| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Commentary |
1 INSERM U650, Laboratoire de Traitement de l'Information Medicale (LATIM), CHU Morvan, 5 avenue Foch, 29609 Brest, France, 2 Department of Nuclear Medicine, University Hospital Brest, Brest, France and 3 Northern Ireland Regional Medical Physics Agency, Royal Victoria Hospital, Belfast, UK
Positron emission tomography (PET) is a functional imaging technique developed in order to use specific ligands labelled with positron emitting isotopes for the monitoring of in vivo molecular processes. The use of a wide range of biologically significant elements (such as 13N, 11C, 15O and 18F) provides a strong basis for molecular imaging using PET. On the other hand, the ligands used by PET originate from pharmacological agents that demonstrate specific biochemical interactions. The most widely used PET radioisotope to date is a 18F labelled glucose analogue known as 2-fluoro-2deoxy-D-glucose (FDG). FDG facilitates the monitoring of molecular glucose metabolism since it is phosphorylated and trapped in cells in proportion to the rate of glycolysis.
Although PET began to demonstrate its clinical utility as early as the mid 1980s, it has only recently (over the past 5 years) found widespread acceptance in routine clinical practice. This can be mainly attributed to the high cost associated with running a PET facility, particularly for applications such as functional brain imaging which require the presence of a cyclotron on site. As the focus in PET utilization "shifted" from functional brain imaging to oncology applications, with FDG being the main radiopharmaceutical of use, a "host satellite" model emerged leading to a wider availability of PET. This model is based on the use of a centrally positioned cyclotron in order to supply a number of peripherally arranged PET facilities requiring only an imaging device.
PET technology development over the past 15 years has accordingly undergone a number of significant changes, driven mainly by the requirements imposed by whole body acquisitions, which dominate diagnostic oncology investigations. The combination of PET and FDG has already demonstrated its advantages over conventional imaging modalities in tumour diagnosis and staging for a number of oncology applications [1]. These include particular areas such as lymphoma, melanoma, lung and colorectal cancer where evidence-based studies in a number of European Union countries have indicated the need for a PET scanner for every 1–1.5 million inhabitants [2]. The demonstrated success of PET in such oncology areas has posed certain requirements in patient throughput, considering that 95% of such studies require whole body investigations.
Today, work is concentrating on the use of PET for more specific oncology areas such as response to therapy applications and radiotherapy treatment planning. Although FDG has demonstrated certain success in areas such as these, the advent of new 18F radiopharmaceuticals, developed to probe particular molecular targets, are expected to be in the forefront of these new imaging applications. The requirements for the continuous success of PET in existing oncology applications, as well as its expansion in new specific oncology areas such as those mentioned above, are the development of more specific imaging molecules in combination with improvements in PET hardware and software technology. The goals for current PET technology developments have been an improvement in patient throughput coupled with increased diagnostic accuracy. The recognition of PET as the diagnostic tool of choice in other major cancers as well as the development of new oncology application areas will continue to impose requirements for higher patient throughput without an associated increase in cost. This requirement, coupled with the need for improved image quality and quantitation for the accurate and detailed characterization of new radiopharmaceuticals, should dominate any future advancement in the PET technology arena. As such, these advancements will most certainly be concentrating on higher sensitivity devices. The present commentary concentrates on the current status and future developments in PET technology (excluding the area of radiopharmaceuticals) and how these developments may further influence the role of PET in clinical practice, particularly in the area of oncology.
From the beginning, certain main principles of PET imaging, including performance assessment parameters, have to be defined in order to allow a relationship to be drawn between PET technology developments and their eventual impact in clinical practice. PET is based on the detection of two 511 keV photons originating from the annihilation of a positron (emitted by a nucleus of the injected radiopharmaceutical) with an atomic electron. Following the coincidence detection of the two annihilation quanta, a line may be subsequently drawn through the point where the annihilation took place between the two detectors. With a large number of these events it is possible to reconstruct an image of the in vivo radioisotope distribution. Such a coincidence event is known as a true coincidence event (Figure 1a
).
|
). A high probability also exists that either one or both of the annihilation quanta will be scattered inside the body prior to their detection, consequently resulting misplacement of the annihilation site. This type of event is known as a scattered coincidence (Figure 1b). The detection of both scatter and random coincidences results in an increase in background noise, which in turn significantly affects the quantitation accuracy of the reconstructed images. The objective of any advancement in PET detection technology lies in the first instance in minimizing the detection of such erroneous events through hardware developments, while in terms of software developments to reduce their number in the reconstructed images.
One of the first parameters that may significantly affect the performance of a PET system is its mode of operation. In principle, there are two such modes of operation defined as two-dimensional (2D) and 3D PET, characterized by the presence (2D) and absence (3D) of septa in the imaging field of view (Figure 2). The use of septa leads to a reduction in the scanner's solid angle of acceptance and through that a decrease in the detected scatter and random coincidences. In addition, their utilization reduces dead time effects associated with the readout of coincidence events which are not allowed by the electronic collimation in 2D PET, where only coincidences between crystals of the same or adjacent rings are allowed.
|
8 in the number of overall detected coincidences [3]. This fact leads us logically to question the dominance of 2D PET over the last 20 years. There are, however, a number of reasons that may justify this dominance, the stronger being that the observed increase in the overall detected coincidences is also associated with an increase in the number of scattered and random coincidences (by a factor of 3 and 5, respectively). This in turn places strict and difficult requirements in software correction methodologies. The contamination of the data with scattered coincidences is the harder to account for considering that there is no direct way of measuring the number of detected scattered events. Taking into account the necessary use of approximations, scatter correction algorithms based on simulation modelling techniques have so far demonstrated the most promising results [4, 5]. In terms of random coincidences, an accurate correction may be performed either through a measurement using a delayed time window or directly calculated based on the detected single events per detector. Finally, despite the removal of a number of such erroneous events they still form a large fraction of the overall detected coincidences in 3D PET and as such their removal significantly reduces the statistical quality of reconstructed images. Considering all these issues, it is therefore not surprising that, until recently, 3D PET has been constrained as the gold standard in brain imaging studies where the object in the field of view is small and therefore associated with a smaller scatter fraction. Over the past couple of years the emergence of scanners based on new scintillators has led to an increased interest in the use of 3D PET for whole body investigations in oncology.
One of the detector components that may strongly influence the performance of a PET system is the scintillation crystal used. Desirable properties include high density and atomic number for efficient gamma ray detection resulting in improved overall system sensitivity. The emission of a large number of scintillation photons is also necessary in order to improve positional resolution and through an associated improvement in energy resolution reduce the number of detected scattered coincidences. Finally, a crystal exhibiting short scintillation light decay times will lead to a substantially improved time resolution facilitating a large reduction in the detection of random coincidences. A short scintillation decay time also allows fast integration of the majority of the emitted light for each photon–crystal interaction, improving the dead time characteristics of a detector employing such a fast scintillator. In turn this will translate into higher overall system sensitivity and count rate capabilities. Such scintillation crystal properties will minimize the detection of erroneous events, potentially allowing the advantages of 3D PET for whole body studies to be realised.
Based on such desirable properties, recent interest in the utilization of new crystal materials has led to the employment of two new scintillators in clinical PET scanners. Both lutetium oxyorthosilicate (LSO) and gadolinium oxyorthosilicate (GSO) possess properties that enable a significant reduction in the detection of scatter and random coincidences in comparison with BGO-based detectors. In direct comparison, although LSO has a higher stopping power and faster decay time than GSO, its intrinsic non-proportionality leads to poorer energy resolution [6].
In addition to the employment of new crystals in clinical PET systems, continuing research and development in this area has yielded some other promising candidates, albeit for the moment available in small quantities. The more interesting scintillators emerging from these new developments are lutetium aluminium perovskite (LuAP) and lutetium pyro-silicate (LPS). Both potential candidates exhibit properties that enable superior system time resolution, while the properties of LPS make it a potential candidate for replacing GSO, if however, it can be grown on a large scale. In addition to lutetium-based materials, there is also interest in lanthanum-based scintillation crystals for use in 3D PET, in particular lanthanum bromide (LaBr3). Lanthanum-based crystals have, in the long run, the potential to be produced more cost effectively than GSO or LSO considering their low melting point, which is comparable with NaI(Tl) (800–900°C).
In principle, the potential exists for a number of these new crystals to have a major impact in the performance of a PET scanner operating in 3D. However any such system needs to demonstrate the level of quantitative accuracy obtained with BGO systems operating in 2D mode, before 3D and the new scintillator can be considered as the gold standard in PET whole body imaging. This evaluation is currently under way with LSO and GSO scanners and will probably continue into the future together with other potential crystal elements given that solutions are found for the existing problems in their large scale production.
Over the past decade, the use of image fusion for the combination of PET, CT and MR images has demonstrated potential improvements possible in clinical diagnosis. One of the disadvantages of FDG PET has been the lack of anatomic detail as a result of the technique's limited resolution (4–5 mm). Although some organs are visualized through the natural uptake of FDG, the precise lesion localization is often compromised by the lack of anatomical detail, with an obvious impact on the patient's clinical management. This has been the incentive for the combination of anatomical and functional imaging through, in the past, the use of fusion algorithms and more recently the combination of PET and CT systems using a common examination bed. Until recently, the majority of routinely used image fusion algorithms have been implemented for brain studies. In other areas such as the thorax, the use of non-linear algorithms have demonstrated improved accuracy but suffer from execution times which are too long to be useful in routine clinical practice [7, 8]. There is currently no hard evidence to support the notion that the use of combined PET/CT systems in areas such as the thorax leads to improved image registration in comparison with the use of non-linear registration techniques of datasets acquired in separate systems [9]. Inaccuracies associated with differences in the conditions of acquisition between the two modalities can cause significant errors for both methods of combining PET and CT images, notably in the use of combined PET/CT systems.
Combined PET/CT systems have been the result of recent hardware developments and are currently dominating the market for installation of new PET devices. Apart from the obvious advantage of inherent image fusion, an additional advantage is the ability to perform PET attenuation correction using the acquired CT maps, leading to significant improvements in terms of overall scanning times per PET examination. This is because the use of radioactive sources such as 68Ge and 137Cs, used for the acquisition of transmission maps in PET-only devices, account for 25–40% of the time required for a whole body imaging session. Therefore, combined PET/CT devices can have an impact not only in further improving the accuracy of PET in staging studies but also in issues related to patient throughput.
A substantial body of evidence exists today on the accuracy of CT-based attenuation correction (CTAC) in PET. In terms of scaling the CT attenuation coefficients (energies of 70–80 keV) to those corresponding to PET (511 keV), a post-processing step such as scaling, segmentation or a combination of the two are currently implemented in clinical combined systems [10]. Studies carried out to date using such methodologies have demonstrated comparable quantitative accuracy combined with an improved signal to noise ratio in the reconstructed PET images compared with radioactive source-based attenuation corrections [11]. Potential issues are associated with the presence of CT contrast agents or metallic implants, which can lead to image artefacts and compromise PET quantitative accuracy in those areas [11]. Potential solutions include the use of post-acquisition processing steps, such as segmentation of areas with contrast or metallic objects and subsequent reclassification of the CT numbers in order to minimize such effects [12, 13]. In addition, artefacts present at the level of the diaphragm as a result of differences in respiration conditions between the two acquisitions, may be reduced by using a "normal expiration and breath hold" CT acquisition protocol [14]. Such a breath hold pattern however, may be difficult to achieve for some patient conditions. Although respiratory artefacts at the level of the diaphragm were more severe in early PET/CT systems utilizing a single slice CT scanner [15], they have been reduced with the introduction of multislice CT scanners (at least 4 slices), allowing whole body CT acquisitions of <30 s.
Future developments in combined PET/CT systems will include the use of state of the art multislice (16 to 32) CT scanners [16], allowing a more efficient dose utilization as well as submillimetre collimation for most imaging protocols. However, faster CT imaging times should not have a significant effect on combined PET/CT oncology investigations, as the present limiting factor is the time required to acquire PET whole body images of sufficient quality.
As already described up to this point, PET technology has in the past been continuously evolving with changes driven predominantly by clinical need and established oncology applications focusing on whole body imaging protocols. This trend should continue over the next few years, with developments concentrating on enhancing patient throughput and establishing new and more focused clinical applications. One such emerging clinical application currently attracting increased interest is the use of PET in radiotherapy treatment planning. Intensity-modulated conformal radiotherapy aims to deliver higher doses in the area of interest whilst simultaneously minimizing the dose to normal tissues. PET/CT systems can be the ideal instruments for radiotherapy treatment planning by providing exact anatomical localization of the area of interest in combination with an accurate delineation of the 3D functional volume to be treated. There are however, certain problems that need to be addressed before PET can become the gold standard in radiotherapy treatment planning. First, if combined PET/CT systems are to fulfil both roles of accurate localization and treatment volume delineation they need to be integrated in the treatment planning process. This involves establishing the CT component of the PET/CT scanner as the treatment simulation system by integrating with such systems a number of additional components. These include re-alignment lasers allowing the unification of the coordinate system of acquired CT datasets with the radiotherapy treatment systems, the incorporation of flat scanning beds as well as potentially larger patient ports. In addition, issues associated with misregistration of PET and CT datasets as well as inaccuracies in the delineation of functional volumes as a result of respiratory motion will have to be addressed. A proposed solution involves the acquisition of respiratory gated PET datasets, resulting in a number of frames corresponding to different points throughout the respiratory cycle [17, 18]. The problem associated with this approach is that the resulting PET images are of reduced resolution since only a fraction of the available counts is used in each of the reconstructed frames. Future directions in addressing this problem will involve combining models that describe the movement of internal organs with reconstruction algorithms utilizing the patient specific respiratory motion information recovered from gated PET acquisitions.
Finally, in the area of combined imaging devices one should not forget the potential of combining PET and MRI devices, the principle of which has already been demonstrated few years ago through the development of a small scale system [19]. Although a number of challenges will have to be met before such a system is scaled up for human studies, it is certainly possible that it may become a reality within the next few years.
As already mentioned earlier, research is continuing into new scintillator materials for PET. However, the performance of a PET system can be also strongly affected by the photodetector employed or the overall detector design itself. Requirements of an ideal detector design include the correct identification of crystal elements, preservation of crystal intrinsic energy resolution and finally, through the use of appropriate electronics, a fast timing response. Despite the domination over the past two decades of the block detector concept [20] in clinical PET systems, new detector designs are appearing. For example, in the case of the Philips AllegroTM there is a continuous light guide coupling the individual crystal elements with a pack of photomultiplier tubes (PMTs) which minimize the variation in light collection from different crystals, thus improving the energy resolution of the detector [6]. In addition, flat panel detectors using a variant of the block detector known as "quadrant sharing design" [21] are also emerging. This detector design has a better encoding ratio than the traditional block detector designs and can therefore resolve smaller crystals for a given PMT size, allowing an improvement in positional information without an associated loss in energy resolution and a large increase in the overall cost. Such small panel detectors have already been used in the construction of a research brain tomograph [22]. A prototype whole body system based on 5 rotating panel LSO detectors (52 cm x 36 cm) has been also constructed. As a result of the large axial extent of this system in comparison with 15 cm in traditional ring designs, it has been suggested that this system will be able to perform whole body acquisitions in 5 min to 10 min without any compromise in image quality [23]. If such a performance can be realised in routine clinical practice it will most definitely shape the future of PET technology.
Finally, in terms of new photodetectors and overall detector designs, a number of alternative technologies already employed in smaller scale animal PET scanners [24] may find their way in future clinical PET systems. These include the combination of small surface area crystal elements (<2 mm x 2 mm) in combination with new photodetectors such as position sensitive PMTs or Avalanche Photodiodes (APDs). A number of such detector designs may either form the basis of a combined PET/MRI system (as a result of the reduced sensitivity of such photodetectors in the presence of magnetic fields) or lead to improvements in resolution uniformity across a scanner's field of view.
In conclusion, major developments in PET technology have already played a major part in defining and establishing the role of this imaging modality in oncology. The introduction of PET/CT has further revolutionized and is in the process of refining this role. New software developments combined with the introduction of new scintillators and PET detector designs hold the potential to improve the throughput of this technique and open the way to new clinical applications. The present is bright but the future will certainly be exciting.
Received for publication February 24, 2004. Revision received June 15, 2004. Accepted for publication July 12, 2004.
 |
References |
|---|
 Top References |
|---|
This article has been cited by other articles:
 |
|
 |
|
  BJR Review of the Year - 2004 Br. J. Radiol., March 1, 2005; 78(927): 181 - 185. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
