Chapter 7 Imaging Techniques

Today, clinicians have two main imaging techniques at their disposal for the investigation of patients with chest disease—plain radiography, which produces a projectional image, and computed tomography (CT), which provides a cross-sectional view. Other techniques, such as magnetic resonance imaging (MRI), radionuclide scanning, and ultrasonography, can provide valuable additional information but are rarely performed without previous chest radiography or CT. Because imaging is an integral part of the practice of respiratory medicine, an understanding of the strengths and weaknesses of these various techniques is vital. The advent of high-resolution and spiral (helical) CT techniques has lent further precision to the clinical investigation of suspected chest disease, but the use of such sophisticated tests should not be indiscriminate; accurate interpretation of the chest radiograph remains the mainstay of thoracic imaging.

Plain Chest Radiography

Technical Considerations

The views of the chest most frequently performed are the erect posteroanterior and lateral projections, taken with the patient’s breath held at total lung capacity. On a frontal (posteroanterior) chest radiograph, just under half of the lung is free from overlying structures, such as the ribs or diaphragm. Many technical factors determine how well the lungs are demonstrated. The characteristics of current digital imaging systems make it possible to adjust the final image and optimize exposure of the least and most dense parts of the chest in a single are image.

Because the coefficients of x-ray absorption for bone and for soft tissue approach one another at high kilovoltage, the skeletal structures do not obscure the lungs on a higher-kilovoltage radiograph to the same degree as on low-kilovoltage radiographs. The high-kilovoltage radiograph thus demonstrates much more of the lung. Improved penetration of the mediastinum also allows some of the central airways to be seen. Although high-kilovoltage radiographs are preferable for routine examinations of the lungs and mediastinum, low-kilovoltage radiographs provide good detail of unobscured lung because of the improved contrast between lung vessels and surrounding lung. Furthermore, dense lesions—for example, calcified pleural plaques—are particularly well demonstrated on low-kilovoltage films.

The past decade has seen a major change in plain film radiography with the development of digital imaging systems, which are now ubiquitous in modern radiology departments. Digital chest radiography, yielding images either stored on a phosphor plate and then digitally scanned or captured directly onto a detector plate, has been combined with computer-based picture archiving and communications systems (PACSs) for distribution of images around the hospital or over wider networks. The much wider latitude of digital systems also allows the image to be “postprocessed” to provide optimum visualization of the relevant structures

The frontal (posteroanterior) (Figure 7-1) and lateral (Figure 7-2) projections are sufficient for most purposes in chest radiography. Other radiographic views are less frequently required, but they should not be overlooked because they may solve a particular problem quickly and cheaply. The lateral decubitus view is not, as its name implies, a lateral view. It is a frontal view taken with use of a horizontal beam and the patient in a side-lying position. Its main purpose is to demonstrate the movement of fluid in the pleural space (Figure 7-3). An adaptation of this view is the “lateral shoot-through” sometimes used in bed-bound patients: A lateral radiograph of the supine patient is taken to show an anterior pneumothorax behind the sternum (not always visible on a frontal chest radiograph) (Figure 7-4). If a pleural effusion is not loculated, it gravitates, to some extent, to the dependent part of the pleural cavity. Thus, in a decubitus patient, the fluid will layer between the chest wall and the lung edge. This view also may be useful for demonstrating a small pneumothorax, because the visceral pleural edge of the lung falls away from the chest walls in the nondependent hemithorax.

Figure 7-1 A standard digital posteroanterior radiograph demonstrating the wide latitude that can be routinely achieved. This image provides good detail of both lung and mediastinal anatomy.

Figure 7-2 Standard lateral chest radiograph. The dorsal spine vertebral bodies are of progressively lower density toward the diaphragm. The metallic density over the cardiac silhouette is an atrial septal defect closure device.

Figure 7-3 Demonstration of small effusions. A, Posteroanterior chest radiograph obtained in a patient with a ventriculoperitoneal shunt. More soft tissue than usual is present between the gastric air bubble and the base of the lung because of a subpulmonic effusion. B, Decubitus view shows redistribution of fluid to the dependent part of the chest (arrows).

Figure 7-4 Lateral shoot-through digital radiograph of the chest obtained in a patient in the intensive care unit. The anterior pneumothorax (arrowheads indicate the visceral pleural edge) was not obvious on the anteroposterior portable radiograph.

For the lordotic view, now rarely performed, the x-ray beam is angled 15 degrees cranially, either by positioning the patient upright and directing the beam up or by leaving the beam horizontal and leaning the patient backward. On this view, the lung apices are demonstrated free from the superimposed clavicle and first rib. It may be useful to differentiate pulmonary shadows from incidental calcification of the costochondral junctions (Figure 7-5).

Figure 7-5 The value of lordotic views. A, Method of obtaining a lordotic view of the lung apices: The x-ray beam is angled upward. B, Selected area from a standard posteroanterior view of the upper lung zones in a patient who presented with hemoptysis, with a suggestion of a small opacity projected over the anterior end of the left first rib. C, A lordotic view confirms that the small opacity is intrapulmonary (rather than calcified costochondral cartilage).

Portable Chest Radiography

Portable or mobile chest radiography has the obvious advantage that the examination can be carried out without moving the patient from the ward. However, the portable radiograph has disadvantages. Use of the shorter-focus film distance results in undesirable magnification, and most portable machines are unable to deliver the power required for high-kilovoltage techniques. Furthermore, the maximum current is limited so that longer exposure times are needed, which potentially increases blurring of the image. Portable lateral radiographs are even less likely to be successful because of the extremely long exposure times required to obtain adequate penetration.

Patient positioning for portable radiography is difficult, and the resultant radiographs often are suboptimal. Even with use of the so-called erect position, in which the patient sits up, the chest is rarely as vertical as it is in a standing patient. Because many patients are unable to move to the radiography department for a formal radiograph, any method of improving the quality of a portable chest radiograph, such as digital radiography, represents a significant advance.

Digital Chest Radiography

The most widely employed systems use conventional radiographic equipment but use a reusable photostimulatable plate instead of conventional film. The reusable phosphor plate is housed in a cassette and stores some of the energy of the incident x ray as a latent image. On scanning the plate with a laser beam, the stored energy is emitted as light that is detected by a photomultiplier and converted into a digital signal. The digital information is then manipulated, displayed, and stored in whatever format is desired. The phosphor plate can be reused once the latent image has been erased by exposure to light. Most currently available computed radiography systems produce a digital radiograph with a resolution of more than 10 line pairs per millimeter. The fundamental requirement to segment the image into a finite number of pixels has resulted in much work to determine the relationship between pixel size, which affects spatial resolution, and the detectability of focal abnormalities. Although it might seem desirable to aim for an image composed of pixels of the smallest possible size, an inverse relationship occurs between pixel size and the cost and speed of data handling. Thus, pixel size is ultimately a compromise between image quality and ease of data processing and storage.

An unequivocal advantage of digital computed radiography over conventional film radiography is the linear photoluminescence-dose response, which is much greater than that of conventional film. This extremely wide latitude coupled with the facility for image processing produces diagnostic images over a wide range of exposures.

Observer performance studies have shown that computed radiography is just as useful as conventional film radiography for virtually any relevant application. However, postprocessing of the digital image has to be used to match the digital radiograph to the specific task. Enhancement of the image for one purpose often degrades it for another but is easily achieved in most PACS reporting systems.

Computed Tomography

The same basic principles that allow film radiography apply with CT—namely, the absorption of x rays by tissues that contain constituents of different atomic number. By use of multiple projections and computed calculations of radiographic density, slight differences in x-ray absorption are displayed as a cross-sectional image. The components of a CT scanner include an x-ray tube that rotates around the patient and an array of x-ray detectors opposite the tube, together contained within the gantry. The patient lies on the examination couch, which moves the patient through the aperture of the CT gantry. The data acquired are then processed by the CT computer, resulting in the final images as displayed on the CT monitor.

An impressive and rapid improvement in CT hardware capability has occurred over the past decade. Most particularly, the advent of multiple-channel CT scanners has resulted in the ability to acquire simultaneous helical datasets. An accompanying increase in gantry rotation speed coupled with the reduction in the size of the individual detectors has resulted in the ability to acquire extremely detailed images in very short scan times. On the current “top specification” scanners from the major manufacturers, up to 320 channels are available, each with a detector size of as small as 0.5 mm. The entire thorax can now be scanned at submillimeter resolution in 1 to 2 seconds. Thus, spiral (also known as volume or helical) scanning entails continuous scanning and table movement into the CT gantry (Figure 7-6). The information is reconstructed into axial sections, perpendicular to the long axis of the patient, identical to conventional CT sections.

Figure 7-6 The principle of spiral (helical) computed tomography. The patient moves into the scanner with the x-ray tube continuously rotating and the detectors acquiring information. The rapidity of data acquisition allows a complete examination of the thorax to be performed in a single breath-hold.

Temporal resolution has been further improved, because data reconstruction algorithms now allow CT images to be generated after a partial rotation of the gantry. Thus, temporal resolution of as little as 65 msec is now possible, enabling modern multichannel CT scanners to acquire cardiac gated images that effectively freeze cardiac motion. This capability in turn can be applied to allow detailed analysis of coronary artery and cardiac anatomy.

The analysis of what is frequently hundreds of individual images that are produced as the result of a single CT examination is undertaken on dedicated CT or PACS workstations. Postprocessing of these thin sections also allows the production of multiplanar reformats (MPRs), maximum and minimum intensity projections (MIPs and MinIPs), and angiographic images. Skeletal structures can be automatically removed, or surface-rendered images that mimic appearances familiar to the bronchoscopist can be produced with a few mouse clicks. These images are visually pleasing and allow an exquisite appreciation of anatomy. They also have a role in the planning of interventional procedures, including transbronchial needle biopsy and endoluminal stent insertion (Figure 7-7).

Figure 7-7 Dataset from a multidetector computed tomography study. A, This axial image demonstrates a bronchogenic carcinoma in a right paratracheal position. B, A reformatted image derived from the same data demonstrates the same abnormality in the coronal plane without loss of resolution. C, A coronal volume-rendered image obtained in a different patient demonstrates a left superior sulcus tumor encasing the left subclavian artery (arrow). D, Virtual bronchoscopy image from a normal subject.

Technical Considerations

The CT image is composed of a matrix of picture elements (pixels). A fixed number of pixels make up the matrix, so the size of each pixel varies according to the diameter of the circle to be scanned. A typical matrix in a modern CT system would be 512 by 512 pixels. The smaller the image field of view, the smaller the area represented by a pixel and the higher the spatial resolution of the image. In practical terms, the field of view size is adjusted to the size of the area of interest, usually the chest diameter.

Often, marked differences can be seen in the “look” of the images obtained on the various models of CT scanners. Such differences are largely the result of specific features of the software reconstruction algorithms used to smooth the image, to a greater or lesser extent, by averaging the density of neighboring pixels. The lung is a high-contrast environment, so less smoothing is needed than in other parts of the body. Higher-spatial-resolution algorithms (which make image “noise”—a granular appearance—more conspicuous) generally are more desirable for lung work.

Section Thickness

Although a CT section is viewed as a two-dimensional image, it has a third dimension of depth. The depth, or section thickness, is determined by a combination of factors, depending on the exact parameters utilized, including focal spot size, thickness of the individual detector elements, and width of the x-ray beam collimation. Because a section has a predetermined thickness, each pixel has a volume and this three-dimensional element is referred to as a voxel. The computer calculates the average radiographic density of tissue within each voxel, and the final CT image consists of a representation of the numerous voxels (not individually visible without magnification) in the section. The single attenuation value of a voxel represents the average of the attenuation values of all of the various structures within the voxel. The thicker the section, the greater the chance that different structures will be included within the voxel and the greater the signal averaging that occurs. This is known as the partial volume effect; the easiest way to reduce this effect is to use thinner sections (Figure 7-8).

Figure 7-8 The partial volume effect on computed tomography (CT) scan appearance. A, This 10-mm CT section shows a poorly defined opacity, adjacent to the left superior mediastinum, apparently within the lung. B, The 1.5-mm section through the same region reveals that the appearance in A results from a partial volume effect—that is, the aortic arch is partially included in the 10-mm-thick sections.

When the entire chest is examined, contiguous thin sections are reconstructed for analysis. If the study is undertaken on a multichannel system, the dataset may be reconstructed at thinner intervals predetermined by the thickness of the detector rows, and these thinner sections may be used for reporting or multiplanar reconstructions. Thinner sections also are used to study fine detail and complex areas of anatomy, such as the aortopulmonary window and subcarinal regions. Another specific example for which narrow sections may be useful is to display differential densities (which would otherwise be lost because of the partial volume effect) of the small foci of fat or calcium that are sometimes seen within a hamartoma.

If exposure factors are otherwise kept the same, the total patient radiation dose varies very little between different multichannel systems. Of note, however, a striking difference in the radiation dose to the patient is associated with use of contiguous sections versus interspaced fine sections. Thus, the effective dose to the patient with interspaced fine sections (e.g., 1 or 2 mm) every 10 mm, such as used for high-resolution CT of the lung parenchyma, is 5 to 10 times less than that imposed by single-channel or multichannel spiral CT of the entire chest volume. The disadvantage of interspaced sections is the inability to view the data in any plane, but for the purposes of assessment of the lung interstitium, this added refinement usually is not of sufficient added diagnostic value to warrant the increased radiation burden. This consideration is especially important in the relatively younger patient.

Window Settings

The average density of each voxel is measured in Hounsfield units (H); these units have been arbitrarily chosen so that zero is water density and −1000 is air density. The span of Hounsfield units reflecting density in the thorax is wider than in any other part of the body, ranging from that for aerated lung (approximately −800 H) to that for ribs (+700 H). Two variables are used that allow the operator to select the range of densities to be viewed—window width and window center (or level).

The window width determines the number of Hounsfield units to be displayed. Any densities greater than the upper limit of the window width are displayed as white, and any below the limit of the window are displayed as black. Between these two limits, the densities are displayed in shades of gray. The median density of the window chosen is the window center or level; this center can be moved higher or lower as desired, thus moving the window up or down through the range. The narrower the window width, the greater the contrast discrimination within the window. No single window setting can depict the wide range of densities encountered in the chest on a single image. For this reason, at least two sets of images are required to demonstrate the lung parenchyma and soft tissues of the mediastinum, respectively (Figure 7-9). Standard window widths and centers for thoracic CT vary between departments, but generally for the soft tissues of the mediastinum, a window width of 400 to 600 H and a center of +30 H is appropriate. For the lungs, a wide window of 1500 H and a center of approximately −500 H are usually satisfactory. For bones, the widest possible window setting at a center of +30 H is best.

Figure 7-9 The effect of window settings on computed tomography scan appearance. A 5-mm-thick computed tomography section is displayed on different window settings. A, On lung windows (center, 500 H; width, 1500 H), nodules in the lungs and pulmonary vessels are clearly visible. B, On soft tissue windows (center, 35 H; width, 400 H), the contrast-enhanced vessels in the mediastinum and the soft tissue structures are delineated, but the lung detail is lost.

Window settings have a profound influence on the size and conspicuity of normal and abnormal structures. Nonetheless, it is impossible to prescribe precise window settings because of the element of observer preference and also differences between machines. The most accurate representation of an object seems to be achieved if the value of the window level is halfway between the density of the structure to be measured and the density of the surrounding tissue. For example, the diameter of a pulmonary nodule, measured on soft tissue settings appropriate for the mediastinum, will be grossly underestimated. When inappropriate window settings are used, imaging of smaller structures (e.g., peripheral pulmonary vessels) will be affected proportionately much more than that of larger structures.

Intravenous Contrast Enhancement

Intravenous contrast enhancement is needed only in specific instances, because of the high contrast on CT between vessels and surrounding air in the lung and between vessels and surrounding fat within the mediastinum. One such instance is to aid the distinction between hilar vessels and a soft tissue mass. The exact timing of the injection of contrast material depends most on the time the CT scanner takes to scan the thorax. With multichannel CT scanners, the circulation time specific for the patient becomes an important factor.

Contrast medium rapidly diffuses out of the vascular space into the extravascular space, so that opacification of the vasculature after a bolus injection with a “power injector” quickly declines, and structures such as lymph nodes steadily increase in density over time. Such dynamics result in a point at which a solid structure may have exactly the same density as an adjacent vessel. The timing and duration of the contrast medium infusion must therefore be taken into account in interpreting images obtained in a contrast-enhanced CT study. Rapid scanning protocols with automated injectors tend to improve contrast enhancement of vascular structures at the expense of enhancement of solid lesions because of the rapidity of scanning. With spiral CT, it is possible to achieve good opacification of all of the thoracic vascular structures by using small volumes of contrast material. Optimal contrast enhancement is a prerequisite for the diagnosis of pulmonary embolism or aortic and great vessel abnormalities. To achieve optimal contrast enhancement, many CT systems now use an automated triggering system. Thus, in examining the pulmonary arteries, a low-dose repeating scan will monitor the density in the pulmonary outflow tract once every second. When a predetermined density threshold is reached as a result of the arrival of intravenous contrast, the preplanned examination is triggered. The couch rapidly moves the patient from the monitoring position to the start position, a prerecorded breath-hold instruction is given to the patient over a loudspeaker, and the data acquisition commences. The acquisition is timed to correspond with appropriate enhancement of anatomic structures if contrast has been administered.

For examining inflammatory lesions, such as the reaction around an empyema, it may be necessary to delay scanning by 30 seconds, to allow contrast to diffuse into the extravascular space. For examining the liver and adrenals in evaluation of a patient with suspected lung cancer, the optimal phase of contrast enhancement to maximize the conspicuity of hepatic metastases is during the portal venous phase of contrast enhancement, and this occurs 60 to 80 seconds after contrast injection.

High-Resolution Computed Tomography

Technical Considerations

Over the past two decades, the development of high-resolution computed tomography (HRCT) has had great impact on the approach to the imaging of diffuse interstitial lung disease and bronchiectasis. Images of the lung produced by HRCT correlate closely with the macroscopic appearance of pathologic specimens, so in the context of diffuse lung disease, HRCT represents a substantial improvement over chest radiography. Three factors associated with significantly improved spatial resolution of CT—hence the designation “high-resolution”—are narrow beam collimation, use of a high-spatial-frequency reconstruction algorithm, and a small field of view.

Narrow collimation of the x-ray beam reduces volume averaging within the section and so increases spatial resolution compared with standard 10-mm collimation. For routine HRCT scanning, 1.50-mm beam collimation generally is regarded as optimal. Narrow collimation has a marked effect on the appearance of the lungs, notably the vessels and bronchi—the branching vascular pattern seen particularly in the midzones on standard 10-mm sections has a more nodular appearance with narrow sections, because shorter segments of the obliquely running vessels are included in the section. In addition, parenchymal details become more clearly visualized (Figure 7-10).

Figure 7-10 The effect of computed tomography (CT) section thickness and edge enhancement on image appearance. A, A 10-mm-thick section reconstructed with edge enhancement from a multidetector CT (MDCT) dataset. B, A 1.5-mm-thick section with high edge enhancement from the same dataset, typical of a high-resolution CT (HRCT) lung image from a volume acquisition.

In HRCT lung imaging, a high-spatial-frequency algorithm is used to take advantage of the inherently high-contrast environment of the lung. The high-spatial-frequency algorithm (also known as the edge-enhancing, sharp, or formerly “bone” algorithm) reduces image smoothing and makes structures visibly sharper but at the same time makes image noise more obvious (see Figure 7-10).

Several artifacts are consistently identified on HRCT images, but they do not usually degrade the diagnostic content of the images. Nevertheless, it is useful to be able to recognize the more common ones. Probably the most frequently encountered is a streaking appearance, which arises from patient motion. Cardiac motion sometimes causes movement of the adjacent lung with consequent degradation of image quality. Some CT scanners are able to eliminate this artifact by triggering the acquisition of the slice from the electrocardiogram (ECG) tracing so that the data are collected during diastole, when cardiac motion is minimized. To optimize this technique, the scanner must have a short rotation time and also be capable of formatting a CT image from data from a partial rotation. This reduces the data acquisition time window to as little as 360 msec.

The size of the patient has a direct effect on the quality of the lung image—the larger the patient, the more conspicuous the noise, which is seen as granular streaks because of increased x-ray absorption by the patient. This artifact is particularly evident in the posterior lung adjacent to the vertebral column. The phenomenon of aliasing results in a fine, streaklike pattern radiating from sharp, high-contrast interfaces. The severity of the aliasing artifact is related to the geometry of the CT scanner, and, unlike quantum mottle, aliasing is independent of the radiation dose. These artifacts are exaggerated by the nonsmoothing, high spatial-resolution reconstruction algorithm but do not mimic normal anatomic structures and are rarely severe enough to obscure important detail in the lung parenchyma (Figure 7-11).

Figure 7-11 High-resolution computed tomography image demonstrating artifact caused by aliasing and quantum mottleing. Detail is obscured in the posterior parts of the lungs. The patchy parenchymal opacification results from desquamative interstitial pneumonitis.

The degree to which HRCT samples the lung depends primarily on the spacing between the thin sections. An HRCT examination also may vary in terms of the number of sections, the position of the patient, the phase in which respiration is suspended, the window settings at which the images are displayed, and the manipulation of the image by postprocessing. No single protocol can be recommended to cover every eventuality. However, the simplest protocol entails 1.5-mm collimation sections at 20-mm intervals from apex to lung bases. Any given scanning protocol may need to be modified—a patient referred with unexplained hemoptysis ideally is scanned with contiguous standard sections through the major airways (to show a small endobronchial abnormality) and interspaced narrow sections through the remainder of the lungs (to identify bronchiectasis).

When early interstitial disease is suspected, for example, in asbestos-exposed persons in whom the chest radiograph is normal in appearance, HRCT scans often are performed with the patient in the prone position, to prevent any confusion with the increased opacification seen in the dependent posterior basal segments in many normal subjects scanned in the usual supine position. The increased density seen in the posterior dependent lung with supine positioning disappears in normal persons when the scan is repeated at the same level with prone positioning. No advantage is gained by scanning a patient in the prone position if no obvious diffuse lung disease is found on a contemporary chest radiograph.

A limited number of scans taken at end expiration can reveal evidence of air trapping caused by small airway disease, which may not be detectable on routine inspiratory scans. Areas of air trapping range from a single secondary pulmonary lobule to a cluster of lobules that give a patchwork appearance of low attenuation areas adjacent to higher attenuation, normal lung parenchyma (Figure 7-12).

Figure 7-12 A, High-resolution computed tomography (CT) scan through the lower lung lobes of a patient with severe asthma. The inspiratory image through the upper zones is normal. B, A high-resolution CT image at end expiration emphasizes the regional air trapping due to small airways obstruction.

Alterations of the window settings of HRCT images sometimes make detection of parenchymal abnormalities impossible when there is a subtle increase or decrease in attenuation of the lung parenchyma. Uniformity of window settings from patient to patient aids consistent interpretation of the lung images. In general, a window level of −500 to −800 HU and a width of between 900 and 1500 HU are usually satisfactory. Modification of the window settings for particular tasks is often desirable; for example, in looking for pleuroparenchymal abnormalities in asbestos-exposed patients, a wider window of up to 2000 HU may be useful. Conversely, a narrower window of approximately 600 HU may emphasize the subtle density differences that characterize emphysema and small airway disease.

The relatively high radiation dose to the patient inherent in all CT scanning needs to be appreciated. The radiation burden to the patient is considerably less with HRCT than with conventional CT scanning. It has been estimated that the mean radiation dose delivered to the skin with HRCT by use of 1.5-mm sections at 20-mm intervals is 6% that of conventional 10-mm contiguous scanning protocols. A further method of reducing the radiation burden to the patient is to decrease the milliamperage; it is possible to reduce the milliamperage by up to 10-fold and still obtain comparably diagnostic images. Although continuous refinement in CT technology is reducing the radiation burden to patients, CT still delivers a relatively high radiation dose to patients, so this imaging modality must not be used indiscriminately.

Clinical Applications of High-Resolution Computed Tomography

Increasingly, HRCT is used to confirm or refute the impression of an abnormality seen on a chest radiograph. It may also be used to achieve a histospecific diagnosis in some patients who have obvious, but nonspecific, radiographic abnormalities.

It probably is impossible to determine the frequency with which HRCT will show significant parenchymal abnormalities when the chest radiograph appears normal. Studies of individual diseases show that HRCT demonstrates abnormalities despite normal chest radiographs in 29% of patients with systemic sclerosis and in up to 30% of those with asbestosis. For hypersensitivity pneumonitis, the proportion may be even higher. As indicated by the average sensitivity results of several studies, HRCT seems to have a sensitivity of approximately 94%, compared with 80% for chest radiography; this increased sensitivity does not seem to be achieved at the expense of decreased specificity.

In patients with clinical and lung function evidence of diffuse lung disease, HRCT is now central in the diagnostic workup, with clinical performance greatly exceeding that of plain chest radiography and may obviate the need for lung biopsy. In the original study that compared the diagnostic accuracy of chest radiography and CT in the prediction of specific histologic diagnosis in patients with diffuse lung disease, Mathieson and associates showed that three observers could make a confident diagnosis in 23% of cases on the basis of chest radiographs and in 49% of cases with use of CT; the correct diagnosis was made in 77% and 93% of these readings, respectively (Figure 7-13).

Figure 7-13 High-resolution computed tomography patterns in diffuse interstitial lung disease. A, Subpleural reticular pattern typical of established fibrosing alveolitis. B, Multiple irregularly shaped cystic spaces within the lungs in a young patient with preserved lung volumes. Images through the lung bases were more normal. This high-resolution computed tomography pattern and distribution combination is virtually pathognomonic for Langerhans cell histiocytosis.

A number of subsequent early HRCT studies acted as the forerunners of a large body of work that has established HRCT as a cornerstone in the assessment of patients suspected of having diffuse lung disease but for whom the clinical features and appearance on the chest radiograph do not allow a confident diagnosis to be made. A number of diffuse lung diseases can have a “diagnostic” appearance on HRCT when findings are interpreted by experienced chest radiologists; such diseases include fibrosing alveolitis, sarcoidosis, Langerhans cell histiocytosis, lymphangioleiomyomatosis, pneumoconiosis, and hypersensitivity pneumonitis (Figure 7-14). An intriguing observation is that the ability of HRCT to allow observers to provide correct histospecific diagnoses seems to be maintained in advanced end-stage disease.

Figure 7-14 High-resolution computed tomography scans of the chest in a patient with subacute hypersensitivity pneumonitis. A, Widespread nodular and ground glass patterns. B, The areas of decreased attenuation evident posteriorly are made more obvious on this scan obtained at end expiration.

However, HRCT is sometimes used indiscriminately for patients in whom the high certainty of diagnosis from clinical and radiographic findings does not justify the extra cost and radiation burden. No evidence shows that an HRCT examination adds anything of diagnostic value for a patient who has progressive shortness of breath, finger clubbing, crackles at the lung bases, and the typical radiographic pattern and lung function profile of fibrosing alveolitis. Nevertheless, the ability of HRCT to characterize disease, and often to deliver a definite and correct diagnosis in patients with nonspecific radiographic shadowing, frequently is helpful.

Much interest has been shown in defining the role of HRCT in staging disease activity, particularly for fibrosing alveolitis, in which cellular histology indicates disease activity and is used to predict both responses to treatment and prognosis. As shown by more recent evidence, a predominance of ground glass opacification in fibrosing alveolitis predicts a good response to treatment and increased actuarial survival compared with patients with a more reticular pattern, which denotes established fibrosis. Similar observations about the potential reversibility of disease can be made with use of HRCT in patients who have sarcoidosis, in whom a ground glass or a nodular pattern predominates. In other conditions, the identification of ground glass opacification on HRCT, although nonspecific, almost invariably indicates a potentially reversible disease—for example, extrinsic allergic alveolitis, diffuse pulmonary hemorrhage, and Pneumocystis jiroveci pneumonia (Box 7-1). An important exception is bronchoalveolar cell carcinoma, in which areas of ground glass opacification that merge into areas of frank consolidation or a more nodular pattern may be seen. Another caveat applies with the situation in which fine, intralobular fibrosis is seen on HRCT as widespread ground glass opacification; in this rare occurrence, evidence of traction bronchiectasis usually is present within the areas of ground-glass opacification.

Box 7-1

Causes of Ground Glass Opacification

Pneumocystis jiroveci or cytomegalovirus pneumonia

Acute respiratory distress syndrome–acute interstitial pneumonia

Hypersensitivity pneumonitis—subacute

Desquamative interstitial pneumonitis

Pulmonary edema

Idiopathic pulmonary hemorrhage

Bronchioloalveolar cell carcinoma

Alveolar proteinosis

Lymphocytic interstitial pneumonia

Respiratory bronchiolitis–interstitial lung disease

The ability of CT to discriminate among various patterns of disease has clarified the basis for the sometimes complex mixed obstructive and restrictive functional deficits found in some diffuse lung diseases. A good example is hypersensitivity pneumonitis, in which both interstitial and small airway disease coexist; patterns caused by these different pathologic processes can be readily appreciated on HRCT. The extent of the various HRCT patterns correlates with the expected functional indices of restriction and obstruction, respectively. Other conditions in which CT is able to tease out the morphologic abnormalities responsible for complex functional deficits include fibrosing alveolitis with coexisting emphysema and sarcoidosis associated with a combination of interstitial fibrosis and small airway obstruction by peribronchiolar granulomata.

In patients for whom lung biopsy is deemed necessary, HRCT may be invaluable to indicate which type of biopsy procedure is likely to be successful in obtaining diagnostic material. The broad distinction between peripheral disease versus central and bronchocentric disease is easily made on HRCT. Thus, disease with a subpleural distribution, such as fibrosing alveolitis, is most unlikely to be sampled by transbronchial biopsy, whereas diseases with a bronchocentric distribution on HRCT, such as sarcoidosis and lymphangitis carcinomatosa, are consistently accessible to transbronchial biopsy. In patients for whom an open or thoracoscopic lung biopsy is contemplated, HRCT assists in determining the optimal biopsy site. Pathologic examination of a lung biopsy specimen can still justifiably be regarded as the final arbiter of the presence or absence of subtle interstitial lung disease. Because HRCT images provide a kind of “in vivo big picture,” many lung pathologists now combine the imaging and pathologic information before assigning a final diagnosis, and in many centers, the benefits of a team approach to the diagnosis of diffuse lung disease are recognized. The indications for HRCT that have been developed over the past 20 years are summarized in Box 7-2.

Box 7-2

Indications for High-Resolution Computed Tomography of the Lungs

To narrow the differential diagnosis or to make a histospecific diagnosis in patients with obvious but nonspecific radiographic abnormalities

To detect diffuse lung disease in patients with normal or equivocal radiographic findings

To elucidate unexpected pulmonary function test results

To investigate the underlying problem in patients presenting with hemoptysis

To evaluate disease reversibility, particularly in patients who have fibrosing alveolitis

To guide the type and site of lung biopsy

Magnetic Resonance Imaging

Plain radiography, CT, ultrasound imaging, contrast angiography, and isotope scanning constitute the mainstays of thoracic disease imaging. Although magnetic resonance imaging (MRI) has developed a role complementary to these techniques, it generally is considered a problem-solving tool rather than a technique of first choice.

MRI entails placing the subject in a very strong magnetic field (typically 0.2 to 1.5 tesla) and then irradiating the area under examination with pulses of radiowaves. Anatomic MRI depends on the presence of water within tissue to produce the signal required for interpretation. Protons within this water exist within different local atomic environments and, consequently, have different properties. These differences, measured as magnetic resonance, can be exploited by sequence manipulation to generate differences in contrast between tissues in the final image. Thus, the frequency of the radiofrequency pulse transmitted into the patient is carefully selected so that it causes hydrogen protons within water to be disturbed from the orientation that they have assumed as a result of being placed inside the powerful magnetic field within the bore of the magnet. After the transient disturbance caused by the radiofrequency pulse, these protons, which are acting akin to small bar magnets, relax back into their original resting position. As they do this, they release energy as a further pulse of radio waves, which are detected by the receiver coils located in the wall of the bore of the magnetic coil or, more commonly, in a variety of receiver coils placed more directly around the area under investigation. These coils frequently are known by the body part they have been designed to examine—thus, a knee, head, neck, or body coil is placed appropriately at the start of the examination. In the case of thoracic imaging, the body coil usually consists of a pair of coil mats placed in front of and behind the patient.

Historically, the main strengths of MRI are the high intrinsic soft tissue contrast generated, the lack of artifact from bone, the absence of exposure to ionizing radiation, and the ability to produce images in any chosen plane. The major weaknesses of MRI in the thorax have, until recently, been its susceptibility to image degradation secondary to respiratory and cardiac motion, as well as the relatively long times required to perform an examination. In general, the quality of MR images is related to the field strength of the scanner and the peak power and speed of the amplifiers that generate the interrogating radiofrequency pulses.

For thoracic imaging, ECG-triggering facilities, whereby the acquisition of imaging data can be coordinated with the cardiac cycle to reduce flow artifact, are essential. Various methods of compensation for respiratory motion have been developed. Some approaches use external devices such as respiratory bellows, which detect movement of the chest wall, with data collection occurring when motion is at its least. Other methods are essentially software developments that compensate for respiratory disruption of magnetic spins. Most of these techniques have been superseded on modern scanners by the ability to acquire images of the thorax with use of breath-hold techniques.

Mediastinal and Chest Wall Imaging

The most common indications for the use of MRI in respiratory disease are for investigation of neoplastic disease, most commonly bronchogenic carcinoma. In addition to the primary disease, secondary complications such as cerebral secondaries, spinal metastases, and retroperitoneal fibrosis all lend themselves to evaluation by MRI. MRI also permits assessment for invasion of mediastinal structures such as the major airways, heart and great vessels, chest wall, and diaphragm and allows differentiation among different forms of soft tissue, fluid, hemorrhage, local hematoma formation, and aneurysms (Figures 7-15 and 7-16). With modern multichannel CT techniques, MRI now holds relatively little advantage over CT in assessing chest wall invasion, except with superior sulcus tumors. However, MRI does provide superb anatomic detail without subjecting the patient to radiation exposure—an important consideration in the pediatric age group, in which a number of follow-up studies may be required (Figure 7-17). The disadvantage of MRI in the very young child is the necessity for general anesthesia in many cases.

Figure 7-15 Magnetic resonance imaging of a right upper zone lung mass in an 11-year-old boy. A, A coronal T1-weighted sequence demonstrates a high-signal-intensity apical mass. B, With the addition of fat saturation, reducing the signal returned from fat, the signal intensity in the mass falls significantly. This imaging feature confirms the fatty nature of the mass, which was a large pleural lipoma.

Figure 7-16 Chest wall invasion by tumor demonstrated with magnetic resonance imaging. Oblique sagittal T2-weighted image through the long axis of the left ventricle demonstrates an adjacent chest wall mass (arrows) extending through the interior chest wall into the overlying breast tissue. This was due to recurrent breast carcinoma.

Figure 7-17 Extralobar pulmonary sequestration. A, A coronal contrast-enhanced breath-hold image demonstrates the avidly enhancing pulmonary sequestration at the left lung base. Note the clear plane between the triangular sequestrated segment and the diaphragm and underlying spleen. B, Volume-rendered angiographic image demonstrating the same triangular sequestrated segment (asterisk) with two supplying branches from the aorta (arrowheads) and complex venous drainage. The largest vein drains subdiaphragmatically (arrows) into the left renal vein.

Lung Parenchymal Imaging

Use of magnetic resonance techniques for imaging the lungs has been limited by a number of significant technical challenges. First, the lungs are constantly moving because of respiratory and cardiac motion. Second, they have a low water content relative to other biologic tissues, so they have a low proton density and return relatively little signal. Third, because of the multiple interfaces between air and soft tissue, innumerable small disturbances arise in the magnetic field. This loss of homogeneity at air-tissue interfaces results in a phenomenon known as magnetic susceptibility artifact, further reducing signal and increasing noise. Thus, on standard sequences, normal lung exhibits little signal and often is obliterated by artifact. Attempts to tackle this problem are showing increasing promise—for example, in the area of suppurative lung disease in younger patients.

Ventilation Studies

Another area of intense interest has been the use of polarized gases (helium 3 and xenon 129) to show pulmonary ventilation. With this technique, a process of heating and irradiating with polarized light produces polarized gases. The gases (which have a short half-life) are inhaled and imaged using optimized sequences. The use of dual-frequency probes allows gas and proton images to be acquired and registered, enabling function and anatomy to be correlated.

Magnetic Resonance Angiography

Magnetic resonance also can be used to demonstrate vascular anatomy by differential visualization of flowing blood and stationary tissue; this may be achieved with or without intravenous MRI contrast agents. Generally, the use of contrast increases the signal returned from blood, increases the signal-to-noise ratio, and allows acquisition times to be shorter. This modality, known as magnetic resonance angiography (MRA), can be used to look at venous or arterial flow, together or separately (see Figure 7-17).

The contrast agents used in MRI generally and MRA in particular are based almost exclusively on gadolinium chelates. Most such agents are sequestered in the extracellular spaces; they cause shortening of the T1 relaxation time and thus increase the signal from the enhanced tissue on T1-weighted sequences. The distribution of these agents is very similar to that of the iodinated contrast agents used routinely in CT.

Pulmonary Emboli and Infarction

At present, MRI is not routinely used in patients with suspected pulmonary embolism and infarction, but it has been the subject of much research: In a number of published series, breath-hold pulmonary MRA has been found to show fifth-order pulmonary vessels and to permit diagnosis of emboli to the segmental level. Presence of smaller pulmonary emboli can be inferred by lack of segmental and subsegmental perfusion. Three-dimensional MRA datasets can be acquired and displayed on workstations as moving projections to demonstrate areas of deficient perfusion.

Vascular Malformations and Congenital Anomalies

Increasing evidence suggests that MRI can clearly define a number of vascular and developmental anomalies of the lungs by combining anatomic and flow studies. Such anomalies include the scimitar syndrome, hypogenetic lung syndrome, pulmonary artery agenesis, bronchopulmonary sequestration, and vascular malformations (see Figure 7-17).

Cardiovascular Imaging

MRI is now a key technique for imaging the heart and great vessels and is widely used for the assessment of cardiac anatomy and function. Rapid and accurate assessment of wall motion, determination of ejection fraction, and stress testing for reversible ischemia, hibernating myocardium, and valvular disease are now routine. The ultimate challenge—namely, accurate imaging of the coronary arteries—is under intense investigation, although this application has not yet reached routine practice. Nevertheless, MRI is now able to provide comprehensive noninvasive cardiac assessment that is likely to challenge more established techniques such as nuclear medicine and echocardiography.

Pulmonary Angiography

Pulmonary angiography is used to investigate pulmonary circulation when other, less invasive, methods have failed to provide the requisite information. The most frequent indication is for suspected pulmonary embolism, often after ventilation-perfusion scanning. In the acute assessment of pulmonary embolism, the angiogram is undertaken within 24 hours of clinical presentation. However, a delay of 48 to 72 hours should not preclude the use of pulmonary angiography, although the diagnostic yield progressively declines because of fragmentation of thrombi over time, especially if anticoagulation has been instituted.

Pulmonary angiography is now rarely used. Apart from the relative expense and invasive nature of angiography, it is perceived to have a high complication rate (although this is not supported by the published evidence), so it has been largely replaced by CT.

The technique of pulmonary angiography involves fluoroscopically directed insertion of a guidewire followed by a modified pigtail catheter into the right and then the left main pulmonary arteries in turn, with injection of a nonionic contrast administered at an appropriate flow rate. At least two views per side are required, with additional oblique or magnification views as necessary. Catheter access usually is through the femoral vein, with use of the internal jugular and subclavian veins as possible alternatives. Most departments undertake angiography with digital subtraction vascular equipment (Figure 7-18). Problems with misregistration artifact, inherent in digital subtraction systems and caused by respiratory or cardiac cycle phase differences between the mask image and the contrast image, usually can be overcome. Crossing the tricuspid valve may induce an arrhythmia that usually is transient. Therefore, electrocardiogram (ECG) monitoring is mandatory, and the use of prophylactic antiarrhythmic agents or temporary pacing-wire insertion is common practice in some centers. Right-sided heart catheter pressure measurements and gas analysis also may be undertaken.

Figure 7-18 Digital subtraction pulmonary angiogram. A large thrombus causes a filling defect within the contrast in the artery of the left lower lobe (large arrow). Smaller thrombi are present within the proximal branches to the upper lobe (small arrows).

When a pulmonary embolus is present, it most frequently is situated in the posterior segments of the lower lobe. Thrombi beyond the segmental vessel level are detected less reliably than more central thrombi. However, the significance of thrombi confined to subsegmental vessels is unclear. The typical angiographic findings with pulmonary embolism are those of vascular cutoff or, when vascular occlusion is not complete, an intraluminal filling defect with contrast passing around and beyond the clot. Indirect signs of embolism include areas of relatively delayed or reduced perfusion, late filling of the venous circulation, and vessel tortuosity. When the angiogram is performed to investigate suspected chronic thromboembolic disease, vascular changes to look for include local stenosis or thin webs, luminal ectasia, and irregularities of the normal tapering pattern.

Bronchial Artery Embolization

Bronchial artery embolization usually is performed to treat massive hemoptysis in patients who are unsuitable candidates for surgical management. The most common causes of bronchial artery hypertrophy and consequent hemorrhage are suppurative lung diseases (particularly bronchiectasis) and fibrocavitary disease that involves mycetomas. Less common causes of hemorrhage from the bronchial circulation include bronchial carcinoma, chronic pulmonary abscess, and congenital cyanotic heart disease. No absolute contraindications to bronchial artery embolization are known, although the patient should be hemodynamically stable and able to cooperate.

The most common anatomic arrangement on bronchial arteriography is that of one main right bronchial artery arising from a common intercostobronchial trunk, which comes off the thoracic aorta at approximately the level of T5, and two left bronchial arteries arising more inferiorly. However, bronchial arteries may arise from the thyrocervical trunk, the internal mammary artery, the costocervical trunk, the subclavian artery, a lower intercostal artery, or the inferior phrenic artery or even the abdominal aorta. The right intercostal bronchial trunk takes off from the aorta at an acute upward angle, whereas the left bronchial arteries leave the aorta at more-or-less right angles, and special catheters have been designed to facilitate selective catheterization. Superselective catheterization of the bronchial circulation allows precise delivery of embolic material, thereby preventing spillover into the aorta or inadvertent embolization of the spinal artery.

Fiberoptic bronchoscopy often is advocated before bronchial artery embolization to establish the site of hemorrhage. However, a large-volume hemoptysis almost invariably results in vigorous coughing, thereby spreading blood throughout the bronchial tree, which makes localization impossible. CT angiography also is a useful preliminary investigation, delineating bronchial artery anatomy, guiding intervention, and sometimes localizing the lobe or segment from which the bleeding originates. Few criteria exist to determine which angiographically demonstrated bronchial arteries should be embolized. Guidelines are particularly relevant when several bronchial arteries have been identified and the site of hemorrhage is not obvious from previous thoracic imaging. Embolization is directed at the vessels considered most likely to be the source of hemorrhage (Figure 7-19). Bronchial arteries of diameter greater than 3 mm may be considered to be pathologically enlarged. In patients with diffuse, suppurative lung disease, most commonly cystic fibrosis, attempts are made to embolize all significantly enlarged bronchial arteries bilaterally. If no abnormal bronchial arteries are identified, a systematic search is made for aberrant bronchial arteries. When a patient continues to experience hemoptysis after embolization, all suspicious systemic arteries should be examined for a source of bleeding, and it may be necessary to angiographically investigate the pulmonary circulation for a source of bleeding.

Figure 7-19 Bronchial arteriogram obtained in a patient who presented with hemoptysis. A, Marked hypertrophy of the bronchial artery to the right upper zone is evident. These changes were caused by cystic fibrosis. Previous embolization coils are seen over the right upper lobe. B, After embolization with a combination of small coils and particles, no further flow into these branches occurs.

A variety of embolic materials have been used for the embolization of bronchial arteries, ranging from spheres of polyvinyl alcohol in a variety of sizes to small pieces of surgical gel (Gelfoam). Although coils lodged proximally in the bronchial artery have been used, they can prevent subsequent catheterization.

After bronchial artery embolization, many patients experience transient fever and chest pain. Some patients cough up a small amount of blood, which possibly arises from limited infarction of the bronchial mucosa. Serious complications after bronchial artery embolization are rare, the most serious being transverse myelitis, probably caused by contrast toxicity rather than inadvertent embolization. Inadvertent spillover of embolization material into the thoracic aorta may cause distant ischemia in the legs or abdominal organs.

The aim of bronchial artery embolization is the immediate control of life-threatening hemoptysis, which is achieved in more than 75% of patients. Failures usually result from nonidentification of significant bronchial arteries and an inability to maintain the catheter position to allow subsequent embolization. Up to 20% of patients rebleed within 6 months of an initially successful bronchial artery embolization. The reasons cited for recurrent hemorrhage are recanalization of previously embolized vessels, incomplete initial embolization, and hypertrophy of small bronchial arteries not initially embolized. However, bronchial artery embolization usually can be satisfactorily repeated in patients who rebleed.

Superior Vena Cava Stenting

Superior vena cava obstruction (SVCO) is characterized by facial and upper limb swelling, headache, and shortness of breath and usually is caused by advanced mediastinal malignancy. Conventional palliative treatment relies on radiotherapy, chemotherapy, and sometimes surgery. Radiotherapy usually produces an initial improvement, although subsequent recurrence of symptoms is frequent. Balloon angioplasty for treatment of both benign and malignant causes of SVCO has been reported, but not surprisingly, symptoms are liable to recur soon after angioplasty alone.

The percutaneous placement of metallic stents for the treatment of SVCO has several attractions. With increasing experience, reliable and successful palliation of SVCO has been reported with use of various stent designs. A superior venacavagram is necessary to identify the length of the stenosis and its site in relation to the confluence of the brachiocephalic veins and the right atrium. Identification of intraluminal thrombus or tumor may require thrombolysis before stent insertion, or the use of a covered stent. After balloon dilatation of the superior vena cava stricture, the stent is positioned across the stricture, and a postplacement venacavagram is performed to confirm free flow of blood into the right atrium (Figure 7-20). Subsequent to angioplasty and stent placement, relief of SVCO symptoms usually is rapid and dramatic. Recurrence of symptoms may be caused by venous thrombosis or tumor progression. Although rupture of the superior vena cava at the time of angioplasty is a risk, this complication seems to be extremely rare, possibly because of the tamponade provided by surrounding tumor or postirradiation fibrosis.

Figure 7-20 Stenting of superior vena cava obstruction. In this patient, the obstruction was caused by mediastinal malignancy. A, Superior venacavagram showing a tight stricture in the midportion of the superior vena cava. B, Balloon dilation of the stricture. C, Placement of a mesh-wire stent in the now-patent superior vena cava.

The role of intravascular stents in the management of nonmalignant SVCO has not yet been defined. Patients who have SVCO caused by benign fibrosing mediastinitis have been treated successfully, although occlusion of the stent secondary to the progression of the mediastinal fibrosis or with endothelial proliferation may occur.

Normal Radiographic Anatomy

Mediastinum and Hilar Structures

The mediastinum is delineated by the lungs on either side, the thoracic inlet above, the diaphragm below, and the vertebral column posteriorly. In the context of radiographic anatomy, the various structures that make up the mediastinum are superimposed on each other, so they cannot be separately identified on a two-dimensional chest radiograph; for this reason, the normal anatomy of the individual components of the mediastinum is considered in more detail in the later section on CT of the mediastinum. Nevertheless, because a chest radiograph usually is the first imaging investigation, it is necessary to appreciate the normal appearances of the mediastinum and the considerable possible variations resulting from the patient’s body habitus and age.

The mediastinum is conventionally divided into superior, anterior, middle, and posterior compartments (Figure 7-21). The practical benefit of use of these arbitrary divisions is that specific mediastinal pathologies show a definite predilection for individual compartments (e.g., a superior mediastinal mass most frequently is caused by intrathoracic extension of the thyroid gland; a middle mediastinal mass usually results from enlarged lymph nodes). However, localization of a mass within one of these compartments does not normally allow a specific diagnosis to be made, and neither do the arbitrary boundaries preclude disease from involving more than one compartment.