European Nuclear Medicine Guide
Historical developments
Positron emission tomography has been successfully used in cardiac, neurological and oncological research applications since the 1970s, and numerous milestone publications have shown the potential of this technique for the characterization of functional tissue properties covering metabolism and perfusion. Since then, PET has been developed into a clinical imaging tool, but, due to high costs and lack of reimbursement in many countries, PET stayed in the research domain for a long time. However, this changed in 2001, when Townsend and Beyer implemented the first hybrid PET/CT system, which combined high-resolution morphological CT images with informationally valuable, albeit at a lower spatial resolution, metabolic specificity of PET[151]. From a logistical point of view, another advantage has to be appreciated: the time-consuming acquisition of a transmission dataset for attenuation correction using external sources, taking up to 15 minutes per bed position, was replaced with a scan lasting a few seconds using the CT. This enabled higher patient throughput in whole-body, oncological imaging and paved the way for the success of PET/CT in today’s medical imaging world. Although this success is almost entirely based on tumour imaging, given the clinical availability of PET/CT systems, neurological and cardiac applications are widely used (Fig. 1).
Figure 1. Three PET image examples from neuro-oncology, prostate imaging and cardiac viability scanning.
Technical principles
PET imaging uses the rapid radioactive decay of positron-emitting isotopes (such as 18F, 11C, 13N, 68Ga, 82Rb, etc.), which are chemically “inserted” in molecules with high biological relevance (e.g. for assessment of metabolism, perfusion, and receptors). These radiopharmaceuticals (otherwise called tracers) are intravenously injected into a patient either prior to the scan or after positioning in a PET scanner. Then, positrons are emitted and will annihilate with electrons within a very short time under the emission of two 511 keV photons (this process is discussed in more detail earlier in this book). PET tomographs utilize the near-simultaneous (coincidence) detection of these photons with rings of detectors (Fig. 2). The high-energy photons are detected with dedicated materials which convert them into lower-energy photons, which are then, in turn, detected and amplified (mostly) with photomultiplier tubes. These electrical signals are subsequently processed to estimate the so-called line of response (LOR) along which the decay must have taken place. This is possible since the two annihilation photons travel in an anti-parallel manner (i.e., they are emitted at an angle of approximately 180o), implying that the decay must have taken place somewhere along the LOR (Fig. 3). Thus, in contrast to SPECT scanners, where collimators are used to associate a detected event with the direction it came from, PET relies on an “electronic collimation” by using a very short-acceptance timing window of typically a few nanoseconds or less for the detection of both events. Fundamentally, three types of events can occur (Fig. 4). The first kind is a true coincidence, if indeed the two annihilation photons arrive unscattered at the two detectors along the LOR. The second, a scattered coincidence, occurs when one or both photons from a single positron decay undergo a scatter event in the body, but arrive within the time window; obviously, this results in a wrong spatial association. Finally, a random coincidence can occur, where rays from two unrelated decay events are registered within the time window. Whereas the first possibility results in a “correct” measurement, the two other cases will yield image degradation. Originally, PET systems operated in 2-dimensional (2D) mode with interplane septa which reduce scattered photons as coincidence measurements. Basically, in this 2D mode, coincidence measurements are performed only in one plane of the PET camera. Removing the septa and accepting coincidences between scanner planes, however, increases sensitivity and consequentially shortens scan time. This so-called 3D mode has shown many advantages especially in oncological imaging protocols, including increased patient throughput[152,153].
Figure 2. Basic principles of PET imaging: a radioactive tracer is injected into a patient and (B) emits two 511keV photons after the positron annihilates with an electron. Using a ring design, photons are detected and measured with photomultiplier tubes (PMTs) surrounding the patient.
Figure 3. Millions of signals per second coming from all the PMTs are processed in coincidence mode: only those two events that are detected within a few nanoseconds (or less for time-of-flight) are considered to stem from the same annihilation event.
Figure 4. Three possibilities exist for such a measurement: true, scattered and random coincidences. The true events are identified using sophisticated algorithms
Today, the most commonly used detector materials are lutetium oxyorthosilicate (LSO) and gadolinium oxyorthosilicate (GSO). Both of these tracers are attractive due to their physical properties and they are increasingly used instead of bismuth germanate (BGO) in PET/CT systems[154]. Their relatively fast light decay time and high light yield enable the use of short coincidence time windows[155]. Consequently, this improves the count rate capabilities and reduces randoms. With the aim to improve spatial resolution, recent PET/CT systems are equipped with smaller crystals, which is of particular importance as these potentially allow the detection of smaller structures. As the enhancement of the spatial resolution from 7.0 to 4.5 mm leads to an increase of about 30% in count recovery (given an average target size of about 10 mm), the use of high-resolution PET ameliorates the assessment of regional tracer distribution in the targets and allows more accurate quantification of physiologic parameters such as blood flow, metabolism, or receptor density.
From the CT components perspective, the available options range from 16-slice to 128-slice systems depending on the desired application. This corresponds to the variety of protocols found in (typically) oncological clinical scenarios. This spectrum ranges from very high-throughput centres with low dose CT protocols to sites where a full contrast media enhanced diagnostic CT workup is combined with PET. The dominating tracer used worldwide is clearly [18F]FDG.
Attenuation and Scatter Correction
Attenuation correction in PET is the prerequisite for any quantification of the radiotracer uptake signal (Fig. 5). Such an absolute quantification is the key to superior diagnostic performance and enables comparisons between serial examinations and the performance of any pharmacokinetic modelling. A large fraction of the 511 keV annihilation photons from the positron decay are actually scattered by the patient’s body. Consequentially, they are discriminated due to a lower energy or do not reach the PET detectors at all. To account for these effects and thus compute activity-wise correct PET images, it is necessary to determine an attenuation map with the appropriate attenuation coefficients for 511 keV photons at each voxel. In hybrid PET/CT systems, this is achieved by using the information about the tissue electron density, provided by the CT, and adjusting it with regard to the difference in photon energy. However, although the CT scan is very fast and PET scan times are being constantly reduced, misalignment can still occur: mis-registration between emission and “transmission” might lead to uptake errors (Fig. 5).
Figure 5. The most relevant correction in PET is attenuation correction (AC). These examples of a cylindrical phantom and a brain study using profiles along the orange dotted line show that ignoring AC yields inacceptable signal distortions, whereas a misalignment (i.e., shift by 8 mm in this example) between CT and PET data also severely degrades the means of quantification.
In addition to the effects of misalignment, metal implants, or other interventional devices may affect the quantification of the PET tracer uptake. These PET artefacts are primarily due to the reconstruction artefacts of the CT and migrate through the overestimating of attenuating tissue into the PET images. The relevance of these artefacts differs depending on their position and thus, hybrid reading or reviewing of also non-attenuation corrected data is always advisable.
Recent technical developments
Since the introduction of PET/CT at the turn of the millennium, a series of technical improvements have happened. From a more logistical aspect, many developments of the CT systems per se were made available for the “hybrid cousin”. Today, with the exception of rather dedicated, high-resolution cardiac CT systems, a variety of CT components in a PET/CT cover the area from rather cost-effective systems to the high end.
On the PET side, more intrinsic developments happened both on the detector, the acquisition, and the image reconstruction side.
Utilizing the time-of-flight information is actually not a new concept, but was described already in the early days of PET. As outlined before, PET scanners utilize a short temporal coincidence window to decide whether the detected photons stem from the same annihilation event. With improved detector technology, the temporal resolution of PET systems allowed measuring the difference in the arrival time of these photons, which, in turn, allowed the estimation where the event along the LOR happened[156]. Since 2006, all major vendors have implemented this technology in their systems, which now show, especially in obese patients, improved lesion detectability.
In the same extent the PET detector technology improved, the computer processing power did too. This resulted from a complete migration from analytical reconstruction techniques (“filtered back projection”) to iterative approaches where, basically, the particular image was estimated such that its projection data optimally fitted the acquired projection data. As this also allowed the integration of the CT data for attenuation and scatter correction, it paved the way for dramatically improved image quality[157]. This led to a long series of incremental improvements in the PET image reconstruction, where many factors describing the properties of the imaging apparatus are integrated into the reconstruction algorithms (9). This approach is known as point-spread-function modelling or resolution recovery. It is worthwhile noting that all these complex computations have an impact on the quantification and require harmonization, especially in respect of serial examination and multi-centre studies[158,159].
Even though cardiac PET gating has been clinically implemented for decades, the use of respiratory gating in oncological imaging is less widespread. What contributes to this effect is the acquisition duration relative to the object size: in cardiac PET, 10-20 minutes are used to cover the heart in a single bed position, whereas in whole-body oncological PET imaging, with scan times of 1-3 minutes in the thorax, there is simply less data acquired.
However, respiratory motion during the acquisition interval discernibly blurs the signal from the affected structure and thus reduces the spatial resolution which could theoretically be achieved. Technically, any respiratory and/or cardiac gated PET uses an implementation where, parallel to the measurement of all coincidence events, the gating signals are recorded and stored together in a so-called list-mode stream. For cardiac gating, an ECG is used, and for respiratory gating, typically an optical or a pneumatic device is used. The first is positioned on the chest and monitors by a camera and the latter is integrated into an elastic belt, which is then fastened around the lower chest of the patient. Recently, data driven approaches were introduced: those derive respiratory motion from the tracer distribution within the body, e.g. a tumour within the lung. For all these methods of detection, the association of a given annihilation event with the motion state allows for sophisticated approaches in motion compensation, thus, increasing the effective spatial resolution. From a technical perspective, it is worthwhile mentioning that the frequency distribution of human respiration is significantly different to that of cardiac gating as the respiratory frequency is much more irregular and its distribution may vary substantially over the length of a PET acquisition as the patients might go from an anxious into a more relaxed state or even fall asleep.
Another acquisition mode is continuous bed motion: historically, when PET examinations required an extended coverage of the patient (i.e. not only heart or brain), this was realized with several, overlapping bed positions along the body axis (“step and shoot”). Technically, the approach of a continuously moving bed (similar to CT and MRI) enables - together with adaptive table speeds - a more homogenous sensitivity along the scanner axis, but also increases patient comfort as the more or less abrupt change from one to the next bed position is avoided[160].
The final major technical improvement was the replacement of the conventional photomultiplier tubes (PMTs). Initially driven by the PET/MR systems, which simply do not allow for PMTs(which are not compatible with magnetic fields), the new (“digital PET”) devices, based on avalanche photodiodes (APD) or silicon photomultipliers (SiPM), show increased sensitivity and allow for both a compact design and are of a massively increased simultaneous scan range. The latter concept is utilized in so-called “whole-body” or “total-body-systems”, which contain either large parts of the body or even the entire body[161,162]. Despite discussions of cost effectiveness, those systems offer a huge potential for either very low activity/dose scanning (utilizing the increased sensitivity) or even parametric imaging based on dynamic acquisition protocols.
Without any doubt, molecular imaging with PET/CT is an invaluable tool in research and clinical routine. Although a technically ambitious and logistically challenging methodology, it provides an unprecedented level of performance. However, with respect to reliable quantification, the need for harmonization of quantification is still an important element.
An extensive collection of PET/CT artefacts can be found in PET/CT Atlas of quality control and image artefacts. IAEA Human Health Series no. 27, IAEA, Vienna 2014
https://www.iaea.org/publications/10424/pet/ct-atlas-on-quality-control-and-image-artefacts (accessed 2020 05 07)