肺内多发小结节鉴别

DOI: 10.1378/chest.129.3.805 2006;129;805-815 Chest Naidich Suhail Raoof, Alexey Amchentsev, Ioannis Vlahos, Ajay Goud and David P. Algorithm Pictorial Essay: Multinodular Disease: A High-Resolution CT Scan Diagnostic This information is current as of September 13, 2006 http://www.chestjournal.org/cgi/content/full/129/3/805 located on the World Wide Web at: The online version of this article, along with updated information and services, is ISSN: 0012-3692. may be reproduced or distributed without the prior written permission of the copyright holder. 3300 Dundee Road, Northbrook IL 60062. All rights reserved. No part of this article or PDF published monthly since 1935. Copyright 2005 by the American College of Chest Physicians, CHEST is the official journal of the American College of Chest Physicians. It has been by Kim Yong Mi on September 13, 2006 www.chestjournal.orgDownloaded from Pictorial Essay: Multinodular Disease* A High-Resolution CT Scan Diagnostic Algorithm Suhail Raoof, MD, FCCP; Alexey Amchentsev, MD; Ioannis Vlahos, MD; Ajay Goud, MD; and David P. Naidich, MD, FCCP The evaluation of patients presenting with multinodular pulmonary disease provides an impor- tant clinical challenge for physicians. The differential diagnosis includes an extensive list of benign and malignant processes making the management of these cases frequently problematic. With the introduction of high-resolution CT (HRCT) scanning, the ability to assess various patterns of diffuse multinodular disease has evolved into an essential part of the diagnostic process. The purpose of this article is to develop an approach to the diagnosis of multinodular parenchymal disease using HRCT scan pattern recognition as a point of departure. (CHEST 2006; 129:805–815) Key words: algorithm; multinodular; multiple nodules Abbreviations: HP � hypersensitivity pneumonitis; HRCT � high-resolution CT; ILD � interstitial lung disease; LCH � Langerhans cell histiocytosis; LIP� lymphocytic interstitial pneumonitis; RB� respiratory bronchiolitis F or the purposes of this report, multinodulardisease will be defined in a patient in which there are too many nodules to easily count on routine CT scan studies, with most of these nodules measur- ing � 1 cm in diameter. While the most common cause of multiple pulmonary nodules is metastatic disease, it is apparent that this definition encom- passes a wide range of lung diseases, both benign and malignant. It is our contention that use of a dedi- cated diagnostic algorithm1 based on characteristic high-resolution CT (HRCT) scan features coupled with clinical findings can provide either a specific diagnosis or a markedly shortened list of differential diagnoses in a majority of patients presenting with diffuse lung nodules. Algorithm Overview Due to its ability to evaluate the lung parenchyma in cross-section, eliminating the superimposition of densities, CT scanning offers a unique opportunity to evaluate lung nodules in exquisite detail.2 This in- cludes first the ability to assess lesions by anatomic distribution, and second by morphology.3–5 Anatomic Localization This includes the consideration of the following patterns: diffuse vs focal or clustered; central (peri- bronchovascular) vs peripheral (subpleural or peri- fissural); and upper vs lower lung distribution. Most importantly, nodules also need to be characterized by their relation to secondary lobular anatomy allow- ing a distinction between centrilobular nodules and those that predominantly involve the lobular periph- ery, including the interlobular septa3–5 (Fig 1). For example, diseases such as sarcoidosis that localize within or adjacent to lymphatics predomi- *From the Division of Pulmonary and Critical Care Medicine (Drs. Raoof and Amchentsev), New York Methodist Hospital, Brooklyn, NY; the Department of Radiology (Drs. Vlahos and Naidich), Tisch Hospital, New York University Medical Center, New York, NY; and the Department of Radiology (Dr. Goud), Brigham and Women’s Hospital, Boston, MA. No financial or other potential conflicts of interest exist for any of the authors. Manuscript received January 9, 2006; revision accepted January 14, 2006. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (www.chestjournal. org/misc/reprints.shtml). Correspondence to: Suhail Raoof, MD, FCCP, Chief, Pulmonary and Critical Care Medicine, Medical Director, Respiratory Ther- apy Department, New York Methodist Hospital, Pulmonary Division, 506 Sixth St, Brooklyn, NY 11215; e-mail: Sur9016@nyp.org CHEST Chest Imaging for Clinicians www.chestjournal.org CHEST / 129 / 3 / MARCH, 2006 805 by Kim Yong Mi on September 13, 2006 www.chestjournal.orgDownloaded from nate in those regions in which lymphatics are most extensive, specifically along the pleural and fissural surfaces, within the interlobular septae, and along the peribronchovascular axial interstitium (Fig 2). Diseases that are primarily hematogenous in origin, such as miliary infections or hematogenous metasta- ses, give rise to nodules that are randomly distrib- uted throughout the secondary lobule, with the greatest profusion in the lung bases (Fig 3). These patterns are clearly separate from nodules that result from inhalational disorders such as occur in patients with endobronchial spread of infection or hypersen- sitivity pneumonitis (HP), in which nodules are predominantly centrilobular in distribution, sparing the lobular periphery (Fig 4, 5). Morphologic Characterization This includes assessing a number of characteristics including whether nodules are as follows: uniform or variable in size; sharply or poorly marginated6–8; solid or subsolid in density (so-called ground-glass opacities) [Fig 5]5; or have a so-called tree-in-bud appearance (Fig 4).9 Additionally, nodules may ei- ther be calcified, as occurs in fungal disease, or cavitary, as is seen, for example, in patients with septic emboli, metastatic disease, or Langerhans cell histiocytosis (LCH).10 It should be emphasized that many of these characteristics are best evaluated on high-resolution CT scan images. With the introduction of multide- tector CT scanners, it is now possible to routinely, prospectively reconstruct both thick and thin sec- tions through the lungs after a single breathhold, provided that the initial data are acquired using appropriately thin collimation. This approach also enables the use of high-definition, multiplanar re- constructions, the use of which may be of value in further characterizing lung nodules.11 Multinodular HRCT Algorithm: A Step- Wise Approach The use of this algorithm begins by dividing CT scans into two broad arms based on the presence (group 1) or absence (group 2) of pleural or perifis- sural involvement (Table 1). Step 1 Group 1: Those cases in which a striking propor- tion of nodules demonstrate pleural or perifissural involvement characterize nodules as predominantly perilymphatic or lymphohematogenous in origin, constituting a separate arm of the algorithm (Table 1). The explanation for this pattern lies in the greater density of lymphatic channels seen in the interlobu- Figure 1. Secondary lobular anatomy. A side-by-side diagram- matic representation of two normal secondary pulmonary lobules. Secondary lobules represent fundamental anatomic units of the lung and are defined by centrilobular structures, including pulmonary arteries/arterioles and their accompanying bronchi/ bronchioles, and peripheral structures, including the pulmonary veins and lymphatics within the interlobular septae. As shown, most of these structures are � 1 mm in size and therefore, with the exception of the centrilobular arteries, lie below the resolu- tion of even HRCT scans. Most importantly, note that centri- lobular structures do not extend to the pleural or interlobular septal surfaces. As will be illustrated, knowledge of basic lobular anatomy is the key to differentiating between different etiologies of diffuse pulmonary nodules. Figure 2. Perilymphatic disease. A diagrammatic representation of the characteristic distribution of lung nodules in patients with perilymphatic disease. Note that nodules are preferentially sub- pleural, peribronchovascular within the axial interstitium, or along lobular septae. While this appearance is especially charac- teristic of nodular sarcoidosis, less commonly a similar pattern may also be seen in patients with silicosis or coal-workers pneumoconiosis. 806 Chest Imaging for Clinicians by Kim Yong Mi on September 13, 2006 www.chestjournal.orgDownloaded from lar septa and subpleural regions, including along the fissures. Step 2 Once nodules are characterized as predominantly perilymphatic or lymphohematogenous in origin, further assessment requires determining whether or not nodules are distributed diffusely or are patchy or clustered, with particular attention paid to the pres- ence or absence of the extent of axial interstitial involvement. It is recalled that the axial interstitium envelops the main pulmonary vessels and bronchi extending from the hilum outward toward the lung periphery.12 Step 3 If nodules prove to be clustered in a predomi- nantly subpleural/axial distribution, they are deemed to be perilymphatic in distribution (Fig 2). In this category, the main disease to be considered is sar- coidosis (Fig 6, 7).2,13,14 This diagnosis is further suggested by nodules that are typically ill-defined, frequently measuring only a few millimeters in size. Clusters of these nodules often have a “grainy” appearance and when sufficiently profuse may result in an appearance of poorly defined nodules or masses on corresponding chest radiographs (so- Figure 4. Bronchiolar disease. A diagrammatic representation of the typical appearance of bronchiolar inflammation resulting in so-called tree-in-bud opacities. These characteristically result in clusters of ill-defined nodules “attached” to adjacent branching or tubular structures due to extensive bronchiolar mucoid impac- tion. Most importantly, note that, unlike the situation in patients with either perilymphatic disease or random nodules, mucoid impacted bronchioles do not extend to the pleural, fissural, or septal surface. This pattern is nearly always due to infected secretions resulting from virtually any cause of acute or subacute bronchiolar infection. Figure 3. Random nodules. A diagrammatic representation of the characteristic distribution of randomly distributed nodules in patients with lymphohematogenous disease. Note that in distinc- tion with patients having predominantly perilymphatic disease, random nodules may been seen adjacent to all secondary lobular structures. Some nodules may also appear to be attached to pulmonary arterial branches (so-called feeding vessels). Random nodules are most commonly due to metastatic disease, and may vary considerably in size and edge characteristics. The differential diagnosis most importantly includes miliary infection. Lym- phangitic carcinomatosis, while hematogenous in origin, is easily distinguished from random metastatic nodules by the presence of characteristically thickened interlobular septae, preferentially involving the lung bases, and usually associated with asymmetric hilar adenopathy and pleural effusions. Figure 5. Centrilobular disease. A diagrammatic representation of the distribution of diseases that predominantly affect the centrilobular portion of secondary lobules, excluding those dis- eases that result in predominantly mucoid impaction due to infected secretions. The most common cause of diffuse centri- lobular disease is subacute HP. This characteristically results in poorly defined, poorly marginated ground-glass opacities. Similar to tree-in-bud opacities, these rarely involve the pleural or fissural surfaces. While a number of different entities may result in predominantly centrilobular opacities, the differential diagno- sis most often includes RB/RB-ILD. In distinction with subacute HP, RB in particular is less extensive, typically upper lobe in distribution, and almost always occurs in smokers. www.chestjournal.org CHEST / 129 / 3 / MARCH, 2006 807 by Kim Yong Mi on September 13, 2006 www.chestjournal.orgDownloaded from called alveolar sarcoid). When coalescent, these may simulate progressive massive fibrosis. Ancillary find- ings include a predominant upper lobe distribution, focal air-trapping due to bronchiolar obstruction, and diffuse adenopathy, often calcified. Calcified nodules may also be present in later stages of the disease. The most important differential diagnoses for this pattern of disease are silicosis and coal worker pneumoconiosis.15,16 In both of these occupational diseases, perilymphatic nodules are the primary ab- normality, typically involving the mid and upper lung fields. While these entities may simulate the appear- ance of sarcoidosis, they are usually easily diagnosed when correlated with clinical history.15 This includes other rare occupational lung disease, for example, siderosis, that may also simulate the appearance of sarcoidosis.17 While lymphangitic carcinomatosis may result in perilymphatic nodules, in fact, CT scan findings are most often characterized by markedly thickened nodular interlobular septae usually asymmetrically involving the lower lobes and usually associated with adenopathy and effusions.18 Nodules, when present, tend more often to be well-defined and are often associated with discrete feeding vessels, further identifying them as hematogenous in origin. Lym- phangitic carcinomatosis rarely mimics findings that are characteristic of sarcoidosis. Step 4 If nodules prove to be diffuse instead of clustered, they are properly considered to be random in distri- bution (Table 1). By definition, true random distri- bution will lead to nodules being identified along pleural and fissural surfaces as well as along the axial Figure 6. Perilymphatic disease: sarcoidosis. An HRCT scan of a 1-mm section at the level of the carina shows innumerable ill-defined small nodules clustered in the mid-portions of both lungs with relative sparing of the anterior aspects of both upper lobes. Note that these preferentially involve the left major fissure (arrow on left lung) as well as the walls of the peripheral airways (curved arrow on right lung). Table 1—HRCT Algorithm for Multinodular Disease* * CWP � coal workers pneumoconiosis; MAI�M avium intracellulare; MTB �M tuberculosis; PMF � progressive massive fibrosis. 808 Chest Imaging for Clinicians by Kim Yong Mi on September 13, 2006 www.chestjournal.orgDownloaded from interstitium. However, in distinction from primarily perilymphatic disease, random nodules may also be identified in even greater numbers when dispersed randomly throughout the lungs. Included in this category most importantly are hematogenous metastases.19 Unlike nodules in pa- tients with sarcoidosis, metastatic nodules tend to be smooth, well-defined lesions (Fig 8, 9).8 However, a wide variety of morphologic appearances has been noted. In a study20 comparing the HRCT scan features of pulmonary metastatic lesions with au- topsy findings, while nodules most often proved to have well-defined margins (38% of cases), nodules with well-defined irregular margins, poorly defined smooth margins, and poorly defined irregular mar- gins could be identified in 16%, 16%, and 30% of cases, respectively. While nodules range from a few millimeters to � 1 cm, they are frequently similar in size. A basilar predominance is typically noted due to preferential blood flow to the lung bases. Individual nodules may have “feeding vessels” consistent with their hematogenous origin. On HRCT scans, a con- nection between nodules and the adjacent pulmo- nary vessels (ie, the mass-vessel sign) may be seen in approximately 75% of cases.21 Nodules may also be either cavitary or surrounded by a “halo” of ground- glass attenuation, which is typical of hemorrhagic metastases such as those due to choriocarcinoma.22 Features of lymphangitic cancer may also be present, which again is consistent with a hematogenous origin of disease.19 It should be noted that the reported incidence of malignant disease as a cause of multiple pulmonary nodules has been shown to vary greatly, from as low as 10% to as high as 58% in some surgical series.23 In 133 patients with a known malignancy who under- went video-assisted thoracoscopy for multiple pul- monary nodules, 64% proved to have at least one malignant nodule.24 A number of malignancies can result in a miliary pattern, rendering differential diagnosis more problematic. This includes tumors, such as renal cell carcinoma, head and neck cancers, and testicular tumors, that have their primary venous drainage in the lungs.25 The differential diagnosis includes a number of additional entities that result in random nodules. The most important of these is miliary infection (Fig 10).26,27 In fact, while differentiation between miliary infection and a miliary tumor may be impossible to determine by imaging features alone, in general, close correlation with the clinical history renders these diagnoses relatively straightforward. Miliary metastases are frequently due to metastatic thyroid cancer, renal cancer, and melanoma, among other cancers, while larger less profuse metastases tend to be adenocarcinomas in adults, typically originating from the lung, breast, or the GI tract.19,28 Less commonly, diffuse nodules may be identified in patients with septic emboli, invasive fungal infec- tions, and pulmonary vasculitides.29 These entities frequently result in cavitary nodules, some with a distinct “halo” of ground-glass attenuation,22 and have even been described in patients with organizing pneumonia.30 Despite similarities between these en- tities and routine metastatic disease, it should be emphasized that the numbers of nodules identified in these cases usually fail to meet the criterion of “too many nodules to count,” with the differential diag- nosis again further aided by close clinical correlation. Step 5 Group 2: In distinction with the patterns de- scribed in patients in group 1, group 2 includes those Figure 7. Perilymphatic disease: sarcoidosis. An HRCT scan of a 1-mm section through the right mid-lung in a different patient than the one in Figure 6 shows evidence of innumerable ill-defined small nodules. Note that these tend to be clustered with relative sparing of the right upper lobe anteriorly and clearly preferentially lie adjacent to the right major fissure (arrow), along pleural surfaces, and along central vascular structures (arrow- heads). This distribution of nodules is rarely seen in any other disease. www.chestjournal.org CHEST / 129 / 3 / MARCH, 2006 809 by Kim Yong Mi on September 13, 2006 www.chestjournal.orgDownloaded from patients in whom no or very few nodules are perifis- sural or subpleural in distribution. Anatomically, these nodules are grouped together as being centri- lobular in distribution.4 By definition, these entities primarily involve centrilobular bronchioles and/or their accompanying pulmonary artery branches. An- atomically, these structures taper peripherally, stop- ping 5 to 10 mm short of the pleural or interlobular septal surfaces and consequently fail to involve pleu- ral and fissural surfaces (Table 1). As will be dis- cussed, these nodules typically fall into the following two broad categories: those with a “tree-in-bud” configuration; and those that appear as amorphous “ground-glass” nodules. Step 6 Once nodules are characterized as being primarily centrilobular in distribution, further assessment re- quires determining whether or not these have a tree-in-bud configuration. Tree-in-bud opacities are characterized by the appearance of centrilobular micronodular branching structures that end several millimete