The following parameters described in Table 1-1 determine data acquisition:

The number of active detector rows or simultaneously acquired sections is one for single detector-row CT, two for dual detector-row CT, and 4 to (presently) 128 for multidetector- row CT. Section collimation is determined by the total width of the acquired volume in the center of the scan field divided by the number of sections. The rotation time is the time it takes for the tube–detector unit to rotate once around the patient.

Table feed per rotation and pitch are parameters that are relevant for spiral scanning: The table feed indicates how much the table moves during the rotation time. The pitch factor provides the relation between table feed and total width of the acquired volume. The maximum pitch depends on the scanner manufacturer and usually varies between 1.5 and 2. For sequential scanning, the table feed between consecutive sections is usually chosen identical to the section collimation, and the resulting pitch is 1.

Scan length depends on the clinical indication. Scan duration can be derived from scan length and table speed, and thus the other parameters as described in Table 1-1.

Tube load, pitch, and tube voltage are main determinants of patient dose. Tube load, also called the mAs setting, is proportional to the radiation dose, while pitch is inversely proportional to it. This has led to the definition of effective tube load (or effective mAs): mAs_eff = mAs/P. Tube voltage (kVp setting) determines the energy of the x-ray beam: At higher energy, there is more penetrating power. However, the x-ray attenuation of elements with a higher atomic number than water (for example, calcium or iodine) will be relatively higher at lower x-ray energy. At constant mAs, the dose decreases if lower kVp settings are chosen.

Image reconstruction is influenced by the following parameters (Table 1-1):

Image reconstruction starts with choice of a reconstruction field of view that determines which part of the data will end up in the images. The FOV is usually chosen from reference images that include the whole body cross sections and allows focusing on the region of interest. The matrix size of the reconstructed image usually is 512², but modern scanners allow for other matrices as well (768² or 1,024²).

The reconstruction filter is required to reconstruct usable images from the projectional raw data. It determines the trade-off between spatial resolution in the imaging plane (xy-plane) and the image noise. The raw data interpolation
algorithm is necessary in spiral scanning to compensate for the continuous motion during the scan, but more sophisticated raw data interpolation algorithms are also used to compensate for cone beam effects in multidetector-row CT (p. 20).

Reconstructed section width or section thickness refers to the thickness of the CT sections in the final images. In sequential scanning, section width and section collimation are identical. With spiral scanning, section width is commonly larger than section collimation (p. 13). With multidetector-row CT, section width can be varied almost arbitrarily but is always larger than or equal to the chosen section collimation (p. 14). Section collimation therefore determines the thinnest section width that can be reconstructed.

The reconstruction increment is a parameter for spiral scanning: Here, images can be reconstructed at arbitrary increments that are smaller than the chosen section width or section collimation. In sequential scanning, the reconstruction increment is identical to the table feed between sections.

The contrast-to-noise ratio (CNR) determines how well a given vessel can be displayed. At a high CNR, even small vessels can be detected, and vessel contours on 3D displays appear well defined (Fig. 1-4A). At a low CNR, small vessels may no longer be sufficiently visualized to make a diagnosis, and 3D displays of larger vessels may be irregular with insufficiently defined contours (Fig. 1-4B).

CNR is determined by the difference in CT numbers between vessel of interest and surrounding soft tissues in relation to the image noise:

CNR = (CT_vessel – CT_{soft tissue})/Noise

Spatial resolution determines how small a detail can be evaluated and how sharp a given structure is depicted. Spatial resolution in a CT image depends on matrix size and system resolution. System resolution is determined by the scanner geometry and the chosen acquisition parameters. If the matrix is fine enough (small enough pixels, or picture elements), the spatial resolution in the image equals the system resolution. Commonly, the spatial resolution of an imaging system is described in one of the following three ways.

Temporal resolution is described by the temporal acquisition window, the time period from which data are sampled to reconstruct an image. If raw data are weighted before reconstruction, the temporal resolution is frequently defined in analogy to spatial resolution as the full width at half maximum of the temporal weighting function.

Single detector-row scanners use a single detector arc or detector ring that consists of parallel detector elements that are much smaller in the direction of the arc (xy-plane) than in the z-direction (Fig. 1-2A). In fact, the detector elements are wide enough in the z-direction to accommodate the maximum section collimation (usually 10 mm) that is available on such scanners.

Dual detector-row scanners are based on a detector array that is twice as wide as a conventional CT detector and is split in half (Fig. 1-2B). These scanners became available in early 1992, but in 1998, the introduction of 4-detector-row scanners brought a technological breakthrough.

In multidetector-row CT, each element on the detector array is subdivided along the z-axis (Fig. 1-2C). These parallel detector rows can be electronically combined to yield between 4 and 128 separate sections per rotation, depending on the type of scanner (1,2,3). With the exception of most 64- and 128-detector-row scanners, the actual number of detector rows is much larger than the number of active detector rows (channels) in order to accomplish more than one collimation setting. Different collimation settings are achieved by adding the signals of neighboring detector rows.

While there were three distinct types of detector arrays with 4- to 8-detector-row scanners—namely matrix arrays,
hybrid arrays, and adaptive arrays (Fig. 1-15)—10- to 16-detector-row scanners use only hybrid arrays, and 32- to 64-detector-row scanners use hybrid or matrix arrays. Matrix arrays consist of detector rows of identical thickness; hybrid arrays have a central group of detector rows whose thickness is half the thickness of the outer detector rows; while adaptive arrays have detector rows that get wider the more peripherally they are located.

All 4- to 16-detector-row scanners provide multiple collimator settings with the maximum number of detector rows (Table 1-3). For hybrid and matrix arrays, the available section collimations are multiples of the thinnest detector width (Fig. 1-16A–B). For adaptive array detectors, the available section collimations are gained by combining signals from one, two, or more parallel rows with appropriate collimator settings (Fig. 1-16C). Most 64- and 128-detector-row scanners, however,
do no longer provide the maximum number of sections for the largest available collimation (Fig. 1-17A–B). These scanners use matrix detector arrays of between 64 × 0.5 mm (128 × 0.5 mm) and 64 × 0.625 mm width. In these arrays, all 64 to 128 detector rows are read out simultaneously. As a consequence, only half the number of sections are available if the next wider collimation settings are chosen.

The only exception to this rule is a scanner that uses 32 active detector rows and z-flying focal spot technology (Fig. 1-17C). This scanner has a hybrid array that allows for section collimation of 32 × 0.6 mm and 20 × 1.2 mm. While this scanner actually provides a maximum of 32 sections per rotation (in the nonspiral, or sequence, mode), the manufacturer has opted for calling it a 64-slice scanner because the z-flying focal spot technology effectively doubles the data points (and thus the simultaneously acquired slices) along the patient axis without having to increase the number of detector rows.

Flying focal spot (FFS) technology refers to the position of the focal spot of the x-ray tube. The focal spot varies very rapidly between various positions on the anode, thus creating slightly different views of the object despite an (almost) identical position of the detector. The technology can be used to increase the number of x-ray projections for image reconstruction and, that way, allows for improving the spatial resolution of a scanner without creating “aliasing artifacts.” The technology is not new. It has been used for years to increase the number of projections within the imaging plane (xy-plane).

Recently, an additional movement along the z-axis (z-FFS) was introduced that allows for increasing the number of projections also in the z-direction. The idea behind the technology is as follows. For image reconstruction, it is irrelevant whether projections from the tube to the detector or in the opposite direction, from the detector to the tube, are considered. This knowledge has been used for more
than a decade for 180-degree interpolation algorithms in spiral CT, in which projection data from a conjugated spiral trajectory (detector to tube) are used for interpolation. Similarly, it can also be used for z-flying focal spot techniques (Fig. 1-18A): The linear trajectories from the tube to the centers of adjacent detector elements are identical to the trajectories from a single detector to multiple positions of the focal spot. However, there is a marked difference in the actual regions within the scanned object that are “seen” along each trajectory using these two techniques. In case of separate detector rows, these regions do not overlap, while they markedly overlap in the case of multiple focal spots (Fig. 1-18B). Currently, z-FFS technology is implemented in a way that two overlapping sections are irradiated, which are exactly 50% offset in the scan center. These sections converge (i.e., become identical) toward the detector and diverge (i.e., become separate) toward the tube with its flying focal spot (Fig. 1-18C). One has to keep in mind that despite the fact that the projections overlap in the center of rotation (by 0.3 mm), the thinnest available section width that can be reconstructed with z-FFS technology remains identical to the collimation (i.e., 0.6 mm).

Because the z-flying focal spot technology provides twice the data points in the z-direction over conventional
technology with a single focal spot, aliasing artifacts in the z-direction are markedly reduced, even at higher pitch values (4). This is especially apparent for CTA images of the skull base and posterior fossa. If the number of projections is increased with conventional technology, however, similar artifact suppression can be reached. This requires low pitch factors or increasing the reconstructed section width by at least 30% over the chosen collimation (Fig. 1-19). For this reason, z-FFS yields the biggest advantages whenever very thin sections have to be reconstructed or high pitch factors have to be used.

Finally, it has to be noted that the number of focal spots plays no role for the calculation of pitch factors or other crucial scanner characteristics. A scanner with 32 active detector rows and 2 focal spots will therefore behave like a 32-detector-row scanner and not like a 64-detector-row scanner when it comes to pitch, scanning speed, or even detector width. To mark this difference, it might be useful to distinguish between detector rows on the one hand and slices on the other: A 64-slice scanner that employs two z-flying focal spots should therefore correctly be addressed as a 32-detector-row scanner with dual z-FFS technology. The technique holds a lot of promise, and in the future, tubes with multiple focal spots can be expected.

Multiple source technology refers to scanners with more than one tube–detector array combination (5). They are primarily being developed for cardiac CT. Presently, dual source scanners are becoming available that have the advantage that the time required to acquire data for one image is
cut in half by the two acquisition systems. Another potential advantage is the availability of a dual energy technique: The two tubes can operate at different kVp settings, and the differences in x-ray absorption at identical projection angles can be used to separately display materials of differing atomic number, such as calcium, iodine, or soft tissues. It remains to be seen whether the ultimate goal of reducing the negative influence of vessel wall calcifications in coronary CTA can be reached.

In dual source scanners, one of the detector arrays has a normal size (in xy-direction) to be able to perform standard CT examinations. The second detector array is narrower because it only is used for cardiac examinations (Fig. 1-20). The effective rotation time is reduced from 330 ms for a single source scanner to 165 ms for a dual source scanner. Maximum temporal resolution with these scanners is below 90 ms.

The most important factor that determines scanner performance for CTA is the relation between scanning speed and z-resolution. This performance parameter can be defined as the ratio between maximum table speed and thinnest section width that can be reconstructed using that table speed. Scanner performance, when defined that way, has been doubling approximately every 2 years since the mid 1980s (Fig. 1-21). As compared to 4-detector-row scanners, performance of 32- to 64-detector-row scanners has increased more than 20 times owing to more detector rows and faster rotation speed.

This increase in performance made it possible to acquire isotropic data and even reduce scan duration with multidetector-row scanners. While spiral CT scanning with single detector-row scanners made it possible to perform thoracic or abdominal CTA during a single breath-hold
phase of approximately 30s, the advent of 4-detector-row scanners made it possible to acquire the data in a near isotropic fashion. With 16-detector-row scanners, near-isotropic resolution could be combined with sub-mm resolution and scan durations of between 8 and 15s for chest or abdomen (Table 1-6) (6). With 32- to 128-detector-row scanners, scan duration can always be kept well below 10s, even for submillimeter resolution. However, this decrease in scan duration may not always be advantageous: Especially for large aneurysms or vessel occlusions, it takes longer to completely enhance the vessel of interest, and scanning too early or too fast will lead to suboptimum contrast enhancement (see also Fig. 1-31). For this reason, 16-detector-row units suffice for providing excellent spatial resolution and good handling for the vast majority of CTA examinations.

Further increase in performance is only desirable for the heart or other applications in which electrocardiogram (ECG) synchronization of data acquisition or reconstruction is required. This is where the present advantages of 32- to 128-detector-row scanners lie. Further improvement in effective rotation time and thus, temporal resolution, is especially important for coronary CTA. For CT perfusion imaging, wider detector arrows would be advantageous because whole organ systems could then be imaged within one rotation.

In sequential scanning, the table remains stationary during acquisition of a single section. The reconstructed section width is identical to the section collimation. The table feed between sections usually is identical to the section collimation. Sequential scanning is not used for CTA.

In general, data from a 360-degree rotation are sampled to reconstruct an image. However, data from multiple rotations can be combined in order to increase dose, for example, for brain scans. In order to increase temporal resolution, so-called partial scans can be performed, in which the minimum amount of projections are used for reconstructing an image: For each pixel in the final image, a 360-degree circle of evenly spaced projectional rays are required. Since direction of the rays plays no role, projectional data from 180 degrees have to be available. For pixels in the center of the scan field, projections from a 180-degree rotation suffice. Outside the center, more data have to be sampled: This minimum amount of data consists of a rotation that covers 180 degrees plus the fan angle of the x-ray beam or, more precisely, the fan angle that just covers the field of view (Fig. 1-22).

These data or reconstruction techniques are also used for dynamic scanning. The basic principles of data reconstruction, however, also hold true for the spiral data acquisition described below.

Spiral scanning is the basis for CTA. The patient is moved through the scan field during continuous acquisition of raw data. To be able to reconstruct stationary images from such a scan, motion correction is necessary. Basic raw data interpolation techniques are used for single detector-row CT, while more complex algorithms are required for multidetector-row CT. With spiral scanning, a CT image can be generated from any segment within the scanned volume, so the table feed is unrelated to the site of image reconstruction. Sectional images can be produced at arbitrary table positions, and individual images can be overlapped as desired without increasing radiation exposure. The spacing between the reconstructed sections is called the reconstruction increment, interval, or index.