image noise mitigation for the dose-reduced CTAL, accurate CT number reconstruction (i.e., no CT number bias leading to under correction of CT-based attenuation correction for PET reconstruction), and accurate PET SUV calculation. The use of iterative reconstruction has allowed the implementation of dose-reduced CTAL acquisition while preserving appropriate CT image quality for anatomic co-localization between CT and PET images.

CTD for CT attenuation correction, anatomical localization, and diagnostic analysis

Modern PET/CT systems combine into one system fully diagnostic CT scanning systems with the traditional full, multi-ring PET units, thus enabling the CT portion of the PET/CT exam to be acquired and interpreted in the same manner as any other diagnostic CT scan [60]. The benefits of implementing diagnostic quality CTD scanning for attenuation correction are that most PET/CT studies are ordered in tandem with a diagnostic CT scan as part of a clinical analysis for response to cancer therapy treatment, and the substitution of a diagnostic quality CT scan for the CTD portion of the PET/CTD provides potential radiation dose savings for the patient, potentially frees up clinical resources on a separate standalone diagnostic CT scanner that might otherwise be occupied performing a diagnostic CT on the PET patient, whose diagnostic CT could be provided as the PET/CTD, and provides high-quality CT images for anatomical co-localization with the reconstructed PET image. In order for the CTD to be considered diagnostic by definition, the scan must have the appropriate spatial resolution and image signal-to-noise ratio (SNR) [61] (Fig. 2c). To produce diagnostically acceptable spatial resolution for the clinical task at hand, beam collimation and the diameter of the area being imaged, or display FOV, should be appropriately set to provide correct pixel dimensions and reconstruction image magnification, where display FOV is established to accurately sample in-plane transaxial resolution, and beam collimation is established to accurately sample out-of-plane resolution along the main axis along the patient’s body length. Image SNR is directly affected by establishing appropriate dose levels to allow diagnostic information to be obtained in low contrast regions that may be negatively impacted by image noise (i.e., quantum mottle) superimposed over the anatomical information [62]. Intravenous and oral contrast agents are often administered to improve image tissue differentiation in diagnostic CT exams (if no contraindications are present). The use of IV contrast in PET/CT may lead to enhanced lesion conspicuity in the CTD, where normal radiotracer, such as [18-F]-FDG, uptake, and accumulation, is known to be limited, and for oral contrast, to distend the bowel for increased conspicuity of luminal, mural, and extraluminal bowel disease and help distinguish physiologic from pathologic gastrointestinal uptake [21].

The disadvantage of combining PET and CTD is that the acquisition parameters for high-quality PET and CT exams have competing requirements that will require compromises to both CT and PET diagnostic image quality [42, 60]. Five competing acquisition parameters for PET/CTD are, first, the CT display FOV is a fixed acquisition parameter and cannot be decreased. The CT display FOV cannot be reduced for greater reconstruction magnification of smaller objects/patients required for diagnostic quality CT (e.g., the mapping of a smaller portion of the display FOV, such as the neck region, to the full 512 × 512 CT image matrix).

Second, image acquisition differences between diagnostic CT and PET led to discrepancies in contrast bolus injection timing schemes for multiphase contrast imaging. For diagnostic quality, CT requiring contrast enhancement, specific arterial, or venous contrast enhancement phases may require scan delays and scanning different scan series over different lengths of the patient’s body in noncontiguous scanning schemes [21, 63]. Whereas for PET imaging, a single pass CT is generally acquired over the majority of the patient’s body in one continuous motion (i.e., from the base of the skull through mid-femur) [63, 64]. If a single scan acquisition is acquired, contrast enhancement can appear muted or at best present as a mix of arterial and venous phase in the final reconstructed CT images rendering them non-diagnostic. If multiple scan series are prescribed in a traditional multiphase protocol, then misalignment between the CTD and PET data may occur due to the temporal mismatch of the CT and PET images from respiratory and cardiac motion, potential physical patient motion between the CT and PET scans, and physical misalignment due to potential repositioning of the patient arms [63]. Another limitation of the use of contrast agents is the formation of CT artifacts due to photonic starvation of the CT detectors. This limitation presents largely as a visual artifact when interpreting the CT image, and has been shown to not statistically effect the attenuation correction of the PET image, since strong smoothing filters are applied to the CT image that minimizes CT artifacts prior to PET reconstruction [67]. In addition, the patient may experience physiological effects from IV and/or oral contrast, such as nausea and vomiting that may lead to patient noncompliance and exam cancellation. In addition, most CT contrast media for IV contrast are hyperosmolar that may accelerate bladder filling during the PET acquisition, leading to patient discomfort, motion artifacts, and suboptimal evaluation of the pelvis. Minimizing the potentially limiting physiological effects from using contrast in a CTD would be to perform the CT component of the PET/CTD acquisition after the PET emission scan. However, this approach would cause delays in image reconstruction, as scanners generally begin processing images from a completed PET axial position, while images from the next axial position are being acquired. This would effectively delay the image reconstruction until both the PET and CT acquisitions are complete, and not be ideal for busy clinics.

Third, diagnostic CT is generally acquired with a breath hold (either inspiration-for chest imaging or expiration-for abdominal imaging), whereas the traditional CTAC or CTAL imaging is acquired during free breathing or normal expiration breath hold and PET imaging is acquired during free breathing [19, 30, 42, 68]. As previously discussed, the difference between a full inspiration CTD and a free breathing PET acquisition may create spatial mis-registrations between the CT and PET images and potential attenuation correction inaccuracies [19]. Techniques to reduce respiration motion-induced artifacts are well established, but none of the established techniques are effective at mitigating the difference in spatial positioning of a full inspiration versus free breathing [68]. Most require a form of respiratory gating for both the CTD and PET exams. Introducing dynamic respiratory gating (or “4DCT”) increases the imaging time for PET and the radiation dose from CT to the patient by ~33 mSv [49].

Fourth, diagnostic CT scan length is different than the traditional scan lengths needed for regular eyes to thighs or whole body PET/CTD scan lengths. As a dose-efficiency standard, diagnostic dose levels are used to scan the appropriate region of the body needed for diagnosis (e.g., lung apices to mid liver for chest-only). In an ideal setting when acquiring a PET/CTD, where the diagnostic CT order is for a chest-only (as an example), the CTD would comprise a standard or low-dose CTAC or CTAL starting at the beginning of the scan volume down to the point of the upper chest-only scan boundary (e.g., lung apices), where the scanner would then increase the radiation output to an appropriate level to acquire a diagnostic quality CT through the chest-only scan region followed by the remainder of the scan performed at standard or low-dose CTAC/CTAL. Most modern PET/CT scanners do not support a continuous motion dose-differential acquisition scheme as provided in the previous example. For a continuous motion CT, a typical PET/CT scanner would require diagnostic quality radiation levels over the whole of the CTD scan volume (i.e., eyes to thighs), thus, unnecessarily exposing sensitive organ systems to higher than needed radiation dose for attenuation correction purposes. For different CT series dose acquisition schemes, most modern PET/CT scanners can be programed to acquire separate CT scan series with different technique factors, as envisioned in the previous example; however, these scanners require inter-series scan delays that may introduce spatial misalignments, due to patient motion, between the various CT series and/or the whole CT series (spliced together) with the PET scan. Typical inter-series scan delays are due to the scanner needing time to adjust technique factors (i.e., tube potential, rotation speed, etc.), and reposition the patient table (in the case of helical acquisition which needs to acquire extra scan volume before and after the programed CT series start and stop locations, i.e., “over-ranging”) [69].

Finally, in most cases, patient arm positioning for diagnostic CT of the body is conducted with arms above the patient’s head to remove beam hardening artifacts that would manifest as streaks through the scanned volume. However, in PET imaging, the patient’s arms are generally placed along the patient’s side; thus, any streak artifacts produced, because the arms in the transaxial SFOV may compromise CT imaging of the thorax and upper abdomen. Similarly, if arms were placed above the head, the potential streak artifacts would impact imaging of the head and neck regions [19, 67, 70]. Options as to where the patient arms can be placed for the duration of the PET/CT exam (which can last upwards of 15–45 min) may be limited due to patient compliance (i.e., comfort) [67, 70, 71] and medical device access (e.g., in the case, where general anesthesia is administered, arm placement is dependent on accessing the arm for drug administration, etc.). In addition, when the arms are placed either above, or below the patient’s head, if the arms are not fully included in the transaxial SFOV, truncation artifacts will occur and have been shown to degrade both the CT and PET image quality [20]. It has been recommended that patient arms should be placed in a comfortable location for the duration of the exam, and when the arms are to be included in the PET diagnosis, they should be placed fully within the transaxial SFOV, but in a position as not to affect the CT or PET diagnostic image quality [70]. When the patient arms need to be placed at their side, the use of different arm pillows and sponges to elevate the arms at an approximate 45° may allow the patient to maintain the arms at their side for comfort/compliance while minimizing the beam hardening streak artifacts in the throrax and abdomen regions of the images without affecting reconstructed image quality.

When considering the compromises required for the acquisition of PET/CTD, customization of specific protocols has been suggested that would require a separate diagnostic CT and PET with low-dose CTAC or CTAL for the initial disease staging, and then allow for PET/CTD exams for follow-up [60]. In addition, when imaging for a particular disease, such as lymphoma [72] or melanoma [60], it has been suggested that PET/noncontrast-CTAL removes the need for additional diagnostic contrast-enhanced CTD.

Current and future trends in CT dose optimization

The current and future methods for reducing patient dose from CT have been largely technology driven. Siemens has introduced technology to only scan the required volume prescribed by the supervising physician or technologist by introducing continuous motion PET/CT. The current CT scanners are required to acquire CT imaging over the full volume of the PET axial position (which may run between 16 and 18 cm along the patient’s z-axis). Recent technology has been implemented to allow for variable, continuous table motion (FlowMotion, Siemens Medical) instead of a traditional step and shoot acquisition. Continuous table motion allows more flexible prescription of the volume eliminating potential over scanning of patient anatomy outside of the prescribed scan range. The ability to vary the table speed allows for optimization of image quality over areas of the body requiring higher resolution (i.e., slower table motion for better image quality) and to go faster over areas not requiring the same level of image quality [73]. In addition, the continuous motion technology has been shown to improve the clinical workflow efficiency of respiratory gating [73].

In another technology, GE has introduced a very low-dose CT, on the order of CTAC at 0.3–0.62 mSv, but with image quality sufficient for anatomical localization (Q.AC, GE Healthcare) [74]. The higher image quality low-dose CTAC is made possible through a variety of factors: more sensitive CT detectors (able to produce greater light scintillation per X-ray photon, which leads to less signal noise for a better image reconstruction), lower dose acquisition parameters (X-ray generator that can operate at lower technique factors), and new iterative reconstruction techniques. The new Q.AC allows for low-dose attenuation correction and respiratory gating. Similarly, Toshiba has introduced iterative reconstruction (AIDR 3D), for enhanced CT noise control in reconstruction—allowing for low-dose acquisition techniques, and a wide detector, 32 slice, CT for greater dose efficiency.

Other approaches for CT dose mitigation seek to use emission data only, such as from time of flight (TOF), to provide attenuation correction [75], and the use of nonionizing techniques to acquire transmission attenuation correction of PET data in a PET/MRI [76,77,78,79,80,81]. With the introduction of PET/MRI, attenuation correction for PET reconstruction could not be handled using the traditional methods. Since transmission and CT-based techniques are incompatible with PET/MRI geometries and physical constraints (such as being immersed within a strong magnetic field), the use of MRI acquisition techniques, such as the Dixon-VIBE method, direct segmentation with or without atlases, and templates [82], are routinely used; such methods provide to the patient inherent radiation dose sparing from what would have been a CT attenuation correction acquisition. In addition, with the adoption of PET/MRI, it has been demonstrated that overall PET injected dose levels may be reduced substantially due, in part, to the longer dwell time required for each PET axial position to acquire the MRI sequence(s) prescribed with the PET scan [83].

Due to the unique challenges that come from providing attenuation correction data for PET imaging within an MRI environment, research into alternative methods for PET attenuation correction has been investigated. Recently, Defrise et al. [84] demonstrated that TOF PET data contain within the emission sinogram all values necessary to estimate the attenuation and the activity information to reconstruct a PET image. From this study [84], an analytical method demonstrated a unique solution (except for a constant that can be estimated efficiently) to the mathematical construct that establishes, from the available TOF information, the estimation of attenuation and activity. The authors suggest that the TOF analytical method, combined with Poissonian noise modeling and iterative algorithms, may provide a future framework for direct, simultaneous reconstruction of emission and attenuation data.

Summary

Advances in CT technology have provided opportunities to acquire CT for attenuation correction in a variety of ways, allowing for the optimization of the CT for specific clinical purposes, such as PET/CTAC, PET/CTAL, or PET/CTD. Table 2 provides a summary list of acquisition parameters and typical delivered radiation doses for these three optimization approaches as found in the scientific literature.

Table 2 Following is a compilation of published CT exam technique factors and their typical radiation dose levels
Full size table

Some institutions obtain PET/CTD only, while others obtain PET/CTAL for localization only, followed or preceded by diagnostic quality CT restricted to a particular section of the body. With modern dose reduction techniques for CTAL, and some technological advances allowing CTAC exams to have a sufficient image quality for anatomical localization, the additional radiation exposure from a non-diagnostic CT attenuation correction exam is low. Thus, the major justifications for combining diagnostic quality CT as CTD (understanding the challenges described previously of acquiring a truly diagnostic CT during a CTD) would be to provide excellent anatomic image correlation capabilities for PET images, in the cases, where subtle functional or lesion uptake requires greater localization refinement, and to reduce patient dose by removing the CTAC or CTAL exam in conjunction with a separate diagnostic CT in lieu of a single CTD exam.

ImageWisely has recommended three broad categories for the dose optimization of CT for attenuation correction purposes in PET/CT imaging: (1) CTD for fully diagnostic CT quality; (2) CTAL for anatomic localization of PET images; and (3) CTAC for PET attenuation correction only. All three categories were discussed in this review along with an analysis and discussion of the estimated patient dose levels for each category.