3.1.1. History

Computed tomography (CT) was the first non-invasive radiological method allowing the generation of tomographic images of every part of the human body without superimposition of adjacent structures. Even more importantly, a contrast resolution could be achieved that for the first time in radiology permitted the differentiation of soft tissue inside the highly attenuating skull. The technological improvements in the first 20 years of CT can best be characterized by a comparison of typical scan parameters from 1972 and 1994, as shown in Table 3.1.

Table 3.1

Technological Improvements in the First 20 Years of Computed Tomography.

The greatest progress has been made in reducing scan times and in improving spatial resolution. Further increasing the contrast resolution will require an increase in the patient dose, since the efficiency of the detector systems is already close to its theoretical limits. An improvement in the transaxial or planar spatial resolution is sensible only if the resolution as measured along the scanner axis (i.e., the slice thickness) is improved simultaneously. With today's typical scan parameters there is a discrepancy between the planar resolution, which is less than 1 mm, and the slice thickness, which customarily is several millimeters. With the recent introduction of spiral CT (section 3.3), x-ray CT, like other imaging modalities, has started to move from a slice-by-slice to a volume imaging method, with more isotropic spatial resolution in all three dimensions allowing a better three-dimensional representation of anatomical structures.

3.1.2. Principle of Operation

The attenuation of monochromatic x-rays inhomogeneous objects is governed by

where I is the x-ray intensity behind the object, I ₀ is the x-ray intensity without the object, x is the length of the x-ray path through the object, and a is the linear attenuation coefficient of the material for the x-ray energy employed.

For inhomogeneous objects like the human body, the attenuation of x-rays consequently can be described by

This model is explored further in section 14.1.1.

In CT scanners the x-ray attenuation according to equation 3.2 is measured along a variety of lines within a plane perpendicular to the long axis of the patient with the goal of reconstructing a map of the attenuation coefficients a for this plane. The resulting attenuation coefficients are usually expressed with reference to water:

Equations 3.1 and 3.2 ignore the fact that the spectrum of the x-rays entering the patient is generally non-monochromatic. It turns out that the resulting ''beam hardening artifacts" can be corrected to a sufficient degree by a simple nonlinear correction to the measured line integrals.¹

Figure 3.1 shows the two principal measurement geometries used in today's CT scanners. In fan beam systems (Fig. 3.1, left) a multicellular linear detector array and the x-ray tube are mounted on the same carrier, which rotates around the patient. A fan-shaped x-ray beam is utilized to successively measure the x-ray attenuation across the patient for the required number of projection angles. The image then can be reconstructed either after rearranging the central projections into parallel projections with parallel beam reconstruction techniques, or more directly by using the central projections with a modified fan beam reconstruction algorithm. A variant of the fan beam system employs a number of x-ray tubes and detector arrays rotating around the patient with the x-ray focal spots shining through the gaps between the detector arrays. This configuration allows for a shorter exposure time since the system has to be rotated only through a small angle rather than 180° or more in order to obtain a complete data set for image reconstruction. However, because of the very high cost involved and because of the improved speed of slip ring or electron beam based systems, this variant of the conventional fan beam system has not gained wide distribution.

Figure 3.1.

Scanning principles of computed tomography systems. Left: Fan beam systems employ a multicellular detector system rotating about the patient together with the x-ray tube. Right: Ring detector based systems have a fixed detector ring, operate with a fan-shaped (more...)

Ring detector based systems (Fig. 3.1, right) use a complete stationary ring of detector elements surrounding the patient port and move only the x-ray tube—or, in the case of electron beam systems, the focal spot—around the patient. The measurement and reconstruction geometry also employs a fan of x-rays through the patient in the same way as with fan beam systems. However, particularly in electron beam based systems, it is often difficult to move the x-ray focal spot in the same plane that is occupied by the ring detector. In this case the focal spot has to follow orbit in a plane adjacent to the detector system, which represents an imperfection in the measurement geometry and can lead to typical artifacts (see also section 3.2.3).

3.2.1. X-Ray Tubes

The dramatic increase in the performance of conventional-i.e., non-electron-beam based-CT scanners over the past 20 years has required significant improvements in the technology of the x-ray tubes used. Today, typically, a metal (tungsten-rhenium-molybdenum) target is brazed to a thick graphite block for improved heat dissipation. Target diameters of up to 20 cm are used with rotational speeds of up to 10,000 rpm. Typically the target is operated at a temperature of 1200°C and the area around the focal spot reaches over 2500°C. The maximum heat dissipation of these tubes can reach up to 10 kW. Still, the major limitation in the use of CT scanners imposed by the x-ray tube is the maximum permissible sustained x-ray power. A further increase will require major developments in target cooling (e.g., liquid cooled-rotating targets), bearing systems, and generation of the electron beam.

3.2.2. Detector Systems

Figure 3.2 gives an overview of the two types of detector systems employed with CT. Both types of systems are usually arranged in linear arrays with up to about 1,000 detector elements or as rings with even more elements in the case of systems with a geometry as shown on the right in Figure 3.1. At least one commercial system uses two adjacent rows of detector elements in order to scan two slices simultaneously. For scintillation based detectors the most commonly used scintillator materials have been cesium iodide (CsI) and cadmium tungstate, as well as ceramic scintillators employing rare-earth oxides and oxysulfides. A particularly critical parameter for CT detector scintillators is their afterglow behavior. Since CT data acquisition systems do not use pulse counting as in most nuclear applications, but rather integrate the detector current, the afterglow leads to measurement errors near tissue/air and tissue/bone interfaces. Data correction schemes have been developed to at least compensate for first-order effects related to the detector afterglow. Compared to the rare-gas ionization detector, the scintillation detector usually has a lower geometric efficiency (ratio of active versus ineffective frontal area), which is, however, generally more than compensated for by the latter's higher quantum efficiency.² In fan beam systems a collimator is usually employed in front of the scintillator elements to reduce the effects of scattered x-rays. In gas detector arrays the chamber walls effectively serve the same purpose by acting as a collimator aligned with the position of the x-ray focal spot. This self-collimating effect of gas detectors, on the other hand, excludes their application in ring detector based systems, since the focal spot of the x-ray tube moves relative to the detector elements.

Figure 3.2.

CT detectors. Left: Scintillation crystal with photo-sensitive semiconductor detector. Right: Rare-gas ionization chamber detector. (Reprinted with permission from E. Krestel, ed., Imaging Systems for Medical Diagnosis, Siemens Aktiengesellschaft, Berlin (more...)

3.2.3. Image Artifacts

Image imperfections or artifacts observed in CT systems can be classified according to their origin:

1.

Physics-related causes

• Quantum noise: Physically unavoidable statistical fluctuations in the x-ray beam intensity will show up as signal-dependent noise in the projection data. A reduction of the noise level requires an increase in x-ray intensity or data acquisition time, along with an increase in patient dose. The signal dependency of the quantum noise leads to typical radial noise structures in the reconstructed images.
• X-ray scattering by the patient: In fan beam type systems it is possible to use detectors that are sufficiently collimating to reduce the effect of scatter adequately. In detector ring based systems, however, mathematical scatter models have to be used to correct for this effect.
• Beam hardening: Simple empirical corrections are usually sufficient to compensate for this effect for all diagnostic purposes. In particular situations such as when the amount of bone in a particular slice is very large, residual beam hardening effects (e.g., the dark bands typically seen between the petrosal bones in head scans) may show up.
• Nonlinear partial volume effects: As a consequence of the exponential, thus nonlinear, attenuation of radiation when high-contrast structures are immersed only partially in the imaged slice, the measured data become inconsistent with the mathematical model used for reconstruction. Although correction methods have been devised, they have not been widely accepted owing to the associated increase in expense, measurement time, and complexity. The simplest and most effective way to reduce nonlinear partial volume effects is to reduce the slice thickness.

2.

Patient-related effects

• Movement of the patient or of individual organs: Typical motion artifacts will show up as stripes or groups of stripes and streaks. The intensity will depend on the contrast of the moving structures. Although in the images these artifacts can usually be easily identified as motion-related, the combination of several artifacts can make it considerably more difficult to interpret the images. The most effective way to reduce motion artifacts, of course, is to speed up the data acquisition time accordingly.
• Metallic foreign objects: Due to the strong attenuation of metal (for example, in dental fillings), the dynamic range of the detector and data acquisition system is often not large enough to accurately accommodate the measurement of the projection data. To make things worse, in such a situation nonlinear partial volume effects usually amplify inconsistencies in the projection data. As a consequence the images show streaking artifacts originating from the metallic objects.

3.

System-related causes

• Insufficient calibration of detector sensitivity
• Time-, temperature-, or irradiation-history-dependent drift in detector sensitivity
• Inadequacies in the reconstruction algorithm
• Non-uniform scanning motion
• Fluctuations in x-ray tube voltage

All of the above errors will show up in the reconstructed images with different typical patterns that vary further with the scanning geometry employed. For example, detector-related problems in fan beam systems usually manifest themselves as ring-shaped artifacts, whereas in systems based on stationary ring detectors they will have the character of streaks or stripes.

3.2.4. Quantitative CT

Although in the overwhelming majority of all examinations the resulting images are inspected only visually and qualitatively by the radiologist providing the diagnosis, a few special applications require absolute evaluation of the attenuation coefficients. These quantitative CT applications have a fairly long tradition, particularly in the case of bone mineral density measurements for diagnosis and therapeutic followup of osteoporosis. Here the measured attenuation coefficients or "Hounsfield numbers" are usually compared with simultaneous measurements made on calibrated, calcium-containing inserts. By using inserts with calcium content similar to that of the vertebra to be investigated, patient-dependent beam hardening effects can be compensated for, at least to a first-order approximation. A further refinement requires measuring attenuation data with two different x-ray spectra by changing the x-ray tube voltage. By using these so-called dual-energy scans it is possible to distinguish the individual contributions of calcium and waterlike tissue to the total measured attenuation coefficient, and thus better determine the bone mineral content of the vertebra.

There are a number of less customary quantitative CT procedures such as measurement of tissue perfusion with dynamic CT or xenon CT.³ Measurement of lung density represented an area of great interest very early on but did not become feasible in practice until spirometric (related to a measurement of the respiratory cycle) control of the scan and suitable automated evaluation methods were available. Apart from documenting findings objectively, as is frequently necessary in occupational diseases, the primary hope is to be able to detect diseases sensitively and early by selective analysis of different areas, e.g., by automated evaluation of the peripheral parts of the lungs.

3.2.5. Requirements for High-Speed CT

Most routine applications of CT ask for the best possible image acquired within the shortest possible time. The requirement for short acquisition times comes both from the necessity to effectively freeze the motion of organs like the heart and from the desire to cover large areas of the patient in the shortest possible time to improve patient comfort and patient throughput. To avoid the relatively greater effects of quantum noise associated with shorter scan times, the most critical and technically most demanding requirement is improvement to both instantaneous and sustained x-ray power from the tube or electron beam system.