Dr. Bouchard is a Fourth-Year Radiology Resident; Dr. Falen is an Assistant Professor of Radiology; and Dr. Molina is a Professor of Radiology, Residency Program Director, and Vice Chairman of Education in the Department of Radiology, University of North Carolina, Chapel Hill, NC. Dr. Molina is also a member of the Editorial Board of this journal.

Lung cancer is the most frequently diagnosed cancer in the world, and the leading cause of death from cancer. ¹ Radiology plays a critical role in the initial detection and diagnosis of thoracic malignancies, as well as in pretreatment staging, which is important in identifying patients with localized disease who are likely to benefit from surgical resection. ² Chest radiography and computed tomography (CT) have been the primary imaging modalities used to fulfill these roles. Positron emission tomography (PET) is rapidly emerging as a clinically useful noninvasive study that complements conventional radiologic imaging in the evaluation of patients with lung cancer. ³ The role of magnetic resonance imaging (MRI) in this regard generally is limited; it is used primarily for specific problem-solving issues, such as the determination of chest wall or mediastinal invasion. ⁴

Epidemiology

Lung cancer is by far the number one cause of cancer death in the United States, regardless of race or ethnic background. This disease burden is a fairly recent phenomenon, arising first in western countries in the 1930s, and rising sharply in subsequent years. Since 1950, lung cancer mortality has risen 197% in men, and 612% in women. In 1998, 20,000 more women died of lung cancer than breast cancer. The survival rate is poor, largely because lung cancer is usually diagnosed at an advanced stage. A cure can only be achieved by surgery, which is feasible only in patients who present in an early stage. However, even in this group, approximately 75% will die of recurrent disease. ^1,5 Lung cancer is rare before the age of 40, after which age-specific rates rise sharply. The ages with the greatest incidence are between 65 and 79. ¹

There are several extrinsic risk factors, but lung cancer is unique among all malignancies in that a single risk factor accounts for approximately 90% of the risk, specifically, tobacco smoking. The smoker is not the only one at risk. There is general agreement that lung cancer risks are increased by 60% from long-term environmental tobacco smoke exposure, making second-hand smoke a significant added risk factor. Other extrinsic factors include exposure to radon, and occupational exposure to asbestos, arsenic, and chromium compounds. Radon and asbestos share the distinction of having a synergistic relationship with cigarette smoking. There is increasing evidence that genetic factors can contribute to lung cancer risk. However, research has not yet identified the gene locus or the mechanism. ^1,5

Histology

Lung cancer is a broad term for cancers arising in epithelial tissue in the lining of the bronchi, and also in the trachea, bronchioles, and alveoli. Approximately 90% to 95% of all primary lung cancers are bronchogenic carcinomas. Histologically, about 20% of these cancers are small-cell carcinomas. The other 80% are grouped together as non­small-cell lung cancers (NSCLC), which are divided into three subtypes: (1) adenocarcinoma (30% to 40%), (2) squamous cell carcinoma (30% to 40%), and (3) large cell undifferentiated carcinoma (10%). These three subtypes are grouped together because NSCLCs overall have similar treatment options, as well as a better prognosis when disease is still localized at diagnosis. ^1,6,7

Approximately 90% of small-cell carcinomas are located centrally, and tend to invade longitudinally along submucosal and intramural portions of the bronchial walls and in the supporting tissues and lymphatics. ⁷ They are often characterized by extensive, bulky, mediastinal lymphadenopathy. Small-cell lung cancer is considered a nonsurgical disease, with the majority of cases presenting in an advanced stage. ⁸ The majority of squamous cell carcinomas are also located centrally, and are more frequently associated with bronchial obstruction. This is in contrast to large-cell tumors and adenocarcinomas, which tend to be peripheral. ⁶ A specific subtype of adenocarcinoma, known as bronchioalveolar carcinoma (BAC), can appear as a single nodule, segmental or lobar consolidation, or as diffuse nodules. ⁹ Other, more rare types of lung cancer include carcinoid tumors, adenoid cystic carcinomas, and mucoepidermoid carcinomas. ⁵

The role of chest radiography

The detection and diagnosis of lung cancer usually begins with a chest radiograph, either in a symptomatic patient or in a patient undergoing a chest radiograph for an unrelated reason. The appearance of lung cancer is variable, and can range from a subtle finding, to the dramatic, depending on location, stage at presentation, and associated findings.

Central tumors may be visible on the chest radiograph as an abnormal convexity or density in the hilar region. In many cases, however, the major radiographic abnormality is abnormal parenchymal opacification due to atelectasis or postobstructive pneumonitis, which may obscure the central tumor. The distribution of parenchymal findings depends on the tumor location, and can range from subsegmental atelectasis to the collapse of an entire lobe or lung (Figure 1). ^10,11 Occasionally, the cancer remains identifiable as a central contour bulge, and if it obstructs the right upper lobe bronchus, it may result in the S-sign of Golden. The peripheral concave portion of the S is formed by the interface between the collapsed right upper lobe and the remaining hyperinflated lung. The more central convex portion of the S represents the tumor itself. Other, less frequently seen manifestations of a central tumor include mucoid impaction, air trapping, and pulmonary vascular occlusion or reflex vasoconstriction leading to oligemia or infarction. ¹⁰

Often, the first indication that a cancer exists is the finding of a solitary pulmonary nodule (SPN) on a chest radiograph. The SPN is usually defined as a single round or oval opacity in the pulmonary parenchyma, measuring <3 cm in diameter. The SPN is one of the most common radiographic abnormalities detected, seen in up to 1 in 500 chest radiographs, with an estimated malignancy rate between 20% and 40%. ¹² With studies of good quality, a SPN larger than 1 to 2 cm is usually not difficult to detect, but can be overlooked easily in certain locations, such as the lung apices, the parahilar and paramediastinal regions, below the domes of the diaphragm, and overlying the spine. As many as 90% of small peripheral lung cancers have been reported to be visible retrospectively on prior chest radiographs, and missed lung cancer is a leading cause of malpractice litigation against radiologists. ¹⁰ Bronchogenic carcinoma is most often located in the upper lobes, particularly the right upper lobe, and not surprisingly, most missed cancers are in the right upper lobe (Figure 2). A 1991 study of 93 patients with SPNs found 63% to be in the upper lobes, with the right lower lobe being the next most common site. ¹³

Once discovered, certain characteristics of an SPN, such as size, calcification, shape, edge characteristics, cavitation, and growth rate can help differentiate between a benign and malignant lesion. The size of a lesion is not only important in terms of its visibility, but also because of the relationship of size to malignancy. Once a nodule reaches a size >3 cm, it is more likely to be malignant. However, the incidence of primary malignancy in smaller lesions, even in those <1.5 cm, is substantial enough that size alone is insufficient for differentiation. ¹⁰

Certain types of radiographically visible calcification, such as lamellated or central calcification in granulomas, and the popcorn pattern in hamartomas, are highly specific for benignity. Caution must be exercised, however, as a growing lung cancer may surround a calcified granuloma. In addition, dystrophic calcifications may rarely be seen in both primary and metastatic tumors, and a minority of carcinoid tumors also calcify. ⁶

The margin of a lesion can also provide useful information. Lobulation of a nodule is a worrisome feature that suggests uneven growth, and supports malignancy. Lobulations are thought to be caused by contraction or differential growth rates within the tumor. Spiculations, defined as linear strands extending from the nodule into the lung parenchyma, are of even greater concern, and are thought to represent a desmoplastic response to local tumor extension. ^6,7,10,13

Cavitation is seen in a minority of lung cancer, mostly squamous cell carcinoma, but also occasionally in adenocarcinoma or large cell types (Figure 3). Usually, the cavity wall is thick (>5 mm) and may demonstrate a nodular internal margin. A maximum wall thickness <4 mm is unlikely to be malignant, but rare cases do exist with thin walls simulating bullae. ¹⁰

Growth correlates with doubling time, which is a measure of volumetric growth, not diameter. The natural history of lung malignancy is to double its volume 40 times before death occurs. Once a nodule has reached the reliably detectable size of 9 mm, it has already undergone 30 doublings. Most tumors have a doubling time of between 30 and 490 days (mean value 120 days), and doubling times that are either shorter or longer suggest a benign etiology. A 9-mm SPN with a doubling time of 100 days will measure 1.1 cm at 3 months and 1.4 cm at 6 months, making accurate measurement critical. Discovery of an SPN on a chest radiograph should prompt every effort to obtain more remote films for comparison. The absence of growth over a 2-year period is the most reliable indicator of benignity. ^6,10 Unfortunately, making a firm diagnosis for all pulmonary nodules, especially for small nodules ¾ 1 cm, remains problematic, as evidenced by a high rate for resection of benign nodules. ¹²

Cancers arising in the lung apex, known as superior sulcus or Pancoast tumors, are a distinct subgroup because of their characteristic location and constellation of symptoms (Figure 4). Radiographic findings can be quite subtle and are frequently obscured by, or misinterpreted as, overlying musculoskeletal structures, brachiocephalic vessels, or benign pleural thickening. Findings suggestive of malignancy include an apical cap >5 mm, asymmetry of apical caps >5 mm, an apical mass, and adjacent bone destruction. ¹⁰ Clinical symptoms of arm pain (secondary to involvement of the brachial plexus) and a Horner's syndrome (meiosis, ptosis, and anhydrosis, secondary to involvement of the stellate ganglion) are classically associated with a Pancoast tumor. CT and MRI are superior to chest radiography in evaluating the lung apex, particularly MRI, because of its ability to image directly in multiple planes and visualize the brachial plexus. ^11,14

Lung cancer occasionally takes the form of focal or multifocal consolidation, typically with bronchioalveolar carcinoma (BAC). Although the most common appearance of BAC is as a SPN (43%), consolidation is the second most common radiographic pattern (30%). This pattern is caused by tumor growth along the framework of peripheral airways and alveoli, combined with mucoid secretions. Air bronchograms and air alveolograms are characteristic, but not specific, features. A pattern of focal or multifocal nodularity can result from involvement of one or more acini, and when confluent, can resemble non-neoplastic conditions, such as pneumonia, aspiration, or edema (Figure 5). However, the clinical course of bronchioalveolar carcinoma is more chronic, and without signs and symptoms of infection. The consolidative pattern has a poorer prognosis than the solitary nodular pattern. Bronchorrhea (copious white mucoid or watery expectoration), once considered the clinical hallmark of this disease, is an unusual and late manifestation seen only with diffuse bronchioalveolar carcinoma. ^6,9,10

Hilar and/or mediastinal adenopathy is sometimes the sole manifestation of lung cancer. Small-cell carcinoma tends to have bulky, central adenopathy with a relatively inconspicuous separate primary lung parenchymal site, but all cell types can have metastatic spread centrally. Careful inspection of the normal contours, lines, and stripes that classically define the mediastinum may reveal enlargement of the aortopulmonary window, right paratracheal thickening or soft-tissue density, a double density adjacent to the aortic knob, or an abnormal convexity in the azygoesophageal recess, all of which are frequent findings of mediastinal metastasis of lung carcinoma. ¹⁰ The lateral film can be especially helpful in the evaluation of suspicious increased hilar or mediastinal density.

Intrathoracic spread of lung carcinoma is not limited solely to mediastinal and hilar adenopathy. The pleura, chest wall, heart, great vessels, diaphragm, and nerves are additional structures that can be involved secondarily. Such involvement significantly impacts tumor staging, treatment, and prognosis. Pleural involvement usually manifests as a pleural effusion, with or without pleural masses. Pleural effusion (either free-flowing or loculated) implies seeding by tumor, but a nonmalignant effusion can result from central lymphatic obstruction, or a coincidental benign cause, such as pneumonia, congestive heart failure, or pulmonary embolus. ^10,11 Direct invasion of the pleura is also possible, but cannot be distinguished on a chest radiograph. Bone involvement is common, and can be due to direct extension or metastatic spread. Pancoast tumors are typically associated with direct extension to ribs or vertebral bodies, but this can also occur with other peripheral cancers. Metastatic disease may also involve other bones on the chest radiograph, as evidenced by bony destruction or lytic lesions in the humerus, sternum, clavicle, and scapula. Bronchogenic carcinoma can also involve the heart, either by direct extension or by manifesting as a malignant pericardial effusion, seen as an enlarging cardiac silhouette on serial radiographs. Elevation of the diaphragm may indicate phrenic nerve involvement by tumor, or be mimicked by a subpulmonic effusion. A history of hoarseness can be secondary to tumor involvement of the recurrent laryngeal nerve in the aortopulmonary window. ^2,6,10

The chest radiograph plays a critical role in the detection and evaluation of the protean appearances and manifestations of lung cancer. However, once the suspected tumor is identified, additional important information is often necessary that cannot be provided by the chest radiograph. Therefore, the next step in the diagnostic work-up of lung cancer is computed tomography.

The role of computed tomography

Thoracic CT scanning plays several vital roles in the evaluation of patients with known or suspected lung cancer. One is to further characterize a suspicious abnormality seen on a chest radiograph, and to provide a more complete evaluation of a primary neoplasm. A second and indispensable role is that of pretreatment or pre-
operative staging, for which CT is the primary imaging modality. Additionally, chest CT helps provide a roadmap for other staging procedures such as bronchoscopy, mediastinoscopy, transthoracic needle biopsy, and video-assisted thorocoscopy. ¹⁴ Most, if not all, of the various manifestations of lung cancer described for chest radiography can be better evaluated with CT. Cross-sectional imaging can help further clarify a tumor's location, whether in a central or peripheral location, and delineate its relationship to pleura, chest wall, and mediastinal structures. The level and degree of obstruction by central tumors leading to atelectasis and postobstructive pneumonitis can be visualized easily with cross-sectional imaging (Figure 6). Squamous cell cancers arise in the cells lining the mainstem, lobar, or segmental bronchi, and are frequently associated with this pattern of obstruction. ⁶ Trapped secretions distal to an obstructing lesion can produce the so-called mucous bronchogram.

Imaging features used to characterize an SPN on a chest radiograph are equally as useful on CT, including size and growth rate, calcification, shape and margins, and cavitation, along with the additional characteristics of density and contrast enhancement. As with chest radiography, increasing size, especially >3 cm, correlates with an increasing chance of malignancy. CT enjoys the added advantage of more accurate measurement of a nodule's size to better detect and quantify interval growth. ^6,7,12 CT can better detect and evaluate calcifications within a nodule. The distribution of calcium, rather than its presence alone, is a more important diagnostic consideration. Thin layers of calcium in a lamellar pattern are indicative of a granuloma, and popcorn calcifications with associated fat density, are associated with a benign hamartoma. ⁶

A smooth peripheral margin on CT is associated more frequently with benign lesions. As with chest radiographs, lobulations and spiculations are worrisome findings. In addition, the finding of a pleural tag, defined as a linear area of high attenuation, surrounded by aerated lung, originating from the edge of the mass and extending peripherally to contact the pleural surface, is also associated more frequently with malignant lesions. As with spiculations, a pleural tag correlates with a desmoplastic response. ^6,7,13

The wall thickness of a cavitary lesion can be measured more accurately with CT. One study found that the majority (94%) of cavitary solitary pulmonary nodules with a wall thickness ¾ 4 mm were benign, and the majority (95%) with a wall thickness >= 15 mm were malignant. Lesions with wall thicknesses between 5 and 15 mm were almost equally divided between benign and malignant. CT also has the added advantage of better evaluating the contour of a cavity's wall. A smooth inner wall is more commonly associated with a benign etiology, while a nodular internal margin reflects focal tumor excrescences. ⁶

Several other characteristics of an SPN can be evaluated with CT, such as attenuation and contrast enhancement. Homogeneous attenuation has been found to be associated more often with a benign, rather than a malignant lesion. Thin collimation CT is more sensitive than standard CT for assessment of attenuation. ¹³ In 1980, an early study with CT reported that benignity was correlated with high Hounsfield units (>164), but further studies have demonstrated that the CT numbers of an SPN cannot be viewed as an absolute value, but rather are a relative indication of attenuation, dependent on scanner and patient variables. ^15-18 These studies resulted in the introduction in the 1980s of a reference phantom that could be used as a standard with which the attenuation of a pulmonary nodule could be compared. Although there is controversy regarding the use of a reference phantom versus thin-section CT alone, a 1991 study found that a phantom is useful if the initial thin-slice CT is indeterminate. Also, if the attenuation of the nodule in question exceeds the attenuation of the phantom by 10%, it has a high probability of being benign. ^13,19,20

A newer technique for the assessment of the SPN is based on differential nodule enhancement with IV contrast material, as measured with thin-slice CT. It relies on qualitative and quantitative differences in the blood supply to benign and malignant nodules. Results from a 1992 study suggest that malignant nodules tend to enhance significantly more (20 HU increase) than benign nodules, with the most diagnostically important measurement made at 2 minutes postinjection. ²¹

One additional CT finding that may be helpful in the evaluation of lobar consolidation and the clinical suspicion of bronchioalveolar cell carcinoma is the CT angiogram sign (Figure 7). This is defined as branching pulmonary vessels extending >3 cm in the midst of completely consolidated pulmonary parenchyma that is of diffusely homogeneous lower attenuation than that of muscle. This finding has been noted to be present in a large majority of cases of BAC, but is a nonspecific finding and can be seen in pneumococcal or tuberculous pneumonia, pulmonary infarction, and even lymphoma. ^22,23

Of all newly diagnosed cases of lung cancer, approximately 80% are NSCLC. Surgical resectability, chemotherapy and radiation options, and survival rates are all related to the stage of disease. Small-cell lung cancer remains a nonsurgical disease, with the majority of patients presenting in advanced stages. The intrathoracic and extrathoracic staging of lung cancer involves assessment of the primary tumor and potential sites for metastases. The TNM staging system (Table 1), most recently revised in 1997, is the most widely accepted and utilized classification for preoperative staging. Accurate and reproducible staging is crucial in the clinical management of this disease. Since the development of CT scanning, this modality has become the mainstay in radiologic staging of chest malignancies. ^8,14,24

With CT, the primary tumor is evaluated for initial size and location, as well as possible extension beyond the lung parenchyma. The primary tumor may involve the pleura, chest wall, mediastinal structures, or vertebral bodies. Hilar tumors may invade the trachea or carina. Metastatic sites include mediastinal lymph nodes, as well as distant sites such as liver, adrenals, brain, bone, and soft tissue. ^2,6,14

Approximately 5% to 10% of patients who present with lung cancer have a pleural effusion. Unfortunately, diagnostic thoracentesis has a high (65%) false-negative rate of cytology for malignant cells. ⁸ The CT hallmark for a malignant effusion is soft-tissue nodularity along the pleural surface accompanying the effusion. Pleural nodularity or fissural thickening without an effusion can also be a sign of pleural metastasis. Pleural tumor dissemination is classified as T4 disease, and is usually considered unresectable. ^2,8,11

Approximately 5% of all lung cancers invade the parietal pleura and chest wall (Figure 8). CT has demonstrated a wide range of results when assessing for chest wall invasion by tumor. Sensitivity ranges from 38% to 87%, and specificity ranges from 40% to 90%, depending on the study. ² The best criterion for diagnosing chest-wall invasion with CT is bony destruction, with or without tumor extension into the chest wall. Other, less reliable signs of chest-wall invasion include pleural thickening, loss of the extra-pleural fat plane, an obtuse angle between the mass and the chest wall, and >3 cm of contact between the mass and the chest wall. Chest-wall invasion does not necessarily exclude resection, but there is increased morbidity and mortality associated with en bloc resection and chest-wall reconstruction in this setting. ^2,8,11

Mediastinal structures that can be invaded by tumor extension include the trachea and carina, esophagus, mediastinal vessels, heart and pericardium, and vertebral bodies (Figure 9). Disease involving these mediastinal structures is classified as T4. Although gross mediastinal invasion by bronchogenic cancer can be diagnosed confidently on CT, attempts to distinguish subtle mediastinal invasion from mere mediastinal contiguity have met with limited success. ¹¹ However, there are certain CT criteria that suggest technically feasible resectability. These include tumor contact of ¾ 3 cm with the mediastinum, less than 90š of contact with the aorta, and the presence of mediastinal fat between the mass and mediastinal structures. Unfortunately, the opposite findings (>3 cm contact with mediastinum, >90š of contact with the aorta, and obliteration of fat planes) are unreliable signs of either invasion or unresectability. ²

CT is the preferred imaging technique for evaluating adenopathy. The accurate localization of abnormal lymph nodes, whether peribronchial, hilar, mediastinal, scalene, or supraclavicular, is important in determining the N classification of the disease. The American Thoracic Society's mediastinal nodal stations are used by most surgeons, radiologists, oncologists, and pathologists to describe sites of nodal metastases in thoracic malignancies specifically ¹⁴ (Figure 10).

Lymph nodes are generally identified on CT as nonenhancing soft-tissue densities surrounded by mediastinal fat. Several studies have shown that the short axis diameter is the best predictor of actual nodal volume. ² In practice, increased nodal size is the only useful criterion for malignancy, and a node measuring >1 cm is generally considered abnormal. Increased nodal size is not an absolute indicator of malignant disease, however, as nodes <1 cm in size may contain microscopic metastases, and enlarged nodes may represent benign reactive adenopathy to an unrelated process in the chest. Central low density can represent fat in a normal node, or necrosis in a malignant node. ^2,25 The accurate description of abnormal nodal location is essential information in assisting the surgeon, bronchoscopist, or radiologist in planning a more invasive staging procedure.

Estimates of the overall prevalence of extrathoracic metastases (M1 classification and therefore stage 4 disease) range from 18% to 50%. ^7,8 The most frequent sites of hematogenous metastases are bone, brain, contralateral lung, liver, adrenals, soft tissue, and skin. Distant metastatic disease implies surgical unresectability, and the median survival once distant metastases are diagnosed ranges from weeks to months, with few patients surviving beyond 1 year. At most institutions, a chest CT in patients with suspected lung cancer is extended inferiorly to include the superior portion of the liver and the adrenal glands. ^8,11 A careful review of bone windows is also necessary in these circumstances, to exclude metastatic deposits in ribs, the sternum, the scapula, and the vertebral column.

The role of MRI

It is well accepted that CT is the imaging modality of choice for staging patients with lung cancer. Compared with MRI, CT is faster, less expensive, has better spatial resolution, offers superior evaluation of the lung parenchyma, and is more sensitive in detecting calcification. However, in certain circumstances, MRI may play a complementary role to CT because of its superior tissue contrast, multiplanar imaging capability, and superb delineation of thoracic vessels. ⁴ There are areas of the chest where the geometry of the structures of interest are better imaged with MRI. Perhaps the best example is the evaluation of Pancoast tumors, in which direct coronal and sagittal imaging with MRI facilitates assessment of invasion of the chest wall, brachial plexus, subclavian vessels, vertebral bodies, and neural foramina ^11,14 (Figure 11).

The superior contrast resolution of MRI suggests an advantage over CT in detecting subtle mediastinal invasion, but its poorer spatial resolution limits this advantage. The replacement of high signal intensity mediastinal fat by lower signal intensity tumor on T1-weighted MR images suggests the presence of mediastinal invasion, but one study comparing CT and MRI for this purpose showed similar accuracies of 56% and 50%, respectively. Some investigators have shown MRI to be superior to CT in detecting mediastinal extension when there is associated vessel involvement. MRI is also believed to be more accurate in establishing superior vena caval patency or obstruction, which may be due to thrombus, compression by soft-tissue mass, or direct invasion. ^4,11

MRI may also demonstrate chest wall invasion better, with one study reporting a sensitivity of 90% and a specificity of 86%. ⁴ Signs of chest-wall invasion on MRI include loss of the subpleural fat stripe and visualization of soft-tissue tumor extension into the chest wall. Some early data suggest that MRI can also differentiate benign (low signal intensity) pleural nodules from malignant (high signal intensity) pleural nodules with T2-weighted and proton density weighted images with a specificity of 87%. ⁴

In early studies it was initially thought that tumor within lymph nodes would generate differential signal changes with respect to normal nodes, leading to increased sensitivity and specificity of MRI when compared with CT. However, several studies comparing these modalities for detection of mediastinal nodal metastases report similar sensitivities and specificities of MR and CT. A significant disadvantage of MRI is its poorer spatial resolution, which can lead to adjacent nodes on CT appearing as an enlarged mass on MRI, resulting in the mistaken diagnosis of abnormal nodal enlargement. ¹¹

The role of PET

CT and MR imaging of the chest provides valuable information about the morphology of a lesion. However, morphologic information alone may not offer all the information necessary to direct proper clinical management. Many lesions are indeterminate as to whether they are benign or malignant by morphologic imaging techniques, such as CT and MRI, and further investigation is warranted. Patients may require an invasive biopsy procedure for further evaluation. Depending on the size and location of the lesion, CT-guided fine needle aspiration biopsy, transbronchial biopsy, mediastinoscopy, video-assisted thorocoscopy, or even thoracotomy may be considered to determine the nature of the lesion. There are risks associated with these procedures and some will be nondiagnostic. Also, >50% of radiographically indeterminate lesions that are thoracoscopically resected are found to be benign. ²⁶

One of the more recent advances in oncologic imaging that has generated a renewed interest in diagnosis, staging, and response to therapy is positron emission tomography (PET). PET imaging with [2-18F]fluoro-2-deoxy-D-glucose (F-18 FDG) allows for the evaluation of the relative level of metabolic activity of a lesion compared with other tissues. PET imaging with F-18 FDG is based on the principle that there is increased utilization of glucose in malignant cells compared with most normal tissues. ²⁷ The mechanism for the increase in F-18-FDG uptake is thought to be the result of an increase in the number of glucose transporters in malignant cells. ²⁸ There is also an increase in the glycolytic rate as a result of increased hexokinase activity. ²⁹ When F-18 FDG is administered intravenously, it competes with glucose for transport into cells. Once intracellular, it also competes with glucose for phosphorylation by hexokinase to F-18 FDG-6-phoshate. However, F-18 FDG-6-phosphate is not a substrate for glycolysis and is not further metabolized. Thus, it becomes trapped in the cancer cells, and the resultant accumulation of F-18 FDG-6-phosphate reflects glucose metabolism and allows for imaging. Malignant lesions demonstrating increased metabolism will appear as abnormal focal areas of increased radiotracer accumulation (ie, focal areas of hypermetabolism) (Figure 12).

F-18 FDG PET imaging has been shown to be an accurate, noninvasive imaging test for the assessment of pulmonary nodules and larger mass lesions. A comprehensive meta-analysis by Gould et al ³⁰ of 40 eligible studies, including 1474 focal pulmonary lesions of any size, found the mean sensitivity and specificity for detecting malignancy were 96.0% and 73.5%, respectively. However, in this analysis, there was little data for nodules <1 cm in diameter.

When a lung mass is shown to be malignant, it is important to stage the extent of disease accurately. Appropriate clinical management depends on whether there is mediastinal involvement and/or distal disease. Several studies have shown that PET is more accurate than CT for the staging of NSCLC. A tabulated summary of FDG PET literature in lung cancer from 1993 through June 2000 showed PET to have a sensitivity of 96% for the diagnosis of malignancy versus 67% for CT. For NSCLC staging, the sensitivity and specificity for PET were 83% and 91%, respectively, versus 64% and 74% for CT. ³¹

PET appears to be more accurate than CT in detecting metastatic mediastinal lymphadenopathy. In one retrospective study of 96 patients, 66 with histologically proven malignant tumors and 30 with benign masses, the sensitivity and specificity of PET for detecting malignancy in the pulmonary lesions was 97% and 89%, respectively. ³² In this study, 111 surgically sampled sites were from lymph nodes. The accuracy of FDG PET in predicting malignancy of nodes was 91% compared with 64% for CT. The sensitivity and specificity of FDG PET in detecting metastatic mediastinal lymph nodes was 98% and 94%, respectively. ³² Imaging with PET was more accurate than CT for diagnosis of both mediastinal and distant metastases.

Valk et al ³³ conducted a prospective study in 76 patients of PET imaging for staging of NSCLC in which mediastinal PET and CT findings were compared with the results of surgical staging. They reported the sensitivity and specificity for the diagnosis of N2 mediastinal nodal disease were 83% and 94% for PET and 63% and 73% for CT, respectively. ³³

In another prospective study by Pieterman et al, ³⁴ a logistic regression analysis was used to evaluate the ability of PET and CT to identify malignant mediastinal lymph nodes and distant sites in 102 patients with resectable NSCLC. The sensitivity and specificity of PET for the detection of mediastinal metastases were 91% and 86%, respectively. ³⁴

PET improves the rate of detection of local and distant metastases in patients with NSCLC. Detection of unsuspected metastatic disease by PET may permit reduction in the number of thoracotomies performed for nonresectable disease (Figure 13). For broad groupings of NSCLC stage, 10% of cases were downstaged and 33% upstaged after PET. Staging that incorporated PET provided a more accurate prognostic stratification than did staging with conventional means. ³⁵

In NSCLC, PET may reduce the need for mediastinoscopy when the primary lesion standard uptake value is <2.5 and the mediastinum is PET negative. ³⁶ The high negative predictive value of PET suggests that lesions that are negative on PET are benign, biopsy is not needed and radiographic follow-up is recommended. ³⁷ After potentially curative therapy of NSCLC, abnormalities or symptoms suggesting recurrence can be difficult to characterize. Early detection is important because salvage therapies are available for localized recurrence. PET better assesses the status of disease and stratifies prognosis than does conventional staging, effects patient management, and should be incorporated into paradigms for suspected recurrence of NSCLC. ³⁸

Though there is much excitement about the use of PET in oncology, the technique has its limitations. False-positive FDG PET scans can occur with infectious or inflammatory processes. Abnormally increased FDG uptake can be seen with sarcoidosis, tuberculosis, histoplasmosis, cryptococcosis, aspergillosis, and other infections. ³⁷

False-negative FDG PET scans have been seen with pulmonary carcinoid tumors. ³⁹ False-negative scans have also been reported with bronchioloalveolar lung carcinoma. ⁴⁰ False-negative PET imaging is also a problem if the lesion is too small for the resolution of the machine. This is most worrisome for lesions <1 cm in diameter.

In a study by Valk et al, a decision tree sensitivity analysis was used to assess the cost-effectiveness of a PET-based strategy for staging of NSCLC. They considered CT alone versus CT with thoracic PET. The CT plus PET strategy in the conservative decision tree showed a saving of $1154 per patient without a loss of life expectancy. ⁴¹ Another decision tree sensitivity analysis by Gambhir et al ³⁸ demonstrated that a strategy using PET plus CT is more economical and has a marginal increase in patient life expectancy compared with the conventional strategy of staging patients with CT alone. ⁴²

Conclusion

Lung cancer is an extremely prevalent disease that most radiologists will encounter on a frequent basis. Familiarity with the various manifestations of lung cancer on chest radiography may help suggest the initial diagnosis, especially in an older patient with a history of cigarette smoking. Once a suspicious abnormality is detected, CT is the next step in the diagnostic work-up. This is necessary to help confirm the diagnosis by identifying CT features of an abnormality that would more likely suggest cancer, and to stage the disease.

The emerging role of PET offers an exciting new diagnostic tool that can quantify the metabolic activity of a tumor or node, and can reveal additional sites of disease unsuspected on CT, thereby increasing the accuracy of the staging process. The role of MRI generally is limited to specific problem-solving areas, or when CT findings are equivocal or indeterminate. AR