PET and SPECT are both emission techniques, and the artifacts discussed here are specific to emission techniques. In an emission scan, the photons traveling to the detectors are attenuated more strongly when they are located more centrally in a transaxial section of a patient than when they are located more superficially. The attenuation decreases exponentially with the distance traveled in the tissues (see Chapter 3). Due to differences in the detection technologies, these attenuation artifacts differ somewhat in PET and
SPECT. The result, however, is basically the same: if the images are not corrected for body attenuation, central foci with the same activity appear to be less active than superficial foci, and the foci are distorted (4). This has to be taken into consideration when scans not corrected for attenuation are clinically interpreted.

To eliminate attenuation effects, the PET and SPECT emission images have to be corrected using an attenuation map. This map is obtained either by gamma radiation–emitting sources in PET or SPECT systems rotated around the patient as a second part of the examination. In PET-CT and SPECT-CT systems, the map is obtained by the CT scanner. With the advent of PET-CT, the acquisition of transmission scans to correct for differences in attenuation has become much less time consuming; thus PET images from a PET-CT scanner are nowadays rarely viewed without attenuation correction. The only exception is when body regions containing metallic implants are evaluated (see “Attenuation Artifacts due to Beam Hardening” later). If a PET-only scanner is used for partial- or whole-body scanning, issues of patient throughput may prevent performances of one from carrying out a transmission scan, as it takes about 30% of the entire imaging time. Although not optimal, it may still be appropriate to view non-attenuation-corrected scans. In SPECT, attenuation correction was introduced successfully into myocardial perfusion scanning a few years ago, but the standard in SPECT imaging for other purposes is still to use uncorrected scans. This may change soon, as various manufacturers are now starting to offer SPECT-CT systems, in which acquisition of a attenuation scan is much faster than with conventional SPECT systems.

In SPECT, collimation is done with lead collimators. In this detection process, activity on the surface of the patient is seen more prominently, because there is less body tissue between the radiation source and the detector system to attenuate the emitted gamma rays (Fig. 33.1A). Furthermore, the spatial resolution decreases with increasing distance from the detector; therefore, the radiation detected by an opposing detector in a dual-detector gamma camera will not contribute much to the image. As a result, for the same activity, superficial lesions appear to be of higher activity and less blurred than deep foci. In PET, the situation is somewhat different. Each gamma ray is subject to the same attenuation as in SPECT, but the coincidence detection mechanism detects two gamma rays, both of which are attenuated differently because the path of the two gamma rays through the body differs (Fig. 33.1A). If the sum distance of the length in the body through which both opposing gamma rays travel is equal, for two decay events located in different body regions, as in the example shown, the net attenuation is the same independent of the location of the radiating focus along the line connecting the two coincidence-detecting detectors, and thus the attenuation is independent of the location of the activity focus. The difference in attenuation of gamma rays from peripheral versus deep foci in PET is due to the fact that more rays are attenuated slightly or not at all for a peripheral focus than for a deep focus (Fig. 33.1B). Unlike SPECT, PET has a better central than peripheral spatial resolution (see Chapter 3). Artifacts due to lack of attenuation are seen in Figure 33.1C.

These types of artifacts are not relevant for CT, as attenuation is in fact the physical principle by which CT images (or x-ray images in general) have contrast in them.

PET and SPECT data typically are based on low photon flux, and the images are noisy. When such images are reconstructed with projection reconstruction algorithms, streaks cross the axial section FOV, particularly when there are great differences in contrast in the image. They are noted only when the lower threshold of image display is set very low. Modern PET and SPECT scans are mostly reconstructed using iterative algorithms, which do not lead to these artifacts (see Chapter 3).

As discussed in Chapter 3, each real imaging system is not able to depict a point source as a point, but the signal is spread to give rise to the so-called point spread function. There are two situations that are useful to distinguish. In the first situation, the focus is as large as or larger than a single voxel. Particularly when the activity or signal in a focus is much higher than that of the surroundings, some of the activity “spills over.” In the images, activity thus is also seen in neighboring voxels, and the lesion appears larger than it actually is. This effect is frequently noted in PET scans with the ureters, which are typically thin and less than 5 mm in diameter but can contain very high activity (Fig. 33.2E–G). In SPECT, this effect is frequently seen in thyroid SPECT scans of remnant thyroid tissue, where again the activity noted on the SPECT scan extends beyond the morphological confines of the activity focus.

The second situation occurs when the lesion is smaller than the voxel or when the voxel is located at the edge of a lesion. Partial-volume effects occur in all tomographic examinations but have different linear scales, as the spatial resolution of PET is on the order of 5 mm, the spatial resolution of SPECT is on the order of 15 mm, and in the anatomic imaging modalities it may be less than 1 mm. In PET and SPECT, many relevant lesions seen in the body, such as lymph nodes, may be smaller than a single voxel, and they might not be detected due to limited spatial resolution. Nevertheless, in both PET and SPECT, smaller lesions can be detected if their specific activity is high (this is also true for the anatomic imaging modalities, but to a lesser extent, because contrast differences are typically not as high as they are in PET and SPECT). The point spread function smears a focus out over the entire voxel. As an example, let us assume, first, that a voxel in a PET examination has linear dimensions of 10 mm and thus a volume of 1,000 mm³; second, that the minimal activity a focus needs to have to become detectable is two times higher than its surroundings; and third, that the
background activity has a value of 1 (arbitrary unit). If a focus has an activity eight times higher than the background, a focus that is only 5 mm in diameter and has a volume of 125 mm³ will still generate an apparent activity that is 8 × 125/1,000 = 1 above the background activity. The voxel containing that focus will then have an activity twice that of the background, which is readily detectable. Hence, a lesion with a volume only 1/8 of the voxel can be detected when its activity is eight times that of the background. The activity of the focus, however, is artificially reduced to 2. In general, partial-volume effects make very small lesions undetectable, but lesions somewhat smaller than the voxel can still be detected, although they will have an artificially lowered activity. In PET, this effect is particularly relevant in the case of small pulmonary metastases, which either may escape PET detection because of their size or appear to have a low SUV despite the fact that the primary tumor has a high SUV. Analogous arguments can be made for the detection of a focus in CT, but scaled to the resolution of CT.

Beam hardening occurs whenever polychromatic rays are traveling through dense material. In this situation, only the higher-energy photons are transmitted, while the lower-energy photons are absorbed. The spectrum of PET and SPECT gamma rays is close to monochromatic, whereas in CT polychromatic x-rays are used; thus beam hardening does not affect PET and SPECT images much but can severely deteriorate CT images. In CT, images affected by beam hardening show areas of higher and lower density, which is a result of CT spatial resolution insufficient to depict the interface of the high-density and low-density material and the lack of lower-energy photons, which have
been filtered out by the high-density material in the FOV (Figs. 33.3B and 34.15A). Furthermore, high- or low-density streaks, mostly in the left to right patient axis, can be noted even distant from the beam-hardening object in the image (Fig. 33.24D,F). These artifacts are noted frequently at the skull base and in the shoulder region, where the horizontal plane the bony shoulder girdle presents a substantial bulk of high-density absorption material. Any body portion containing a metallic implant, most notably hip prostheses and metallic dental work, will also generate beam-hardening artifacts. With low photon flux, these artifacts tend to be more prominent. As in PET-CT and SPECT-CT, frequently low-dose CT images are acquired, and the issue of image quality versus diagnostic necessity has to be carefully weighed.

As stated, there are many other technical artifacts that can occur. They are related to partial loss of detector function, misalignment of the rotation center of the system, and many other causes. They are not discussed here, as they do not interfere with image interpretation when the system functions properly.

Strong CT beam attenuation by metallic implants will affect the quality of PET or SPECT emission images reconstructed using CT attenuation maps. Thus such attenuation affects not only CT images but also images of the integrated systems.

The most relevant artifacts occurring as a result of the integration of PET or SPECT and CT are due to a limited axial FOV in CT or to erroneous attenuation values introduced from CT attenuation maps into attenuation-corrected PET and SPECT images. Some of the basic issues concerning these artifacts are covered in Chapters 7 and 8. Here, the clinical implications are emphasized.