CURRENT AND FUTURE TRENDS IN PET INSTRUMENTATION

Transcription

1 ANALELE ŞTIINŢIFICE ALE UNIVERSITĂŢII AL. I. CUZA IAŞI Tomul III, s. Biofizică, Fizică medicală şi Fizica mediului 2007 CURRENT AND FUTURE TRENDS IN PET INSTRUMENTATION D. Mihăilescu 1, C. Borcia 1 KEYWORDS: PET, PET/CT, PET/MRI, scintillators, photodetectors. The history of PET has been one of continuous improvement in resolution and sensitivity, mainly due to the continuous efforts of both academic community and nuclear medicine industry to integrate the appropriate research findings for the design of different geometries and various detector/readout technologies. As PET has become integrated in clinical and research practice, several design trends have been developed, from those designed to low cost clinical applications, to others designed for high resolution research applications (dedicated PET scanners for brain imaging and small animals PET scanners). Another emerging and very promising technology is PET/CT and PET/MRI dual-modality imaging which leads to significant improvements in both image resolution and sensitivity. The choice of the scintillator is a fundamental element in PET design. The selection is generally made after a carefully examination of their physical characteristics. The aim of this work is to review the current and future trends in PET instrumentation. A special attention is given to block detector design, readout technologies (various types of photomultiplicators, hybrid photodetectors, silicon p-i-n photodiodes, avalanches photodiodes), as well as to the new cerium doped scintillator crystals like lutetium orthosilicate (LSO:Ce), lutetium yttrium orthosilicate (LYSO:Ce) or lanthanum bromide (LaBr:Ce). 1. INTRODUCTION Positron Emission Tomography (PET) is used to obtain information on biochemical processes in living organisms, being a tool of increasing importance for medical diagnosis and bio-medical research. The history of PET [1] has been one of continuous improvement in the resolution and sensitivity of the imaging devices. Despite advances in other imaging methods such as Computer Tomography (CT) and Magnetic Resonance Imaging (MRI), the ability to image the metabolic abnormalities associated with disease has made PET one of the most significant diagnostic tools ever developed. The next generation in PET technology, PET/CT fusion imaging, has the ability to combine CT structural information with PET s metabolic information into a single set of images. This ability to detect the exact location of a metabolic hot spot by overlaying the PET and CT images provides priceless information for physicians in the treatment of cancer and other metabolic diseases. In the years to come, as more is learned about the fundamental processes of diseases and as new radiopharmaceuticals 1 Faculty of Physics, Al. I. Cuza University, Iasi, Romania

2 38 D. Mihăilescu, C. Borcia and analysis tools are developed, PET and PET/ CT scanning will prove to be an invaluable tool in the diagnosis and treatment of some of the most critical diseases challenging modern medicine. The improvement and expansion of these techniques depend on the progress attained in several areas, such as radionuclide production, radiopharmaceuticals, radiation detectors and image reconstruction algorithms. This review paper will be concerned only with the detector technology. We will review in general terms the current and future trends in PET instrumentation with the emphasis put on the developments of high-resolution detector systems for dedicated PET scanners and animal imaging. The present trend to combine two or more modalities in a single machine in order to obtain complementary information will also be considered [2]. All of these systems are undergoing revisions in both hardware and software components. New technologies that are emerging include the use of LSO and GSO as alternatives to BGO crystals, the use of layered crystals and other schemes for depthof-interaction (DOI) determination (for dedicated systems) and for hybrid PET/CT devices. In addition, high-resolution animal scanner designs are plentiful and a number of such devices are now being offered commercially. There are many different design paths being pursued that could be the mainstream of future research systems generally situated between the following major design paths: (1) lower cost yet higher performance whole body imaging systems; (2) (2) very high spatial resolution systems aimed primarily at research applications. The major technologies currently under development seem to be centered on improved, cheaper electronics and better scintillators. One technology that has yet to make a significant impact in this field is the use of solid-state devices either for light collection or primary photon collection. The problem has usually been cost, but new technology may yet allow such devices to be considered for future PET systems [3, 4]. 2. THE PHYSICS OF PET In a PET study, one administers a radio-pharmaceutical labeled with a positronemitting isotope by injection or inhalation. A great number of radiopharmaceuticals labeled with 11 C, 13 N, 15 O, or 18 F positron-emitters have been applied both for biomedical research and clinical purposes in neurology, cardiology, oncology and many other branches of medicine and life sciences. The recent success of PET with rapidly increasing installations is mainly based on the use of 18 F-fluorodeoxyglucose (F-18 FDG) in oncology where it is most useful to localize primary tumors and their metastases. PET has always provided information on the metabolic function of organs or tissues by detecting how cells process certain compounds such as, for example, glucose. Cancer cells metabolize glucose at a much higher rate than normal tissues. By detecting increased radio-labeled glucose metabolism with a high degree of sensitivity, PET identifies cancerous cells, even at an early stage when other imaging modalities may miss them [5].

3 CURRENT AND FUTURE TRENDS IN PET 39 The positron-emitting isotope circulates through the bloodstream to reach a particular organ. The positrons emitted by the radioisotope have a very short range in tissue (from less than 1mm to several mm) and undergo annihilation with an available electron. This process generally results in an emission of two gamma rays, each with energy of MeV, moving away from the point of production in nearly opposite directions. This photon pair is equal for all positron emitters so that a detector system can be optimized according to this constant physical property. The photon pairs are measured by a system of detector rings (usually named PET camera) in which the patient is situated. The detectors consist of blocks of small scintillation crystals coupled to one or more photomultipliers. A photon pair is registered if one of its photons hits a detector crystal and the other hits one of the fan-like opposing crystals within the coincidence time window of e.g. 6 ns. Then the origin of the positron emission, which is the position of the radio-labeled molecule, is assumed to be located on the line between the two detector crystals [5]. The PET machine generates transverse images depicting the distribution of positron-emitting radioisotopes in the patient and uses annihilation coincidence detection to obtain projections of the activity distribution. The transverse images are obtained through the process of filtered back-projection. A PET scan can require 5 60 min, depending on parameters such as the system photon sensitivity, the mode of acquisition, the size of the imaging subject region of interest, and the amount of injected activity [6]. A PET system can be characterized by several important performance parameters, such us: the photon sensitivity, the spatial resolution, the energy resolution, the temporal resolution, the contrast resolution, the counting rate, the deadtime, the signal-to-noise ratio and the molecular sensitivity. The photon sensitivity is the fraction of all coincident 511-keV photon pairs emitted from the imaging subject that are recorded by the system, and is also referred to as the coincidence photon detection efficiency. This parameter determines the statistical quality of image data for a given acquisition time. Photon sensitivity impacts image quality because it influences the noise level of images reconstructed at a desired spatial resolution. The spatial resolution describes a system s ability to distinguish two closely spaced molecular probe concentrations and is important to detect and visualize subtle molecular signals from miniscule structures. Energy resolution is the precision that one can measure the incoming photon energy. Since scattered photons lose energy, good energy resolution may allow a narrow energy window to reduce scatter photon contamination in image data without significantly compromising photon sensitivity. A narrow energy window also helps to reduce the rate of random photon contamination since many of these photons also undergo scatter. The coincidence time resolution determines how well one can decide whether two coincident photons truly arrive simultaneously. Analogous to benefits of good energy resolution, good coincidence time resolution may allow a narrow time window to reduce random events without compromising photon sensitivity. The energy and temporal resolutions work together to define the available system contrast resolution, which is the ability to differentiate two slightly different concentration levels of probe in adjacent targets.

4 40 D. Mihăilescu, C. Borcia The photon sensitivity, spatial resolution, and contrast resolution work together to define the molecular sensitivity of a PET instrument. A detailed discussion about the improving modalities of the main performance parameters of PET scanners can be found in the review paper of C.S. Levin [6]. Theoretically, an ideal PET tomograph should have high sensitivity and counting rate capability in addition to low dead-time losses, and provide a uniform spatial and energy resolution over the whole sensitive volume allowing to reach the highest signal-to-noise ratio (SNR) for the lowest possible injected activity. Perfect correction for variable crystal detection efficiency and geometrical factors, random coincidences, photon scattering and attenuation, partial volume and other background effects is also highly desired to allow a good image quality or accurate measurement of quantitative indices in vivo [4]. 3. TRENDS IN DETECTOR TECHNOLOGY PET has achieved a high level of technological perfection mainly as a result of the improvements in detector technology, computer hardware, and image processing software [7]. Recent developments of new PET detector modules and scanner designs have followed three main trends [2]: (1) Providing general purpose PET systems with DOI information. (2) Designing detector modules and scanners for specific purposes, namely brain, breast, prostate and small animal imaging. For these applications, detectors with very high space resolution, approaching the limit imposed by the range of the positron before annihilation, are required. (3) Constructing cameras capable of performing two different types of imaging modalities in the same scanner, such as, for example, PET and X-ray tomography. The advantage of this combination is that PET provides information on the tissue functioning while the X-ray tomography gives a high-resolution image of the object morphology. 3.1 Trends in PET detector modules The schematic of a conventional PET detector module, also named block detector, is shown in figure 1 [2]. A block of BGO scintillator crystal is partially sawn through to make a group of 8 x 8 = 64 quasi-independent crystals that are optically coupled to four photomultiplier tubes. When a gamma ray interacts in the crystal, the resulting scintillation photons are emitted isotropically but the saw cuts limit their lateral dispersion as they travel toward the photomultiplier tubes. The position (i.e. crystal element) of the gamma ray interaction is then determined by the analog ratio of the photomultiplier tube output signals, and the gamma ray energy is determined and a timing pulse generated by the sum of these four signals. A typical PET detector module has 80% detection efficiency, 20% fwhm energy resolution, 2 ns fwhm timing

5 CURRENT AND FUTURE TRENDS IN PET 41 resolution, 4 μs dead time, and 5 mm fwhm position resolution for 511 kev gamma rays [3, 8]. Fig. 1: Conventional PET detector module. The detector module performance is limited by the BGO scintillator crystal. A scintillator with a faster decay time would improve timing resolution and decrease dead time, while one with a higher light output would improve energy resolution and spatial resolution (by allowing more crystals per block to be unambiguously decoded). However, short attenuation length is critical in order to maintain high spatial resolution and for this reason BGO dominates. However, some recently developed scintillators (see the next section) are being incorporated in experimental PET systems. In order to increase efficiency and reduce the number of detector modules (and hence cost), PET camera designers would like to reduce the diameter of the detector ring. Unfortunately, they are prevented from doing this by a resolution degradation artefact caused by penetration of the 511 kev photons into the crystal ring. The origin of this artefact, variously known as radial elongation, parallax error, or radial astigmatism, is shown in figure 2. Photons that impinge on the detector ring at an oblique angle can penetrate into adjacent crystals before they interact and are detected, which causes mispositioning errors (i.e. events are assigned to chords that do not pass through the source). This spatial resolution degradation increases for objects placed further away from the center of the tomograph ring. This artefact can be reduced significantly or eliminated if the detector module is capable of measuring not only the identity of the crystal of interaction but the depth-of-interaction (DOI) within that crystal. With such information, the event can be assigned to the chord that connects the interaction points (rather than the interaction crystals), and as that chord will pass through the source, no mispositioning errors are generated [8].

6 42 D. Mihăilescu, C. Borcia Fig. 2: Cause of the parallax error. Developing a detector module capable of accurately measuring this interaction depth is an active field of research. Figure 3 schematically shows two general approaches that have been taken to measure depth-of-interaction (DOI). The first, shown in figure 3(a), is a phoswich approach, in which the scintillator block of a conventional PET detector is effectively replaced with two or more layers of different scintillator materials, so each scintillator pixel now contains stratified layers of scintillator material with different decay times. By providing the readout electronics with pulse shape discrimination capability, the type of scintillator where the interaction occurred is identified. This gives rough information on the depth-ofinteraction (DOI) which depends on the thickness of each layer [9]. Fig. 3: Methods of determination the depth-of-interaction of gamma rays within scintillator crystals: (a) using a block made of two layers of crystals with different decay times; (b) sharing the light among a photomultiplier and an array of silicon photodiodes coupled to the opposite sides of a block of optically isolated crystals.

7 CURRENT AND FUTURE TRENDS IN PET 43 The second general technique for measuring depth-of-interaction shown in figure 3b is to utilize light sharing. With this approach, each scintillator element is attached to two photodetectors, usually on opposing ends of the crystal. The amount of light observed by each photodetector depends on the interaction position; the depthof-interaction is estimated from the ratio of the two photodetector signals. Recent advances in pixellated photodetectors have contributed greatly to this design approach. Many combinations of photodetectors have been used, including single anode photomultiplier tubes, PIN photodiode arrays, avalanche photodiode arrays, and multianode photomultiplier tubes. At present, no cameras have been built utilizing any of these schemes, although several are under construction [3]. 3.2 New materials for detection crystals The critical component of PET tomographs is the scintillation detector [4, 10]. The scintillation process involves the conversion of photons energy into visible light via interaction with a scintillating material. Increased light yield, faster rise and decay times, greater stopping power and improved energy resolution, linearity of response with energy, in addition to low cost, availability, mechanical strength, moisture resistance, and machinability are the desired characteristics of scintillation crystals [11]. Table 1 summarizes most of these properties for selected scintillators currently in use or under development for PET applications [12]. Improvements in these characteristics enable detectors to be divided into smaller elements, thus increasing resolution and minimizing dead-time losses [13]. Table 1: Characteristics of scintillation crystals used or developed specifically for the design of current or future generation PET imaging systems [12]. Scintillator BGO LSO GSO LuAP LaBr 3 LYSO Formula Bi 4 Ge 3 O 12 Lu 2 SiO 5 :Ce Gd 2 SiO 5 :Ce LuAlO 3 :Ce LaBr 3 :Ce LuYSiO 5 :Ce Density (g/cm 3 ) Light yield (photons/kev) Effective Z Principal decay time (ns) Peak wave length (nm) Index of refraction Photofraction (%) * Attenuation length (cm) * Energy Resolution (%) * Hygroscopic No No No No Yes No *511 kev

8 44 D. Mihăilescu, C. Borcia The choice of the scintillator is a fundamental element of a PET design. The selection is generally made after careful consideration of the physical characteristics mentioned above. Bismuth germanate (BGO) has a very high physical density and effective atomic number, and is not hygroscopic. These properties rendered it the preferred scintillator for commercial PET units in the 1990s. Its major disadvantages are, however, the low light yield and only a moderately fast decay time that limits coincidence timing and count rate performance [4]. New detection technologies that are emerging include the use of new cerium doped crystals as alternatives to conventional BGO crystals, and the use of layered crystals and other schemes for depth-of-interaction (DOI) determination, taking advantage of excellent timing resolution of new scintillators, along with many other designs [3,4]. It appears that cerium doped lutetium orthosilicate (LSO:Ce) produced by CTI Molecular Imaging (Knoxville, TN), lutetium yttrium orthosilicate(lyso:ce) produced by Photonic Materials Ltd. (Bellshill, UK) and cerium doped lanthanum bromide (LaBr3:Ce), under development by Saint Gobain (France), are the most promising candidates [4,14]. They combine high density and high atomic number necessary for an efficient photoelectric conversion of the annihilation photons, with a short decay time of the scintillation light, which is a key requirement for high counting rates. Phoswich detectors received considerable attention for the design of highresolution scanners dedicated for brain, female breast and small animal imaging. This may be implemented with solid-state photodiode readouts, which also allows electronically collimated coincidence counting. The phoswich technique is able to roughly halve the uncertainty of the DOI, and hence significantly reduces the parallax error, which is inherent to all radial geometries. However, the phoswich approach is still only a compromise between the maximum crystal length, which can be tolerated for parallax error reasons, and the minimum length, which is required to achieve high detection efficiency. In addition, the phoswhich approach demands a delicate pulse shape discrimination of the analog signal delivered by the PMT s in order to identify the hit crystal layer. The readout electronics becomes inevitably more complex and possibly limits the data acquisition rate. The axial arrangement of long finely segmented arrays of long high-z scintillation crystals coupled to highly pixelated photodetectors leads to an essentially parallax free geometry, which allows reconstructing the interaction point of the annihilation photon in true 3-D reconstruction [4]. 3.3 Photodetectors Recent developments in photodetectors for medical applications should enable efficient collection of the light emanating from the scintillation crystals [15]. The design of high resolution imaging devices imposes some additional constraints with respect to the necessity for compact arrays of photodetectors; in turn, this has stimulated the development and use of multichannel position-sensitive photomultiplier tubes (PS-PMT s), silicon p-i-n photodiodes (PDs) and avalanche photodiodes (APDs). Solid-state photodiodes exhibit many advantages compared to conventional PMT s. They are relatively small, operate at much smaller voltage, and more importantly, exhibit higher quantum efficiencies. Furthermore, photodiodes are

9 CURRENT AND FUTURE TRENDS IN PET 45 insensitive to axial and transversal strong magnetic fields and therefore, have the potential to be operated within MRI systems. By using this technology, the sensitive area of the detector could be read out more efficiently, taking advantage of recent developments of monolithic pixilated photodetectors. Thanks to the progress made in the microelectronic industry, high-purity detector grade silicon now is available commercially and can be used to produce low-noise silicon planar photodiodes, avalanche photodiodes, and silicon drift photodetectors [16]. Commercially available APDs typically are not pixelated and can have a sensitive area of ~20 mm 2. PDs and APDs can also be segmented to form a linear or two-dimensional array of pixels on a single device [17]. Their geometry can be modified to suit specific applications, with the configuration of the scintillator matrix adapted to the available read out pixelation of the photodetector. However, given the current cost of solid-state photodiodes, they need to offer very significant advantages to replace the currently adopted technology. Although not previously used in medical imaging, hybrid photodetectors (HPDs) represent an interesting technology with potential for high-resolution photon detection [18]. An HPD consists of a phototube with two key components: a semitransparent photocathode (deposited by vacuum evaporation on the entrance window) with high sensitivity for photons in the visible and near UV range, and a silicon (Si) sensor which serves as the anode [19]. The photocathode receives light from a scintillator and converts the incident photons to electrons. These photoelectrons are accelerated by an electrostatic field (10-20 kv between cathode and anode), which is shaped by a set of ring electrodes, onto the segmented Si sensor; the anode thereby creates a signal proportional to the kinetic energy of the incident photoelectrons. In this way, the HPD combines the sensitivity to single photons of a conventional PMT with the spatial and energy resolution of a solid-state sensor. In addition, this design overcomes the intrinsic limitations of a classical PMT with respect to the statistical fluctuations in the number of electrons at the first dynodes. Some hybrid photodetectors (HPDs) can be operated in strong magnetic fields, as long as the field direction is aligned with the tube axis. Axial magnetic fields even have the beneficial effect of reducing the intrinsic spatial resolution of the device, which is a consequence of the angular and energy distribution of the photoelectrons at emission from the photocathode [20]. HPDs have been already incorporated as major components of the new generation PET cameras. However, the use of semiconductor detectors in PET is far from reaching acceptance; these devices are regarded as especially promising candidates for the design of PET cameras with the aim of overcoming the drawbacks of conventional PMT based PET instrumentation [16]. 4. DEDICATED PET SCANNERS One of the main current trends in PET instrumentation consist in designing specific purpose scanners, such as those dedicated to brain [4,21-23], breast [24-29], prostate or small animals imaging. Conventional whole-body PET cameras can image any part of the body. Their development is mature enough that the main gains to be made are in the cost / performance trade-off, where only relatively small gains are possible [3]. However,

10 46 D. Mihăilescu, C. Borcia cameras can be optimized for imaging a single organ or small animals (mice, rats, small primates), which could result in large performance gains at the expense of limited body coverage. By using small scintillator crystals and multi-anode photomultiplier tubes, spatial resolutions below 2 mm fwhm have been achieved [30-33]. 4.1 Dedicated brain PET cameras In PET systems designed for human brain imaging, the ultimate performance in terms of spatial resolution, sensitivity and signal-to-noise ratio (SNR), can be achieved through careful selection of the design geometry, detector assembly and readout electronics as well as optimized data acquisition protocols and image reconstruction algorithms. The rationale in designing dedicated brain vs multipurpose whole-body PET scanners resides in the fact that unlike whole-body imaging, where a larger detector ring diameter is needed to accommodate large patients, the size of the human head is relatively small, thus allowing to reduce the scanner s ring diameter and to increase the solid-angle coverage, leading to higher sensitivity per unit detector volume [30]. A small ring design has the advantage of improving the inherent spatial resolution degradation due to noncollinearity of the annihilation photons in addition to reducing the overall cost of the PET tomograph at the expense of a higher parallax error, hence an image degradation depending on the emission point in the transaxial plane and/or the angle of incidence of the lines of responses. This effect becomes worse when reducing the scanner s ring diameter or the size of the crystal s crosssectional area. This inherent limitation could be coped with by keeping the radial length of the crystal small, typically on values around the attenuation length at 511 kev, which however strongly compromises the detection efficiency. As mentioned earlier, the phoswich approach, which leads to a better approximation of the interaction point, has recently been implemented in several designs. Such an approach halves the parallax error for a given crystal length, but still only represents a compromise between maximum sensitivity and optimum spatial resolution [4]. Theoretically, a dedicated brain PET scanner should be able to provide higher spatial resolution, while at the same time offering a sufficiently high counting rate capability to allow investigation of physiological processes with fast temporal dynamics [31]. There has been a remarkable progress in PET instrumentation design from a single ring of BGO crystals with a spatial resolution of ~15 mm, to multiple rings of BGO and more recently, GSO and LSO detector blocks resulting in a spatial resolution of about 4-6 mm. Improvements in spatial resolution have been achieved by the use of smaller crystals and the efficient use of light sharing schemes to identify the active detector cell. On the other hand, improvements in sensitivity have been accomplished through fully 3-D acquisition mode by removing the interplane septa, which increases coincidence efficiency by about a factor of around five in comparison to 2-D acquisition with interplane septa extended [32]. Newer tomographs operate essentially in fully 3-D acquisition geometries, at the expense of increasing system sensitivity to scattered radiation and random coincidences. Several innovative developments in PET instrumentation have been proposed or are currently under design or test. These include large or pixelated detectors

11 CURRENT AND FUTURE TRENDS IN PET 47 mounted on a rotating gantry, detectors arranged in a cylindrical partial or full multiring geometry, and detectors assembled in a polygonal ring. A high resolution PET camera design was proposed in the 90 s by optimizing the detector size and system diameter through optimal arrangement of a small number of large detectors in a circular ring format combined with a dedicated sampling scheme [33]. Monte Carlo simulation studies showed that the proposed design allows obtaining uniform highresolution (2-3 mm FWHM) images [34]. Braem et al. [35] have proposed a novel detector design, which provides full 3- D reconstruction free of parallax errors with excellent spatial resolution over the total detector volume. The key components are a matrix of long scintillator crystals coupled on both ends to HPDs with matched segmentation and integrated readout electronics. The data acquisition system is composed of several read out cards (each one associated with a module of the PET scanner) and a main card that controls the whole system [36]. Using fast triggering signals from the silicon sensor back-planes, the main card performs the coincidence event analysis and enables the read out of the two modules involved in case of coincidence. The other modules are left free to perform new acquisitions. This concept based on several independent, event driven and parallel read out chains, considerably reduces the acquisition dead time. Computer simulations and Monte Carlo modeling predict that the detector will achieve excellent spatial (x,y,z) and energy resolution [37]. The design also increases detection efficiency by reconstructing a significant fraction of events that undergo Compton scattering in the crystals. The 3-D axial detector geometry is configured from a matrix of 208 (13x16) long crystals, each with a cross section of 3.2x3.2 mm 2 and with an inter-crystal spacing of 0.8 mm. Scintillation light produced after an interaction of an annihilation photon will propagate by total internal reflection to the ends of the crystal, where it will be detected by the HPD photodetectors. The transaxial resolution depends only on the crystal segmentation and not on its chemical composition, whereas the axial resolution is closely related to the scintillator properties. The scintillator s optical bulk absorption length should be approximately equal to the crystal length to obtain both a high light yield and a significant light asymmetry required to decode the axial coordinate z of the photon interaction in the crystal matrix [38]. 4.2 Dedicated breast PET cameras Early detection of breast cancer is crucial for efficient and effective treatment. Functional breast imaging could improve imaging of such lesions because it provides a non-invasive method in the screening of women for the presence of the disease, as well as for staging disease and tracking its response to therapy. Dedicated breast PET imaging, called positron emission mammography (PEM), is a new technique to obtain images of the breast for detection of radiopharmaceutical-avid tumours. Metabolic images from PEM contain unique information not available from conventional morphologic imaging techniques and aid in establishing the diagnosis of breast cancer. Therefore, PEM might be used for staging, characterizing indeterminate lesions, and finding occult lesions in those patients who are difficult to scan with mammography. PEM promises to achieve low-cost directed functional examination of breast

12 48 D. Mihăilescu, C. Borcia abnormalities, with the potential for performing X-ray correlation and image-guided biopsy. A number of dedicated PEM cameras optimized to image the breast have been proposed or constructed [22-27]. These cameras restrict the field of view to a single breast, and have higher performance and lower cost than a conventional PET camera. By placing the detectors close to the breast, the PEM geometry subtends more solid angle around the breast than a conventional PET camera. In addition, gamma rays emitted in the breast have to pass through at most one attenuation length (~ 10 cm) of tissue in the PEM geometry, but may have to travel through as much as four attenuation lengths of tissue in a conventional PET camera. These two factors significantly increase the sensitivity (the detected coincident event rate per unit activity in the field of view) in the PEM geometry [27]. The most common PEM geometry is the parallel plane one. It consists of a pair of parallel plane detector heads placed above and below a single breast. The size of the plane varies, but is usually less than 20 cm on a side so that it can fit within an X- ray mammography unit. The spacing between detector planes is variable, allowing mild breast compression. Compression serves several purposes. It reduces motion artifacts, it spreads out the breast (making it easier to resolve separate structures) and it thins the breast (which improves the contrast as it reduces the amount of normal tissue in the field of view). Recent advances in scintillator array and position sensitive photomultiplier tube (PSPMT) technologies have simplified construction of PEM camera detectors. PSPMTs can economically and accurately read out arrays of optically isolated scintillator pixels. These devices generally produce four analog output signals. The sum of the four signals is proportional to the total energy deposited into the scintillator crystal array, while the ratio of provides the 2-dimensional center of gravity of the optical signal and thus the position of the crystal that the interaction took place in. While each anode of a multi-anode PMT could be coupled one-to-one to a scintillator crystal, it is more common to use an external resistor network to effectively turn them into PSPMTs. The spatial resolution of the camera is largely determined by the in-plane size of the scintillator crystal. To identify 5mm diameter tumours, high spatial resolution (~ 2 mm) is required, so the scintillator crystals are usually between 2 and 4mm 2. Thick (~ 3 attenuation length) scintillator crystals are desired in order to obtain high efficiency. Scintillator crystals with high density and high atomic number (such as LSO, BGO, GSO, or LuAP) are also desired, although cameras have also been constructed using NaI:Tl scintillator [27]. 4.3 Small animal PET scanners Another field that is growing rapidly is PET scanners for imaging small animals, especially mice and rats. PET s ability to measure biochemical function, rather than structure, can provide crucial insight into the functioning of new and existing pharmaceuticals, the nature of diseases, or the function of specific genes. These experiments are usually performed in small animals, requiring resolutions much higher than those achieved in human PET scanners [39-46].

13 CURRENT AND FUTURE TRENDS IN PET 49 MicroPET [46], for example, is a state-of-the art high resolution PET scanner especially designed for imaging small laboratory animals. It consists of a ring of 30 position sensitive scintillation detectors, each with an 8 x 8 array of small lutetium oxyorthosilicate (LSO) crystals coupled via optical fibers to a multi-channel photomultiplier tube. The detectors have an intrinsic resolution averaging 1.68 mm, an energy resolution between 15 and 25% and 2.4 ns timing resolution at 511 kev. The detector ring diameter of micropet is 17.2 cm with an imaging field of view of 112 mm transaxially by 18 mm axially. The scanner has no septa and operates exclusively in 3D mode. Reconstructed image resolution 1 cm from the center of the scanner is 2 mm and virtually isotropic, yielding a volume resolution of 8 mm 3. For comparison, the volume resolution of a typical PET system used in clinical imaging is in the range mm 3 [44]. The system also includes a computer controlled animal bed with built in wobble motion. MicroPET produces images with a spatial resolution of < 2 mm in all three axes and has sufficient sensitivity to realize this resolution for many applications. For studying biological systems that are easy to saturate, improved sensitivity could be realized by the addition of a second or third detector ring, which would also serve to increase the axial coverage of the system [45]. 5. DUAL-MODALITY IMAGING SYSTEMS Software- and hardware-based correlation between anatomical (x-ray CT, MRI) and physiological (PET) information is a promising research field and now offers unique capabilities for the medical imaging community and biomedical researchers. One of the main advantages of dual-modality PET/CT imaging is that PET data are intrinsically aligned to anatomical information from the x-ray CT without the use of external markers or internal landmarks, thus providing a reliable estimate of the attenuation map to be used for attenuation and scatter correction purposes. On the other hand, combining PET with MRI technology is scientifically more challenging owing to the strong magnetic fields. There has been a strong trend in recent years to equip SPECT cameras (which are optimized to detect 140 kev gamma rays) with coincidence electronics and give them the ability to obtain PET images. The benefits of this are largely economic - SPECT cameras are far more common than dedicated PET cameras and so any hospital with a SPECT camera can, for a relatively small cost, also have the ability to acquire PET images. Some compromises in performance (as compared to dedicated PET cameras) are necessary, but clinically valuable images are often obtained [4]. 5.1 PET/CT imaging systems During the course of patient diagnosis, anatomical imaging is usually performed with techniques such as CT or MRI that have excellent spatial resolution and signal-to-noise-ratio (SNR) characteristics, but that may offer relatively low specificity for differentiating disease from normal structures. In opposition, PET generally targets a specific functional or metabolic signature in a way that can be

14 50 D. Mihăilescu, C. Borcia highly specific, but generally lacks spatial resolution and anatomical evidence that often are needed to localize or stage the disease, or for planning therapy in cancers of the brain, head and neck. The availability of correlated functional and anatomical images would improve the detection of disease by highlighting areas of increased radiotracer uptake on the anatomical CT or MRI scan, whereas regions that look abnormal in the anatomical image can draw attention to a potential area of disease where radiopharmaceutical uptake may be low [4]. Fusion of images from the PET and CT scan can allow for convenient visualization of relevant information, and accurate fusion facilitates radiotherapy planning. The fusion can be performed visually, using software or by so-called hardware fusion. Commercial packages are available for software fusion, but the transformation algorithms used cannot correct for extreme differences in positioning or motion artefacts. The term hardware fusion currently refers to a PET/CT scanner that consists of separate scanners, which are positioned in line at a fixed distance, with projection of the PET image over the CT image. The hybrid PET/CT imaging permits both CT and PET scans to be performed in a single examination, with CT images used for anatomic reference of the tracer uptake patterns imaged in PET, as well as for attenuation correction of the PET data [47]. PET and CT obviously complement each other in providing important diagnostic information. Separate PET and CT images are unfortunately difficult to fuse because the patient is generally not positioned identically on both machines. On the other hand, the recently introduced PET/CT machines, integrating PET and CT technologies into a single device, enable the collection of both anatomical and biological information simultaneously during a single examination, resulting in accurately fused PET and CT images that permit a more accurate tumour detection and tumour localization for a wide variety of cancers. The main advantages of PET/CT machines are as follows: earlier diagnosis of disease, accurate staging and tumour localization, more precise treatment, monitoring of response to treatment and early detection of recurrences, reduction of biopsy sampling errors, reduction of the number of invasive procedures. 5.2 PET/MRI imaging systems The main difficulty with PET is the lack of an anatomical reference frame. Magnetic resonance imaging (MRI) is an excellent morphological imaging modality with a high anatomical resolution. Whole-body MRI produces large amounts of image data, resulting in the possibility of overlooking subtle pathological findings. The fusion of PET with MRI can compensate for their disadvantages and therefore offers several advantages in comparison to PET or MRI alone. The combination of these two excellent diagnostic imaging modalities into a single scanner improves the diagnostic accuracy by facilitating the accurate registration of molecular aspects and metabolic alterations of the diseases with exact correlation to anatomical findings and morphological information. Whole-body PET/MRI is a very promising diagnostic modality for oncological imaging and for use in cancer screening in the decades to come due to the considerably lower radiation exposure in contrast to PET/CT and the high soft tissue resolution of MRI [48].

15 CURRENT AND FUTURE TRENDS IN PET 51 Fused PET/MRI data was also used extensively in a wide variety of clinical neurological applications including cerebro-vascular disorders, brain trauma, stroke, epilepsy, dementia [21], Parkinson and Alzheimer diseases [49], brain tumours, and in mental disorders such as depression, schizophrenia and obsessive-compulsive disorders [4]. 5.3 PET imaging with gamma camera A further trend in dual modality imaging instrumentation is to use adapted dualhead gamma camera systems not only for scintigraphy and SPECT but also to acquire PET images by removing the collimator and operating in coincidence mode. There are also systems that are able to provide CT images by adding an X-ray tube [50]. Obviously, this flexibility is obtained at the expense of a compromise between the performances of the device for each modality. The main motivation for developing such systems is mostly economic since dedicated PET scanners are quite expensive and some clinical situations do not really require the high-quality images they provide [2]. 6. CONCLUSIONS The recent development of high performance detector crystals (the new cerium doped scintillators like LSO, GSO or LYSO) and readout technologies (various types of photomultiplicators, hybrid photodetectors, silicon p-i-n photodiodes, and avalanches photodiodes) has had an enormous impact on the PET imaging instrumentation. These novel technologies have enabled a number of detector module designs that are capable of measuring depth of interaction. By measuring depth of interaction, PET camera makers can maintain high spatial resolution with smaller detector ring diameters, simultaneously reducing cost and increasing performance. New compact PET cameras, with high resolution and optimized for specific applications, such as single organ (brain, breast, prostate) or small animal imaging are already in the market or are being in an advanced stage of development. Another significant step forward was made by designing hybrid imaging systems like PET/CT and PET/MRI capable to provide both anatomical and biological information simultaneously during a single examination. On estimate that dual-modality or even multi-modality imaging techniques will be in the next years the object of significant improving, catalyzed by the multiple advantages that such high performance devices could offer both physicians and biomedical researchers. REFERENCES 1. R. Nutt, Mol Imaging Biol. 4, 11 (2002). 2. M.I. Lopes, V. Chepel, Radiation Physics and Chemistry 71, 683 (2004). 3. W.W. Moses, Nucl. Instr. Meth. A-471, 209 (2001). 4. H. Zaidi and M.L. Montandon, Current Medical Imaging Reviews 2, 3-13 (2006). 5. H. Herzog, Radiation Physics and Chemistry (available online at 6. C. S. Levin, Eur. J. Nucl. Med. Mol. Imaging 32, S325 (2005). 7. Z. Szabo et al., Semin. Nucl. Med. 36, 36 (2006).

16 52 D. Mihăilescu, C. Borcia 8. W.W. Moses, S.E. Derenzo and T.F. Budinger, Nucl. Instr. Meth. A-353, 189 (1994). 9. A. Saoudi et al., IEEE Trans. Nucl. Sci. NS-46, 462 (1999). 10. P.K. Marsden, Nucl. Instr. Meth.A-513, 1 (2003). 11. S.E. Derenzo et al., Nucl. Instr. Meth.A-505, 111 (2003). 12. C.W.E. van Eijk, Pys. Med. Biol. 47 R85 (2002). 13. V. Fidler, Radiol. Oncol. 34(4) 381 (2000). 14. P. Derenbos, Nucl. Instr. Meth.A-486, 209 (2002). 15. A. Del Guerra et al., Nucl. Instr. Meth. A-442, 18 (2000). 16. J.L. Humm, A. Rosenfeld and A. Del Guerra, Eur. J. Nucl. Med. Mol. Imaging 30, 1574 (2003). 17. D. Renker, Nucl. Instr. Meth. A-486, 164 (2002). 18. C. Joram, Nucl. Phys. B 78, 407 (1999). 19. P. Weilhammer, Nucl. Instr. Meth. A-446, 289 (2000). 20. C. D Ambrosio and H. Leutz, Nucl. Instr. Meth. A-501, 463 (2003). 21. W. Jagust, Alzheimer s & Dementia 2, 36 (2006). 22. T.R. Henry and R.L. Van Heertum, Semin. Nucl. Med. Vol XXXIII, No 2, 88 (2003). 23. G.K. Frank and W.H. Kaye, Nuclear Medicine and Biology 32, 755 (2005). 24. K. Murthy et al, Radiology 215, 280 (2000). 25. N.K. Doshi et al., Medical Physics 27, 1535 (2000). 26. R. Freifelder and J.S. Karp, Physics in Medicine and Biology 42, 2463 (1997). 27. W.W. Moses et al., J. Nucl. Med. 36, 69P (1995). 28. F. Giammarile and A. Bremond, Nucl. Instr. Meth.A-527, 83 (2004). 29. W.W. Moses, Nucl. Instr. Meth.A-525, 249 (2004). 30. E.J. Hoffman, M.E. Phelps and S.C. Huang, J. Nucl. Med. 24, 245 (1983). 31. J.S. Karp et al., J. Nucl. Med. 44, 1340 (2003). 32. T.J. Spinks et al., Phys. Med. Biol. 37, 1637 (1992). 33. Z.H. Cho et al., IEEE Trans. Nucl. Sci. 37, 842 (1990). 34. Z.H. Cho, S.C. Juh, IEEE Trans. Nucl. Sci (1991). 35. A. Braem et al., Phys. Med. Biol. 49, 2547 (2004). 36. A. Dragone et al., IEEE Trans. Nucl. Sci. 52 (2005). 37. J. Seguinot et al., Novel geometrical concept of high performance brain PET scanner: Principle, design and performance estimates, CERN Document Server -CERN PH-EP , available at A. Braem et al., Nucl Instr Methods A 525, 268 (2004). 39. A. F. Chatziioannou, S. R. Cherry, Y. P. Shao, et al., J. Nucl. Med. 40, 1164 (1999). 40. A. Del Guerra et al., IEEE Trans. Nucl. Sci. NS-47, 1537 (2000). 41. A. P. Jeavons, J. Nucl. Med. 41, 1442 (2000). 42. S. Siegel et al., IEEE Trans. Nucl. Sci. NS-46, 571 (1999). 43. D. Lapointe et al., J. Nucl. Med. 40, 876 (1999). 44. S.R. Cherry et al., IEEE Trans. Nucl. Sci. NS-44, 1161 (1997). 45. S.R. Cherry et al., in Quantitative Functional Brain Imaging with Positron Emission Tomography, Academic Press, San Diego, CA, 3 (1998). 46. C. Cohade and R.L. Wahl, Semin. Nucl. Med. 33, 228 (2003). 47. S. Senan, D. De Ruysscher, Critical Reviews in Oncology/Hematology 56, 345 (2005). 48. M.D. Seemann, Technol. Cancer Res. Treat. 4, 577 (2005). 49. L. Nichols et al., Biol Psychiatry 59, 940 (2006). 50. J.A. Patton, D. Delbeke, M.P. Sandler, J. Nucl. Med. 41, 1364 (2000).