The ubiquitous presence of carbon makes carbon-11 an attractive and important positron-emitting isotope. However, it is beyond the scope of this chapter to describe extensively carbon-11 radiochemistry and for review the reader is referred to Allard et al. (2). The most commonly applied production method to obtain carbon-11 is the ¹⁴N(p,α)¹¹C nuclear reaction. If the target of the cyclotron is filled with a nitrogen/oxygen gas mixture, [¹¹C]-carbon dioxide is produced (blue box, Fig. 3), whereas when a nitrogen/hydrogen gas mixture is present, [¹¹C]-methane is recovered (green box, Fig. 3). [¹¹C]-carbon dioxide and [¹¹C]-methane are referred to as primary precursors. As [¹¹C]-methane is unable to react directly with the labelling precursor, subsequent rapid reactions are needed to produce more reactive molecules, referred to as the secondary precursors. These react with the labelling precursor yielding the final radiotracer. Even though [¹¹C]-carbon dioxide has very reactive properties, it can be transformed in secondary precursors for some chemical reactions, e.g. the methylation and carbonylation reaction (Fig. 3).

[¹¹C]-carbon dioxide as the primary precursor:

[¹¹C]-methane as the primary precursor:

An important difference between carbon-11 and fluorine-18 radiochemistry follows from the limited chemical methods that are available to incorporate fluorine-18 into the target molecule, which is in contrast to carbon-11 radiolabelling. It is beyond the scope of this chapter to describe extensively fluorine-18 radiochemistry and for a review of this domain the reader is referred to the recent special issues 2 and 3 of the volume 3, edited in 2010 by Current Radiopharmaceuticals (16–18).

Briefly, two main reactions are used: electrophilic and nucleophilic radiofluorination. The first method uses molecular [¹⁸F]-fluorine ([¹⁸F]-F₂) generally obtained via the ²⁰Ne(d,α)¹⁸F nuclear reaction or more recently via the ¹⁸O(p,n)¹⁸F reaction, where the cyclotron target is filled with ¹⁸O gas. These two nuclear reactions involve an exchange step between carrier non-radioactive fluorine and [¹⁸F]-fluorine, inducing a poor specific radioactivity. The second radiofluorination uses the anion [¹⁸F]-fluoride. This method is more selective and provides significantly higher yields and specific radioactivities than the electrophilic radiofluorination. Therefore, nucleophilic radiofluorination is often preferred, particularly for brain imaging studies, where high-specific radioactivities are required.

Electrophilic radiofluorination. [¹⁸F]-fluorine is not the only electrophilic-labelling reagent used for this reaction. It is sometimes suitable to involve less-reactive fluorinating reagents, such as trifluoromethyl-[¹⁸F]-hypofluorite (CF₃O[¹⁸F]-F) (19) or acetyl-[¹⁸F]-hypofluorite (CH₃CO₂[¹⁸F]-F) (20). The labelling precursor, consisting of a nucleophile entity such as an alkene, an enol or an organostannyl compound, is then radiolabelled with one of these [¹⁸F] electrophilic reagents to produce the [¹⁸F]-radiotracer. Starting from [¹⁸F]-fluorine and the 6-trimethylstannyl-labelling precursor, the 6-[¹⁸F]-fluoro-l-DOPA can be obtained using a fully automated computer-controlled module (TRACERlab FX_FDOPA™, GE Medical Systems) (21).

Nucleophilic radiofluorination. As previously mentioned, this method is the most efficient and allows significantly higher yields and specific radioactivity than the electrophilic radiofluorination reaction. The anion [¹⁸F]-fluoride, recovered after an ¹⁸O(p,n)¹⁸F nuclear reaction, is highly soluble in liquid [¹⁸O]-water, which served as a the target for the nuclear reaction. However, the [¹⁸F]-fluoride is a strong base, but a poor nucleophilic reagent. The most efficient and commonly used procedure to increase the nucleophilic force of [¹⁸F]-fluoride consists of fixing [¹⁸F]-fluoride on an anion exchange resin before eluting it with a diluted aqueous potassium carbonate solution. This creates a chelation between the [¹⁸F]-fluoride anion and the potassium cation. This chelation is then weakened using a kryptand, the polyaminoether Kryptofix-222^® (4,7,13,16,21,24-hexaoxa-1,10-diazabicyclo[8.8.8]hexacosane). After concentration to dryness, the complex (K-[¹⁸F]-F-K₂₂₂) is often further dried by azeotropic evaporation with acetonitrile. The reactive nucleophile [¹⁸F]-fluoride reagent is then dissolved in polar solvents, such as DMSO, DMF, acetonitrile or bulky alcohols (22). Typical reaction conditions involve high reaction temperatures, between 80 and 200°C, and even microwave irradiations. In general, the reaction time lasts a few minutes to half an hour. Depending on the target molecule, being an aliphatic or aromatic entity, the nucleophilic radiofluorination occurs by direct or indirect nucleophilic fluorination.

Direct nucleophilic fluorination (left side, Fig. 4) is the method of choice to radiolabel a target molecule with a fluorine-18 because it consists of a one-step process by direct reaction of the radioactive precursor, [¹⁸F]-fluoride, with the labelling precursor. Within direct fluorination, short reaction steps (one to two reactions) can be considered as well, e.g. to remove a protective or to remove/transform a strong electron-withdrawing group.

Indirect nucleophilic fluorination (right side, Fig. 4) is used when direct nucleophilic fluorination is not feasible due to the relatively harsh reaction conditions, such as high pH and high reaction temperatures, not supported by many labelling precursors.

The increasing interest of macromolecules (green box, Fig. 4) for treatment and diagnosis of disease has contributed to the development of [¹⁸F]-fluorine radiolabelling of oligonucleotides, peptides, proteins and antibodies. Many other [¹⁸F]-prosthetic groups reacting in a chemoselective way with the macromolecule have, thus, been designed. Two of these most commonly used prosthetic groups are the N-succinimidyl-4-[¹⁸F]-fluorobenzoate ([¹⁸F]-SFB) (30) and the maleimide 1-[3-(2-[¹⁸F]-fluoropyridin-3-yloxy)propyl]pyrrole-2,5-dione ([¹⁸F]-FPyME) (31) which react selectively with thiol group of cysteine residues.

The vast majority of annihilation events between positions and surrounding electrons, give rise to the emission of two 511keV photons that are emitted in almost exactly opposite directions (180°), albeit with a directional uncertainty of about  ±  0.5° (Fig. 5). This physical particularity yields some uncertainty in the exact determination of the site of annihilation, thereby limiting the spatial resolution of the technique.

The distance that the positron travels prior to annihilation (free path) is dependent on the initial emission energy and the medium through which it is travelling (Table 1). Depending on the isotope, the mean distance travelled by the positron before annihilation varies, affecting the actual spatial resolution in the images. As a consequence, even if the point of annihilation is very close to the point of emission, the image generated from the distribution of annihilation radiation only approximates the actual distribution of the radionuclide itself.

The detection of the positron emission is achieved by the simultaneous (coincident) detection of the two collinear 511 keV photons resulting from the annihilation process by two opposite detectors surrounding the positron-emitting object. Surrounding the subject’s head by a large number of detectors (up to 20,000 for the most recent tomographs) permits a multi-coincidence detection and a spatial reconstruction of the various annihilation events (Fig. 6).

Typically, in a circular set-up, each detector is placed in coincidence with about half of the total detectors in the ring (Fig. 6a, b), whereas in a polygonal array set-up the detector is in coincidence with the opposed detector bank (Fig. 6c). PET systems with only partial detector rings are less expensive, but require rotation of the detector assembly round the longitudinal axis of the patient to complete the acquisition of the projection data (Fig. 6d). Continuous, large-area detectors, such as those found in multi-head Anger camera systems and used for single-photon emission-computed tomography (SPECT), have now been appropriately modified and are as well used as for coincidence imaging of positron emitters (Fig. 6e).

Each coincidence event, i.e. simultaneous detection of a ­photon pair within a time window of a few nanoseconds, represents a line in space, the Line of Response (LOR), connecting the two detectors along which the annihilation process occurred. The raw data collected by a PET scanner are, thus a list of “coincidence events” ­representing near-simultaneous detection of annihilation photons by a pair of opposite detectors. It takes approximately 3 ns for ­photons to traverse the field of view of a PET scanner, and in addition to measure the position of arriving annihilation photons, the detectors placed in coincidence also measure their time of arrival. Each detected photon is, thus, tagged with a detector position and a detection time: if the detection time difference between two photons is smaller than the coincidence window (traditionally 5–10 ns), the two events are considered to be physically correlated to the same annihilation event (Fig. 7). Time-of-flight PET uses this time-of-flight difference between the detection of the two photons to better locate the annihilation position of the emitted positron along the LOR (for review, see Conti et al. (32)).

Besides true coincidence events arising from single positron annihilation followed by detection of the photon pair without any precedent interaction of the photon pair with the medium (solid line in Fig. 8a), two other major events might cause coincidences. Scattered coincidence (solid line in Fig. 8c) results from a single annihilation with one of the photons (dashed line) undergoing a change in direction (due to the Compton scatter effect) in the medium before detection. Though a true incidence, the LOR is displaced and thus degrading image contrast and resolution. In addition to true coincidences, random or accidental coincidence events (Fig. 8b) occur when two photons not arising from the same annihilation event (dashed lines) are detected by two opposite detectors within the coincidence time window of the system. These falsely detected coincidences (full lines) are uniformly ­distributed in time and will cause isotope concentrations to be overestimated if not corrected for. Photons are inherent to attenuation as well, reducing the number of coincidences and thus underestimating the amount of radioligand within a region of interest, if not corrected for. Although the magnitude of this correction for small animal subjects is much smaller than for humans (1.3 for a 3-cm diameter set-up in mouse versus 1.6 for a 5-cm diameter in rat versus 45 for a 40-cm diameter in human), it is important to correct the data for attenuation in order to obtain accurate quantitative images of the tracer distribution (33).

In conclusion, in order to achieve quantitative images, the acquired data need to be corrected for random, scattered and attenuation effects during the image reconstruction process. Random coincidences are usually subtracted by the delayed coincidence method while scatter subtraction is based on model calculation. Correction for photon attenuation is by far the most important correction, especially on large animal brains.

A significant fraction of the photons arising from annihilation events fail to reach the detectors due to their auto-attenuation in the body. In addition, at the surface of the body, the mean ­attenuation is less, as the LOR passes mostly through air. These differences in attenuation will result in distortions in the reconstructed image if not corrected for. If the anatomical properties of the subject/object are known, the measurement along each LOR can be corrected for the attenuation effect. Several methods of correcting for attenuation in PET exists: the material properties of the object/subject can be obtained using an external rotating pin ­single photon-emitting source, e.g. [⁶⁸Ge] or [⁵⁷Co], or by using an X-ray-computed tomography scanner (for more information, the reader is referred to Fahey et al. (33)). In this case, the attenuation coefficients are scaled to the appropriate energy (511 keV) using tabulated values. The attenuation map is then projected along each LOR to give the attenuation coefficient of the coincidences.

Detectors for annihilation radiation should have good stopping power for 511 keV photons and the ability to resolve the time of arrival of incident photons within a few nanoseconds at most. Additionally, because these detectors are configured without collimators, they must be capable of operating in high photon flux environments. These extreme conditions have led to the development of specialized detectors. The most important practical features of scintillation detectors include a high mass density and a high effective atomic number to maximize the crystal stopping power. In addition, a fast crystal (short scintillation decay time) allows the use of a narrow coincidence window and therefore a notable reduction of the random coincidences count rate. Finally, the detector system should have a good energy resolution to reduce the fraction of detected scattered gamma rays as much as possible. Typical choices are bismuth Germinate (BGO), which has excellent stopping power, or lutetium oxyorthosilicate (LSO), which has good stopping power and excellent signal speed. Another choice is gadolinium silicate (GSO), which has intermediate stopping power but excellent signal speed and also offers improved energy resolution (Table 2).

While the first generations of PET tomographs were characterized by limited resolution and sensitivity, the latest generation of PET scanners combines a relatively high spatial resolution, a high sensitivity with a large field of view resulting in better counting statistics and improved region-of-interest definition due to decreased partial volume effects. Table 3 gives a brief overview of the main characteristics of several commercially available PET imaging systems.