Digital Radiography

Digital radiography has also invaded the workspace. This allows collection and storage of data on digital detectors and computers rather than merely relying on hard-copy film for plain-film studies. This has led to a debate between hard and soft copy as it relates to expense versus ease of use versus storage needs, and so on. Some manufacturers have switched to silicon flat-panel detectors, with cesium iodide scintillators to improve image quality, decrease radiation dose, and allow long-term storage of data. The x-ray photons are converted to light by the cesium scintillator, which in turn produces an electrical charge, which is transformed into a digital readout on an electronic processor. Digital images may be read on film, computer monitors, or even video screens. Thus, the innovators have taken a relatively simple technology and made it as complex as rocket science. The trend has been from low cost/low tech to high cost/high tech methodology with constant software upgrades. Digital radiography has largely been implemented in such plain-film bastions as the intensive care unit and mammography unit (where the debate between hard versus soft also rages). As these areas are often services that are “in the red” to begin with, the negative margin only increases.

Subtraction angiography is based on the principle that a baseline film of an area of anatomy without vascular opacification can be subtracted from a film of the same area with vascular opacification, yielding an image of the vascular structures alone. The administration of iodinated contrast allows one to opacify the blood vessels because of the differential attenuation of the x-ray beam by iodine compared with the skull and nonopacified portions of the brain, head, and neck. Taking a “negative image” of the scout film and manually applying that to one where the vessels are opacified yields a composite vessel-only study.

Parameters and Units

Nobel Prize winner Sir Godfrey Hounsfield developed CT for clinical use between 1972 and 1973. The first company to introduce a CT scanner was EMI (Electric and Musical Industries, Ltd), the same company the Beatles used for distributing their music on the Apple label.

The principles of differential x-ray beam attenuation apply to CT, except CT uses a highly collimated x-ray beam. The photons that pass through the patient are collected by CT detectors, which show a differential rate of intensity on a gray scale depending on the degree of absorption along the narrow x-ray beam. The CT scanner’s x-ray beam is rotated over many different angles so as to get differential absorption patterns across various rays through a single slab of a patient’s body. By a mathematical analysis known as projection reconstruction, one is then able to obtain an absorption value for each point (pixel) within a CT slice. To understand the concept of a pixel, one must understand how pixel size relates to the matrix and field of view (FOV).

The matrix refers to the number of imaging partitions in the x-y plane of a slice, assuming an axial slice is in the z plane. The in-plane pixel size is determined by dividing FOV by the matrix dimensions. The FOV is the linear dimension of the space to be imaged. The machine operator can select both the FOV and the matrix size. The matrix sizes of CT scanners have increased several-fold since the original 80 × 80 matrix of the EMI scanner in 1972. At present, matrices on the order of 512 × 512 are used. As an example, a 20-cm FOV scanned with a 512 × 512 matrix would yield pixels that are 0.39 mm (200 mm/512) by 0.39 mm. The final dimension one must know in CT imaging is the slice thickness. At present, CT slice thicknesses can be less than 1 mm. The three-dimensional imaging unit is called a voxel; in the example just given, the voxel size would be 1.0 × 0.39 × 0.39 mm³. For an 18-cm FOV with a 256 × 256 matrix and 8-mm slice thickness, the voxel size would be (180 mm/256) × (180 mm/256) × 8 mm. For high-resolution imaging, as desired in the temporal bone or orbits, a large matrix and a small FOV are used with slice thicknesses of 0.5 mm.

The scale for CT absorption generally ranges from +1,000 to −1,000, with 0 allocated to water and −1,000 to air (Table 1-1). The units are termed Hounsfield units (HU), named to honor the discoverer of the technique. White matter and gray matter are in the 30 to 50 HU range. Hematomas tend to range from 50 to 80 HU, and calcification is generally 150 HU or greater. These values vary by approximately 10 to 25 HU, according to the particular CT machine that is used. Dense bone and metal are the materials at the highest HU range. High protein concentrations (clotted blood, tenacious sinus secretions, the lens of the eye) equate with higher HU values. At values less than 0 one finds the structures that show less CT attenuation than water. Fat is usually in the −40 to −100 HU range. In neuroradiology, the structures with less CT attenuation than fat are relatively limited to air-containing materials (airway, mastoid air cells, sinuses) (Fig. 1-1B).

Table 1-1 HU of Central Nervous System Structures on CT

Evolution of CT Scanners

CT technology has evolved over several generations, each one designed to reduce scan time and increase image quality. The first-generation CT scanner had a thin x-ray beam and one detector. The second-generation scanners used a fan-shaped beam and multiple detectors. The arc of scanner gantry motion improved from 1-degree increments to as much as 30-degree differences. Third-generation scanners used an even wider fan-shaped beam and 10 times as many detectors as the second-generation scanner. The gantry rotated 360 degrees and moved continuously. Fourth-generation scanners used circumferential detectors so that only the x-ray tube moved in a 360-degree arc.

Most CT scanner manufacturers now use “slip ring technology,” which allows continuous data acquisition and gantry rotation throughout the scanning procedure as the table moves without stopping and starting for each slice. This procedure, called spiral scanning, has allowed scan times per slice to be reduced to 1 second or less. The increased heat capacity of the newer x-ray tubes and the increasing sensitivity of the CT detectors have allowed more rapid image acquisition and more slices before x-ray tube heating becomes prohibitive. Other advances in CT collimation have allowed thinner and thinner slice profiles to be obtained. At present, 0.5-mm thick sections are often used for evaluation of fine anatomic structures such as the ossicles in the temporal bone or for CT angiography studies. Because the beam is so well collimated, x-ray exposure to the patient is limited to the area of scanning, and if overlapping sections are not used, the overall dose to the patient is less than 3 rad to the imaged volume. Of course, one cannot help but get “scatter radiation,” which may affect radiosensitive organs such as the thyroid glands or gonads (both of which are hypometabolic in RIG). Helical (spiral) scanning has allowed excellent quality CT angiography studies to be performed, thus enabling CT to compete with ultrasound and MR angiography (MRA) for the evaluation of neck and intracranial vessels.

One of the terms used to define the parameters for helical scanning is pitch, which is defined as the table speed times the tube rotation time divided by the slice width. As an example, if a scanner has a table speed of 5 mm/sec and you scan for 0.8 seconds and have a slice thickness of 3 mm, your pitch would be 5 × 0.8/3 = 1.33 pitch. In the past the pitch floated between 0.75 and 1.50, but as table speeds have increased and slice thicknesses has reduced, pitches of 2.0 to 15.0 are not uncommon. Fast table speed is desirable to cover more anatomy in a scan; however, spatial resolution decreases with higher pitch (like Roger Clemens’ fastball—the faster it moves the more the ball blurs—from a hitter’s standpoint).

Another advantage to helical scanning is that once you have a volume of tissue scanned, you can slice it in as many thin sections or in as many planes as you wish. In general, the best image quality is produced when images are reconstructed using at least half the collimator setting. You can play with overlapping images and thinner slice reconstructions for CT angiography studies.

The latest refinement in CT technology is the multidetector system. In this scenario, instead of 1 mm × 20 mm detector channels, you have 1 mm × 1.25 mm channels. The key to optimizing scanning is deciding which detectors to turn on when. Image thickness is selected by changing collimation, detector configuration, and reconstruction algorithm. In practical terms, with one rotation of the gantry, one is able to perform interweaving helices producing multiple (current scanners for sale are in the 16- to 256-slice mode) slices instead of one per rotation. The speed of data acquisition and patient throughput can be accelerated in this way; alternatively, thinner slices and higher resolution can be achieved. The evolution of CT has thus progressed from 0.5 image per second (single slice with 1 second of scanning and 1 second of table movement), through 2 images per second (helical imaging with 2 images in 1 second as the table moves), to 64 images per second (multidetector mode with 64 images in 1 second as the table moves).

CT Perfusion

Xenon-133 CT is a method for evaluating cerebral perfusion. The xenon is inhaled in combination with oxygen, and CT scans are performed to determine cerebral blood flow (CBF) at multiple locations in the brain. Brain xenon concentration is related to the concentration of xenon absorbed in the bloodstream, the brain blood flow, and the time of exposure to xenon. Decreased flow has been documented in patients with meningitis, vasospasm, head trauma, sickle cell disease, and stroke. Adverse effects of xenon inhalation may include sedation, bronchospasm, and respiratory depression.

Iodine-based CT perfusion has also been introduced recently. A large rapid bolus of contrast is infused during continuous rapid scanning of a single slice, and the wash-in and wash-out of the bolus can be analyzed by a computer to generate semi-quantitative images of brain perfusion. Commonly measured parameters are mean transit time (MTT), cerebral blood volume (CBV), and CBF (CBF = CBV/MTT). On CT perfusion images, the differential density between normal brain and hypoperfused brain can be accentuated through computer manipulation to demonstrate areas of ischemia in the brain. The main disadvantages to this technique are the large-bore catheter required (14 to 16 gauge), the rapid injection rate (5 to 10 mL/sec), patient discomfort (one should use nonionic contrast to reduce the “barf” factor), the limited number of slice acquisitions that limit the region of study, and the reluctance of clinicians to give iodinated contrast to patients with strokes. To that end, MR still has an advantage in that only milliosmoles are delivered with gadolinium injections (see discussion of Magnetic Resonance Imaging in this chapter).

CT perfusion CBF maps are noted to be more sensitive to ischemia than blood volume or time-to-peak maps. Infarctions may occur in most patients in areas of the brain with CBF values no more than 30% of normal tissue and in 50% of patients where the CBF of affected tissue is 30% to 50% that of normal tissue.

CT technology has really taken off in the past 5 years, and the sanctity of MR as the premiere means for evaluating intracranial pathology now rests more in the elimination of radiation exposure and increased soft-tissue contrast than in image resolution and functionality.

Algorithms, Windows, and Contrast Agents

Different reconstruction algorithms, or kernels, can be used to highlight a particular tissue with CT. Thus, bone disease may be best visualized with a bone, edge, or detail algorithm used to accentuate the interface between the bone and the soft tissue. Alternatively, the algorithm for data reconstruction can be set to highlight differences in soft-tissue attenuation of structures. If you save the raw data from a scan, any number of algorithms can be used retrospectively to analyze (target) the tissues studied.

The images from a given algorithm may be displayed with different window widths and levels to photograph the pictures in a manner that accentuates differences in CT attenuation between structures. Window widths refers to the HU range selected for gray-scale display, whereas the window level refers to the center point about which the range is displayed. By using small window widths (80 to 400 HU) and center levels (20 to 80 HU), you can highlight subtle soft-tissue differences. To visualize tissues with wide variations in CT attenuation, as in bone versus air, a larger width (2000 to 3000 HU) and level (300 to 600 HU) are used.

Contrast enhancement is often used in cranial CT to opacify blood vessels and to detect areas of abnormal blood-brain barrier breakdown, where iodinated contrast will seep into the parenchyma. The factors that determine contrast enhancement of a lesion include (1) the volume and delivery of the contrast to the intravascular system, (2) the size of the intravascular space, (3) lesion vascularity, (4) permeability of lesion blood vessels, and (5) size of extravascular intralesional space.

Occasionally, intrathecal contrast is administered through a lumbar, cervical, cisternal, or ventricular approach to visualize intracranial pathology. Approximately 3 mL of nonionic contrast material (iodine, 180 mg/mL) can be administered through the cisternal or ventricular approach without risk of deleterious effects (i.e., seizures, headache, nausea, vomiting, and neuralgia). Myelographic doses are discussed later.

Current Role of CT in Neuroimaging

What is the role of CT in neuroimaging (Box 1-2)? It remains the quickest, most efficient screening technique in patients with head trauma. It is the most sensitive imaging study for the detection of subarachnoid hemorrhage (SAH), and is the study of choice for initial evaluation of patients with signs and symptoms suggestive of SAH. CT angiography has gained ascendancy as the mode for screening for aneurysms in patients with SAH in the ED. In a similar vein (equally large), CT perfusion has become an emergency department technique for the evaluation of possible stroke patients and for guiding thrombolytic therapy.

The sensitivity of CT for calcification is critical in increasing diagnostic specificity, particularly for central nervous system (CNS) tumors (e.g., craniopharyngioma, oligodendroglioma, neurocytoma, retinoblastoma, meningioma), metabolic disorders (e.g., parathyroid dysmetabolism), and congenital lesions (e.g., TORCH infections, tuberous sclerosis). In the head and neck, chondroid and osseous lesions are well depicted on CT and may be confusing in their appearances on MR imaging. CT is the best study for bony (nonmarrow-replacing) lesions and is indispensable in evaluating the temporal bone in general and the skull, face, and spine for fractures. It is critical for defining the intricate anatomy of the paranasal sinuses. In the spine, CT still holds its own in the evaluation of cervical and lumbar bony spinal stenosis, trauma, and postoperative studies where the hardware precludes adequate MR definition.

Magnets

The two main types of magnets used in clinical imaging are permanent magnets and superconducting magnets. Permanent magnets can be thought of as two bar magnets that generate a uniform magnetic field between them. Typically, the magnets are composed of metallic alloys, with iron used as the material that outlines the two magnets and conducts the magnetic field from one bar to the next. Permanent magnets, as the name implies, do not require continual energy to maintain the magnetic field. With a patient lying between the two permanent magnets, the magnetic field is oriented perpendicular to the axis of the supine body.

Superconducting magnets require no additional energy input once they have been “magnetized,” mainly because they are encased in a liquid helium shell that prevents dissipation of the energy. The helium keeps the magnet cold enough to maintain its superconductance (i.e., zero resistance). An outer insulating layer of supercoolant liquid nitrogen is often used to keep the helium cold. The nitrogen requires constant replenishment or the magnet “quenches.” The coils of superconducting material that produce the magnetic field in a superconducting magnet are generally made of niobium-titanium wire. The static magnetic field of a superconducting magnet is oriented parallel to the axis of the supine patient body.

Because of their heavy weight from the iron yoke that surrounds and connects them and the weight of the magnets themselves, the maximum field strength of the permanent magnet scanner is relatively low, usually 0.5 tesla. A tesla (T) is a unit of magnetic field strength and is equivalent to 10,000 gauss (G). By comparison, the earth’s magnetic field is approximately 0.5 G. Aren’t you impressed now with “low-field strength” (0.15 to 0.5 T) magnets? They are much stronger than Mother Earth! Most nonopen superconducting magnets in clinical use range from 1.0 to 3.0 T systems, but scanners with field strengths up to 8.0 T are available and can blow your ears off with the noise they generate. Open magnets range from 0.1 to 1.0 T in strength.

Gradient Coils

The concept of gradient coils is important to understanding MR. To localize a point in space, the magnetic field at that point must be unique. The way to alter the magnetic field (which is uniform in strength) within the bore is to pass current through gradient coils, which create an organized continuous gradation. That is, for a z gradient, there is an orderly increase in field strength of the main magnet from one end to the other. Each coordinate axis has a series of gradient coils, which are essentially loops or half-loops of wire that carry current. By winding different-shaped coils around a cylinder, one is able to achieve an x-, y-, or z-oriented magnetic gradient field. Therefore, at any point in the x, y, and z planes a unique magnetic vector will be present in the scanner. A proton in that location in the scanner will precess with a unique frequency that is proportional to the magnetic field it “feels.” By tuning to a particular frequency one can localize that point and judge the amplitude of its signal. This is one way that spatial localization is achieved with MR. Turning the imaging gradient coils off and on is what gives MR its characteristic loud noise.

Obviously the homogeneity (uniformity of magnetic field strength across an FOV) of a magnetic field affects the image quality. Other coils (called shim coils) are used to correct any unwanted localized inhomogeneities to the main magnetic field. Much emphasis in recent years has therefore been placed on those gradient shim coils that allow better homogeneity to the applied magnetic field. These shim coils should not be confused with the three gradient coils used to localize a particular region of interest within an x-y-z coordinate system.

The MR manufacturers are in a never-ending battle to upstage each other with the strength of their gradient packages. What this translates to for the physician is shorter times to echo (TEs), smaller FOVs, faster scanning, and higher resolution. We encourage these battles with our demands for image quality, but pay through one of our apertures for the cost.

Radiofrequency Receiver Coils

Radiofrequency receiver coils are used to receive magnetic signals from the region of interest within the body. In neuroradiology, except for the head, neurovascular head and neck, and whole body coils, which transmit signal, the surface coils are limited to receiving signal from the imaged volume after the body coil stimulates it. Innovations in surface coil technology have led to the creation of more specialized coils, including those used for the spine, temporomandibular joint, shoulder, and knee, and even rectal and intravaginal coils. The smaller the surface coil, the higher the signal-to-noise ratio (SNR) but the smaller the sensitive volume. One is therefore forced to strike a compromise between sensitivity profile (coverage), SNR, and resolution. If the coil is too small, one obtains excellent resolution and SNR but insufficient coverage. If the coil is too large, one obtains adequate coverage but insufficient SNR to support the resolution one desires. This has led to the concept of phased-array coil and parallel imaging systems, in which signal is obtained from several small coils simultaneously or sequentially to scan a large volume. For example, a multicoil spine system may use a linear array of four coils, each approximately 6 inches in size, ample to cover from C1 to L2. The four coils are electrically isolated from one another (with low-input impedance preamplifiers and overlapping fields) and are each connected to a separate MR receiver, preamplifier, and digitizer. Each coil has limited coverage but very high resolution. The four separate images are then combined (by sophisticated computers) to form one composite image that has maximal coverage, SNR, and resolution.

Larmor Frequency

Most of MR consists of proton imaging of the hydrogen nucleus because it is very abundant within human tissue. The hydrogen nucleus precesses in a magnetic field at its own resonant frequency, called the Larmor frequency. The Larmor frequency of the hydrogen nucleus is linearly related to field strength by the following equation:

The gyromagnetic ratio of each nucleus is unique and does not vary in different field strength magnets. At 1 T the Larmor frequency for hydrogen is 42.6 MHz, whereas at 1.5 T the Larmor frequency is roughly 63.86 MHz. To stimulate the hydrogen nuclei, a radiofrequency (RF) pulse must be tuned to the Larmor frequency of the hydrogen nucleus (its resonant frequency). As the hydrogen nucleus is put into a magnetic field with a gradient of magnetism, the location of a particular hydrogen nucleus can be determined by its resonant frequency within that gradient. Again, the frequency varies slightly, depending on where in the field the nucleus resides. The ability to localize protons by variations in their Larmor frequency as a response to a graded magnetic field allows the spatial characterization that distinguishes MR imaging from nuclear magnetic resonance, or chemical spectroscopy.

Relaxation Times

When a sample containing hydrogen nuclei is placed in a magnet, its magnetization aligns along the direction of the magnetic field (z direction). After stimulation of the hydrogen nuclei by applying a 90-degree RF pulse at the Larmor frequency of its nucleus, the magnetization vector rotates from the z axis to the transverse x-y plane, where the protons precess at the Larmor frequency. According to Faraday’s law of induction, the precessing magnetization creates voltage in a properly oriented receiver coil (the same coil that is used to apply the RF pulse when the head or body coils are used). One can vary the angle at which the vector is tipped from the z axis by varying the amplitude and duration of the RF pulse. The hydrogen nuclei then relax by two mechanisms. The first is termed T1, or spin-lattice relaxation time. As the nucleus relaxes back to equilibrium after being excited by an RF pulse, there is an exponential increase in the amplitude of the z-direction magnetization until there is complete return of the magnetization toward its baseline position. T1 is defined as the time it takes for the hydrogen nucleus to recover 63% of its longitudinal (z-axis) magnetization. At the same time, the transverse magnetization in the x-y plane also decays toward zero in an exponential fashion. This exponential decay is characterized by a time constant that is termed T2, or spin-spin relaxation time. The signal created in a proton’s decay is called a free induction decay (FID). In spin-echo imaging, rather than detecting the FID, a 180-degree RF pulse is given at some time (half the TE) after the initial 90-degree RF pulse. This rephases the spins after another ½ TE and when all the spins are coherent produces the so-called spin echo. Thus, the TE is the time from the 90-degree RF pulse to the echo. The analogy is to a race where the slow and fast runners start together (in phase), but very soon thereafter the fast runners pull ahead of the slow runners. At a certain time (½ TE) in the race the runners are told to turn around and head back to the starting line (180-degree RF pulse). All the runners should return across the starting line at the same time (spin echo) if they have kept up their original pace.

Because T1 and T2 relaxation mechanisms are independent of each other by and large, one can completely lose signal in the x-y axis without having completely returned all the magnetization to the z axis. The T2 or transverse (spin-spin) relaxation is due to dephasing caused by the adjacent hydrogen nuclei, which are not totally in concert with each other. T2 is defined as the time for 63% of the transverse magnetization signal to be lost owing to this natural dephasing process. By and large, T2 values in the CNS are shorter than T1 values. The T1 and T2 values of some normal tissues seen in the CNS are listed in Table 1-2.

Table 1-2 Representative T1 and T2 Relaxation Times of CNS Structures at 1.5 T

The overriding concept of T2 relaxation is that of phase dispersion or incoherence caused by local field inhomogeneity. However, phase dispersion may be due to three factors: (1) the magnetic environment of the hydrogen protons (true T2), (2) the heterogeneity in the main magnet itself (extrinsic variations caused by magnetic field imperfections and other inhomogeneities produce phase dispersion characterized by the time constant T2′), and (3) the paramagnetic substance-induced field inhomogeneities (blood or iron) known as T2″. Now you can understand that various tissues within the human body have varying magnetic susceptibilities (affinities to be magnetized), which result in different local field strengths and which cause phase dispersion as the patient’s body is placed in the “uniform” magnetic field. To reiterate, consider three components: T2 from spin-spin relaxation, T2′ caused by main field inhomogeneity, and T2″ from susceptibility effects. The reciprocal of these relaxation times (relaxation rates), when summed, can be related by this equation:

The reason these three factors are worth emphasizing becomes clear in a discussion of the differences between spin-echo and gradient-echo pulse sequences.

Pulse Sequences