| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Thoracic Imaging |
1 From the Department of Clinical Radiology, University of Munich, Ziemssenstrasse 1, 80336 Munich, Germany (R.E., P.H., O.D., M.F.R., S.O.S.); and Department of Haematology/Oncology, University of Munich, Munich, Germany (C.T.R., H.O.). Received December 5, 2004; revision requested February 2, 2005; revision received November 15; accepted January 2, 2006; final version accepted February 6. Address correspondence to R.E. (e-mail: Roger.Eibel{at}t-online.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Materials and Methods: The institutional review board approved this study; prior consent was obtained. Thirty consecutive neutropenic patients (10 women, 20 men; mean age, 51 years ± 15 [standard deviation]; range, 25–75 years) with fever of unknown origin or clinical signs and symptoms of lung infection were examined with breath-hold single-shot half-Fourier turbo spin-echo MR imaging. To reduce image blurring and increase MR signal in the lungs, the echo time was shortened with generalized autocalibrating partially parallel acquisition (GRAPPA). Patients underwent thoracic CT (four detector rows and 1-mm section thickness [4 x 1 mm]; pitch, 6) as reference standard. Pulmonary abnormalities (ill-defined nodules, ground-glass opacity areas, and consolidation), their location and distribution, and lesion characteristics were analyzed at MR imaging by three readers, blinded to results of CT, in consensus. Frequencies were calculated for each feature; paired Wilcoxon rank sum test was used to examine whether differences between CT and MR imaging features were statistically significant (
< .05). Bonferroni adjustments were performed. Overall sensitivity, specificity, and positive and negative predictive values were determined.
Results: Twenty-two patients had pulmonary abnormalities at CT. In 21 (95%) patients, pneumonia was correctly diagnosed with MR imaging. One false-negative finding occurred in a patient with ill-defined nodules smaller than 1 cm at CT. One false-positive finding with MR imaging was the result of blurring and respiratory artifacts (sensitivity, 95%; specificity, 88%; positive predictive value, 95%; negative predictive value, 88%). There was no significant difference in lesion location and distribution.
Conclusion: With parallel imaging (GRAPPA technique) and fast MR imaging, detection of pulmonary abnormalities is almost as good as with CT. MR imaging has a slight disadvantage in its lower capability to assist in characterization of specific internal features, such as cavitations.
© RSNA, 2006
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Helical computed tomography (CT) is now regarded as the method of first choice for the detection and characterization of pulmonary nodules and infectious disorders (4). This technique, however, is associated with radiation exposure, which is of particular concern in young patients who require multiple follow-up studies. An important feature in reducing the radiation burden of CT studies was the implementation of low-dose techniques. Although preliminary observations were published by Naidich et al (5) in 1990, low-dose protocols became a routine part of screening and examinations for lung diseases in the mid and late 1990s (6,7). With low-dose protocols, the overall dose can be markedly reduced to about one-tenth of the normal CT dose when it is performed at 10 mAs instead of at 100 or 120 mAs; however, the radiation risks are not eliminated.
Pulmonary magnetic resonance (MR) imaging has become a more popular research tool in the last decade, and findings in studies indicate that MR imaging is capable of depicting a variety of pulmonary disorders (8,9). The major problems of pulmonary MR imaging result from the low proton density of normal lung tissue and the susceptibility artifacts caused by the extensive air-tissue interfaces of the parenchyma. Both factors lead to low signal intensity of the pulmonary parenchyma (10). In one study (11), the value of T2-weighted turbo spin-echo sequences has been documented to substantially improve image quality and lesion detectability. With the introduction of parallel acquisition techniques, a further improvement in image quality is possible by means of a shortening of the echo times (TEs) of single-shot sequences; this shortening results in decreased blurring artifacts and less signal decay caused by T2* effects (12).
To our knowledge, few studies have been published in which a direct comparison of pulmonary CT and MR imaging was made (13,14). These studies were performed with extremely long acquisition times of as long as 20 minutes. Thus, the purpose of our study was to compare parallel acquisition MR imaging with thin-section helical CT for depiction of pulmonary abnormalities suggestive of pneumonia in the immunocompromised patient.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The inclusion criterion for our study was a chest radiograph that was either normal or did not show abnormalities suggestive of pulmonary infection such as confluent opacification with air bronchograms; multiple ill-defined nodular, patchy, or confluent opacities; or progressive bilateral reticular opacities. Nonspecific findings might be present on the chest radiograph. Any suspicion of pneumonia at chest radiography was an exclusion criterion for this study. The study was approved by the Federal Office for Radiation Protection, Salzgitter, Germany, and by our institutional review board. Written informed consent was obtained from all patients after participation in the study had been fully explained. Minors younger than 18 years and patients with standard medical contraindications to MR examinations (eg, cardiac pacemaker, cochlear implants) were excluded.
In all patients, blood culture, bronchoalveolar lavage, and, in the case of consolidation depicted on the CT images, a transbronchial biopsy were performed to find out the cause of infection. In 10 patients, we found Aspergillus species pneumonia in six, cytomegalovirus pneumonia in two, and Pneumocystis jiroveci pneumonia in two. In the other patients, clinical diagnosis of pneumonia was established when the three following criteria were met: (a) development of new symptoms such as cough and fever; (b) clinical signs suggestive of pneumonia such as production of purulent sputum, tachypnea, or signs of consolidation at physical examination; and (c) one, two, or all of the following: ill-defined nodules, ground-glass opacity areas, or consolidation on CT images. When no pulmonary abnormality suggestive of pneumonia was found with CT imaging and the patient had ongoing signs and symptoms of pulmonary infection, the CT study was repeated in 3–6 days. Additionally, CT was used to follow up treatment. All of these follow-up examinations, however, were not part of this study.
CT Examination
CT was ordered as part of a routine clinical protocol in these scenarios: (a) the patient had fever of unknown origin for more than 48 hours and/or suspicion of pneumonia was raised by other physical and laboratory examinations, (b) further evaluation of abnormal but nonspecific findings on chest radiographs was needed, and/or (c) the chest radiograph was normal.
Patients with CT findings that were interpreted as normal were not excluded from the study because normal CT findings are necessary to evaluate the false-positive rates of the MR examination.
Unenhanced CT was performed within 48 hours of chest radiography with the patient in a supine position. A scanner (Somatom Sensation 4; Siemens Medical Solutions, Erlangen, Germany) was used with the following parameters: four detector rows and 1-mm section thickness (4 x 1 mm) with a spiral technique, pitch of 6, 0.6-mm reconstruction increment, 1.0-mm reconstructed effective section thickness, 120 mAs, 120 kV, and approximately 35–50-second scanning time. Scanning extended from the lung apices to below the costophrenic angles. The patients were instructed to hold their breath as long as possible and then start shallow breathing. Depending on the patient's size, 350–500 images were obtained. Images were reconstructed with a high-spatial-frequency algorithm for the lung evaluation and in a soft-tissue kernel for better detection of calcification and pleural effusion. The lungs were assessed in the lung window (window width, 1500 HU; window level, –600 HU); the calcifications and pleural effusions were assessed in the mediastinal window (window width, 450 HU; window level, 60 HU). Images were reviewed on a workstation (Sienet Magic View 1000; Siemens Medical Solutions).
MR Imaging
To avoid divergent results from rapid changes of pulmonary infectious manifestations during therapy, MR imaging was performed no later than 24 hours after the CT examination (mean interval, 2 hours). MR imaging was performed with a 1.5-T MR system (Magnetom Sonata Maestro Class; Siemens Medical Solutions), a maximum gradient strength of 40 mT/m, a maximum rise time of 120 µsec, and eight radiofrequency channels. For signal reception, a recently introduced commercially available dedicated 12-element integrated matrix coil system was used that consisted of one anterior and one posterior flexible coil, each with a set of six receiver elements. Outside elements on each side were combined to fit to the limit of the eight receiver channels. A T2-weighted half-Fourier single-shot turbo spin-echo sequence was used with the following parameters: TE, 27 msec; bandwidth, 488 Hz/pixel; echo spacing, 3.76 msec; section gap, 50%; and field of view, 256 x 256 (transverse) or 320 x 256 (coronal). The nonelectrocardiographically gated acquisitions were interleaved to cover the lung parenchyma and were performed in deep inspiration.
The value of T2-weighted turbo spin-echo sequences for detection of pulmonary diseases, especially pneumonia in immunocompromised patients, in comparison with helical CT, was described in earlier studies (14,15). The time of acquisition, however, in these studies was between 9 and 19 minutes. Therefore, in our study, the use of the half-Fourier single-shot turbo spin-echo technique reduced examination time. This sequence provides good image quality with short imaging times (14). To minimize artifacts from blurring and fast T2* decay, this sequence was combined with parallel acquisition techniques.
Parallel imaging was performed with an autosimultaneous acquisition of spatial harmonics–based generalized autocalibrating partially parallel acquisition algorithm (GRAPPA) (16). In contrast to the way other algorithms for parallel imaging such as sensitivity encoding function, correction of the parallel-acquired data is performed in the k-space domain and not in the image domain. This algorithm is less susceptible to artifacts from additional aliasing of tissue outside the field of view that are propagated into the center of the image. The necessary coil profiles are determined within the half-Fourier single-shot turbo spin-echo image itself with acquisition of additional reference lines in the center of the k-space, and this process is referred to as autocalibration. The GRAPPA parameters were set to a factor of two with 24 reference lines (12 additionally acquired autocalibration signal lines). For further reduction of the TE, the reference lines were acquired prior to the half-Fourier single-shot turbo spin-echo readout (external reference image).
Each 15-second breath hold allowed 50% of the thorax diameter to be included. Thus, the entire thorax could be imaged with a section thickness of only 6 mm and an in-plane spatial resolution of 0.8 mm in the coronal and transverse planes within four breath holds. The imaging time to include the thorax in our study was about 30 seconds for one imaging plane and about 60 seconds for transverse and coronal planes. When we included the positioning of the patient on the examination table, positioning of the coils, performance of the topogram, and performance of the examination sequences, the resulting room time in the MR imaging room was less than 15 minutes.
Image Analysis
All CT scans were analyzed at the monitor with window settings appropriate for lung parenchyma (window width, 1500 HU; window level, –600 HU). In addition, the raw data were reconstructed with a soft-tissue kernel, and the images were analyzed in the mediastinal window for detection of calcifications in the lung lesions and for the evaluation of pleural effusion (window width, 450 HU; window level, 60 HU).
According to the standardized nomenclature for parenchymal findings at CT, consolidation was defined as a homogeneous increase in lung parenchyma attenuation at CT or an area of signal intensity that obscures the margins of vessels and airway walls at MR imaging. A ground-glass opacity area was defined as an opacity at CT or an area of hyperintensity that did not obscure bronchovascular margins at MR imaging. A nodule was defined as a round lesion with a diameter of 3 cm or smaller, and a consolidation was defined as a lesion with a diameter larger than 3 cm (17).
The images were reviewed retrospectively by three chest radiologists (R.E. with 10 years, P.H. with 6 years, and S.O.S. with 9 years of experience in CT interpretation), who together analyzed all the images and made final decisions in regard to the findings with consensus. Because to date there are no established MR imaging criteria for the diagnosis of pneumonia, we used the definitions of lung nodule, ground-glass opacity, and consolidation used for CT image evaluation. CT and MR images were analyzed without demographic data and, in each group (CT and MR imaging group), in a randomized order to reduce recall bias effects from the remembering of patients' names or bias from knowledge of the findings from the other modality. In the first step, further explained later, we evaluated the CT scans 4 weeks after the analysis of the MR images so that the CT findings would not influence the radiologists' judgments of the MR imaging findings. In the second step, we evaluated the CT and MR images side by side without blinding of the patients' names, so that direct comparisons of the lesions on CT and MR images were possible.
For CT and MR images in the first step, areas of each lung were classified as upper, middle, and lower zones, and this classification resulted in a total of six zones per patient. Each zone had approximately the same number of sections. CT and MR imaging findings in every lung zone were reviewed in consensus by all three readers for the presence, size, and characteristics of nodules (number, size of every nodule measured in one direction, margins, presence of a halo sign [surrounding area of ground-glass opacity], cavitation, and presence of calcifications within the nodules), of consolidations (size measured in two perpendicular directions, existence of cavitation, and calcifications), and of ground-glass opacities (size measured in each lung zone was assigned a score on the basis of the percentage of ground-glass opacity involvement shown on the image by using a four-point scale [score 0, no involvement; score 1, <25% involvement; score 2, 25%–50% involvement; score 3, >50% involvement]). The overall percentage of involvement was calculated by averaging the scores from each of the images in the associated lung zone.
The margin of a nodule was described as smooth and sharply defined or as ill defined with blurred borders. In cases in which more than one nodule was present, the previously mentioned marginal criteria were determined for the most common type of lesion. After this assessment, pleural effusion and MR imaging artifacts were evaluated. The presence and severity of pleural effusion were assigned scores by using a four-point scale for both CT and MR images: score 0, no pleural effusion; score 1, effusion in the basal third of the thorax; score 2, effusion that reaches the middle third of the thorax; and score 3, effusion reaches the apex of the thorax. The MR images were assessed for quality as follows: score 1, no artifacts; score 2, a few artifacts; score 3, images are of diagnostic value but impaired by artifacts; and score 4, images are of no diagnostic value because of severe artifacts.
In the second step, we compared the previously mentioned parameters (size, margins, internal characteristics) in a consensus reading point by point and lesion by lesion on the CT and MR images, so that an exact comparison was made of the same abnormality with both modalities. Because of the lower spatial resolution of the MR imaging system, CT patterns such as a tree-in-bud appearance were not taken into account.
Statistical Analysis
From the previously mentioned first step, the frequencies of number, size, and internal characteristics for nodules, consolidations, and ground-glass opacity areas and their median size were determined. This evaluation was performed independently for the CT and MR images. From the second step of the evaluation, we determined the frequencies (ie, how often the size, the margins, and the internal characteristics corresponded or disagreed). With regard to size, we analyzed how often the abnormality was under- or overestimated on the MR images compared with that information for the standard of reference. We also analyzed how often the abnormalities contributed to the diagnosis of pneumonia and whether there were differences between CT and MR images that would result in false-positive and false-negative diagnoses of pneumonia with MR imaging. From the third step, we calculated the frequencies of pleural effusion and severity of artifacts at MR imaging.
When we applied the general formula for calculation of total sample size for a study by using the z statistic to compare proportions of dichotomous variables, the sample size was calculated for a test power of 0.8 (equivalent to a ß error of .2), an error of .05, and equal proportions of subjects in each group. On the basis of previous observations (18), it was assumed that the probability of finding pulmonary infiltrates by means of CT in patients as examined in this study would be .6. Since both CT and MR imaging of the lung had been performed in the same patients in this study, it was assumed that imaging findings with both modalities would correlate.
Correlation between imaging findings was calculated by means of the Spearman test on the basis of the comparison of numbers of inflammatory nodules per lung zone (upper, middle, and lower areas of the right and left lungs). The Spearman correlation was calculated separately for lung zones with and without evidence of nodules at CT. For CT and MR imaging, a rank was assigned separately to each lung zone, according to the number of nodules detected by using the respective imaging modality. Results of the Spearman calculations for lung zones with and without nodules were then added, and the sum was divided by two to account for different proportions of lung zones with and without nodules. The resulting Spearman correlation coefficient was then subtracted from one, the result was multiplied by the total sample size as calculated previously, and then this product was divided by two to account for the fact that subjects were the same for both CT and MR imaging.
The overall sensitivity, specificity, and positive and negative predictive values for diagnosis of pneumonia and for detection of the separate abnormalities were analyzed. By using the Kolmogorov-Smirnov test, we found a non-Gaussian distribution of our data. Therefore, the paired Wilcoxon rank sum test was used to examine whether differences between CT and MR imaging in the number, extent, size, margins and internal characteristics of nodules, consolidations, ground-glass opacity areas, and pleural effusions were statistically significant ( < .05). Bonferroni adjustments were performed. The problem of clustering that arose in the multiple nodules per patient and the fact that there is no independence between lesion detection and measurement on CT and MR images in the same patient were addressed by using the generalized estimating equations method, a multivariate analysis method for detection of changes in repeated-measurement data.
All data were analyzed by using dedicated statistical software (SPSS for Windows, version 12.0, SPSS, Chicago, Ill; SAS for Windows, version 8.2, SAS Institute, Cary, NC).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
One false-positive diagnosis was made in the evaluation of the MR images, because a small area of ground-glass opacity, misinterpreted as pneumonia, was in retrospect a result of blurring and respiratory artifacts. The sensitivity of MR images for detection of pneumonia was 95% (21 of 22); specificity, 88% (seven of eight); positive predictive value, 95% (21 of 22); and negative predictive value, 88% (seven of eight). After Bonferroni adjustment, there was no significant difference (P > .0085) in the description of lesion location and distribution.
The number of patients included in this study was 30, and the number of lung zones compared between CT and MR imaging was 180. Spearman correlation between CT and MR imaging findings was 0.9867 in lung zones without nodules at CT (n = 112, 62.2%) and 0.8986 in lung zones with nodules at CT (n = 68, 37.8%). Spearman correlation coefficient r was calculated as 0.9427, such that 1 – r was 0.0573. On the basis of the assumption that on CT images we would find lung nodules in 60% of the cases, we calculated that the total sample size would have to be N = 6223.4 if there were two separate groups of subjects without any correlation of findings, and detection of nodules at MR imaging would occur in 56.5% of all cases.
Because the Spearman correlation was high in this study, however, the total sample size necessary to detect a difference of at least 3.5% between CT and MR imaging decreased to N = 178.4. Thus, the sample of N = 180 lung zones with imaging findings compared between CT and MR imaging would have sufficed to find a statistically significant difference between the modalities, with a test power of 0.8, if the performance of MR imaging had been inferior by at least 3.5% and CT had depicted nodules in 60% of lung areas. Given the percentage of 37.8% of lung zones with nodules at CT in this study, an inferior performance of MR imaging would have been detected with a test power of 0.8 if we had found nodules in 34.4% or fewer lung zones on MR images.
Detection of Nodules and Assessment of Size of Nodules
Fourteen (47%) of 30 patients had no ill-defined nodules. In the remaining 16 patients, 259 nodules were detected in the six lung zones by means of evaluation of the thin-section CT scans, but only 207 nodules were detected at MR imaging (Figs 1, 2). On the MR images, 186 nodules had a correlation with the nodules on CT images; thus, there was a concordance in 186 (72%) of 259 nodules. When we analyzed lung zones separately (Table 1), with CT we were able to depict more lesions in every lung zone. By using the Wilcoxon rank sum test and after Bonferroni adjustment, the differences in the detection rate in each lung zone were not significant.
|
|
|
|
|
Because follow-up was not part of this study, we could not further focus on this aspect. Thus, the sensitivity of MR imaging for detection of ill-defined nodules larger than 10 mm in diameter was 100% (154 of 154). When we took into account all nodules independently of their size, the sensitivity of MR imaging was 72% (186 of 259), and the positive predictive value was 90% (186 of 207). Because no true-negative nodules could exist in this evaluation, no specificity and no negative predictive value could be calculated (Table 1). By using generalized estimating equation modeling, we did not find a relevant contribution of one variable (nodules on MR images) to the other (nodules on CT images). That means that the clustered nature of the multiple nodules was not relevant for detection of nodules with MR imaging in comparison with findings at CT.
Nodule Characterization
In 158 (85%) of 186 nodules that were concordantly identified on the CT and MR images, the description of the margin contour was identical with CT and MR images. In 28 (15%) of 186, the contour of nodules was depicted as being smoother on MR images than on CT images. Smaller irregularities delineated on CT images were not detectable at the MR imaging evaluation. In 169 (91%) of 186 cases, there was an agreement between CT and MR imaging findings concerning the presence (in 56 nodules) or absence (in 113 nodules) of the halo sign. In 15 of 17 nodules in which the findings were discordant, we tended to detect the halo sign for the corresponding nodule with MR imaging more often than with CT.
In two patients, pulmonary nodules with cavitation or the crescent sign were detected on thin-section CT images. With MR imaging, we were not able to delineate the small areas of air in the corresponding nodules. The analysis of the lung zones in the mediastinal window showed no calcifications in the nodules. In one particular patient, we could not differentiate between an active ill-defined nodule and a scar with thin-section CT (Fig 3). The MR imaging study showed low signal intensity values throughout the entire lesion, and this finding suggested the diagnosis of a scar. A follow-up examination with CT 14 days later showed no difference in location, shape, and density of this nodule. This finding supported the diagnosis of a scar without active infection.
|
|
Consolidation
In seven (23%) of 30 patients, nine consolidations could be found (Fig 4). With the assumption that there was an imprecision of measurement of ± 2 mm, an identical size was measured in seven (78%) of nine cases. In the remaining two (22%) of nine cases, the MR images demonstrated a slightly larger size of consolidations, but the size difference did not exceed 15% (mean diameter, 7.1–8.0 mm). This was caused by ground-glass opacity areas adjacent to the lesions and a consecutive imprecise definition of the marginal contour of the consolidation. In no case did this size difference exceed 10 mm. None of the consolidations demonstrated a cavitation on either CT or MR images. As with the nodules, no calcifications in consolidations could be found on the CT images. There was no significant difference between the two modalities in the detection rate and size measurements of the consolidations; the sensitivity for detection of consolidation on the MR images was 100% (nine of nine).
|
|
|
|
|
Pleural Effusion and MR Imaging Artifacts
The detection rate of pleural effusion and the description of effusion size were absolutely identical at CT and MR imaging. Twenty-two patients had no pleural effusion on the right side, two patients had a score 1 effusion, two patients had a score 2 effusion, and four patients had a score 3 effusion. On the left side, 25 patients had no effusion, four patients had a score 1 effusion, and one patient had a score 3 effusion.
Fifteen (50%) of 30 patients had no artifacts on the MR images. In 14 patients, we found score 1 artifacts, and this finding indicated that there was no image degradation and no influence on image evaluation. In only one (3%) patient of 30 was score 2 image degradation from artifacts found, but image analysis was still possible. Score 4 artifacts, which made image interpretation impossible, were not found in our study. The causes of these cases of image degradation included patient movement artifacts, motion artifacts from cardiac and vascular pulsation, ghosting artifacts, or blurring, or a combination of these factors.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
In 16 patients, ill-defined nodules could be visualized at CT; in 14 (88%) of 16 patients, they were visualized at MR imaging. In 14 patients, ground-glass opacity areas were on CT images, and in 16 patients, they were on the corresponding MR images. Seven patients had consolidations, which were equally detected by using CT and MR imaging. The rate of 73% positive findings with CT, although findings at chest radiography were normal or nonspecific, is in accordance with data in previous reports from the literature. In the study of Heussel et al (18) in neutropenic patients with fever that persisted for more than 48 hours despite empiric antibiotic therapy, CT findings were shown to be suggestive of pneumonia in up to 60% of the cases despite a normal chest radiograph.
Of 259 nodules detected with thin-section helical CT, only 186 (72%) were depicted with MR imaging. The mean size of lesions not visualized with MR imaging was 4 mm. All nodules larger than 10 mm were reliably detected with MR imaging. The different sensitivity values for CT and MR imaging data sets likely relate to the section thickness used with CT (1 mm) and MR imaging (6 mm). Even with advanced MR imaging technology employing parallel imaging with a GRAPPA acceleration factor of two, the section thickness could not be smaller than 6 mm if the acquisition had to be performed within one breath-hold period.
In a study by Fischbach et al (19), pulmonary nodules larger than 10 mm were equally well depicted with CT scans with 1.25- and 5-mm reconstruction intervals, whereas lesions smaller than 5 mm were significantly better depicted with a reconstruction interval of 1.25 mm (P < .05). By using 1.25 mm as the reference standard, the sensitivity values of 5-mm reconstruction intervals were 88% and 86% for two observers. In comparison with findings in our study, these results may indicate that the reduced detection rate with MR imaging may be related to the larger section thickness used with MR imaging.
Nodular margins and contours can be classified as smooth or irregular. Furthermore, the halo sign adjacent to pulmonary nodules is an important feature that refers to the condition in which a pulmonary nodule is surrounded by a peripheral area of ground-glass opacity at thin-section CT. In our study, MR and CT images were concurrent in depiction of nodular margins and halo signs in 85% and 91% of nodules, respectively. In most of the cases in which the findings were discordant, MR imaging tended to depict the halo sign more often. It remains to be proved in further studies whether these are false-positive findings or MR images are more sensitive than CT images in the depiction of a small halo surrounding a nodule. In two patients, pulmonary nodules with cavitation or the crescent sign were detected on CT images. With MR images, we were not able to delineate the small areas of air in the corresponding nodules. Because cavitation is an important aspect that can suggest the fungal nature of the infection, the failure of detecting cavitation seems to be an important disadvantage of MR imaging. To clarify whether the failure of detecting small areas of air in nodules and consolidations is a fundamental problem in MR image analysis or only in our study, further investigations are necessary.
In all 14 patients in whom ground-glass opacity areas were found at CT evaluation, the MR images were also positive for ground-glass opacity areas. In addition, in two further patients, ground-glass opacity areas were found with MR imaging without correlation at CT. In one of these two patients, the diagnosis was wrong, because a blurring artifact was misinterpreted as a ground-glass opacity area. In the other patient, a CT study performed 3 days later because of ongoing cough and sputum production showed a ground-glass opacity area in the same location in which a similar finding was positive on the previous MR image. Further studies are needed to evaluate whether the sensitivity of T2-weighted MR images for the detection of fluid is higher than is the sensitivity for the detection of increasing attenuation on CT images. In MR imaging, ghost and blurring artifacts may resemble a ground-glass appearance, thereby making the specificity inferior to that of thin-section CT. CT also may be at a disadvantage because of artifacts caused by involuntary breathing and cardiac pulsations, which may mimic pathologic lung conditions (20). Shortening of examination time in MR imaging and further improvements in sequences may potentially reduce these problems.
In our patients, the detection rate for consolidation with CT or MR imaging was identical. Results from CT and MR imaging may differ in regard to the size and extension of consolidation. In MR imaging, larger areas of ground-glass opacification surrounding the consolidations were frequently depicted. In no case did this size difference exceed 10 mm. None of the consolidations manifested with cavitation formation at CT or MR imaging.
The detection rates for pleural effusions and the description of their extension were identical with the two modalities. In two patients, pleural effusions were found on the T2-weighted MR images.
Our study had some limitations. We included only 30 patients, and therefore every finding or difference between the two modalities had a major effect on the statistical results. Therefore, confirmation of our findings with findings in studies with larger series of patients is important. In our study, three experienced readers reached a consensus decision. This approach was performed because the criteria for ground-glass opacity areas and consolidation are currently not well established in the reporting of findings on MR images. With consensus reading, we were able to avoid a bias from applying too much individual interpretation. Our study protocol did not include assessment of the capability of predicting bacterial, fungal, and viral infections with the thin-section CT and MR images. The limitations of CT in the differentiation of various infectious entities are well known (21). Nevertheless, early detection, characterization, and follow-up of pulmonary abnormalities that are highly suggestive of pneumonia are most important for treatment of immunocompromised patients. Our results indicate that MR imaging is sensitive in the detection of pulmonary abnormalities in this group of patients.
In a study by Leutner et al (14), these researchers indicated that, with MR imaging, they were able to show various features of opportunistic pneumonia that are already well known from CT, including cavitation, air crescent, cysts, and "reticular infiltrations." The same number of lesions was detected on MR and CT images in all patients in their study. Furthermore, the early detection of a necrotizing lesion by using the reverse-target sign, a hypointense rim in a hyperintense infiltration of the lung parenchyma, offers advantages in the immunocompromised host because this sign is a strong indicator of fungal pneumonia and could be detected with a higher prevalence by using MR imaging.
Our results differ from those of Leutner et al (14) in some aspects. The overall detection rate for lung nodules with MR imaging was not as good as it was with CT. The discrepancy between our results and those of Leutner et al may have been caused by the CT section thickness of 10 mm used in the trial of Leutner et al, whereas we employed a 1-mm section thickness, which allows the visualization of smaller lung lesions. To compare the diagnostic value of different imaging modalities, we consider it appropriate to use state-of-the-art techniques. None of our patients had cavitation or air crescent signs in the consolidations at CT; thus, the capability of depicting these types of lesions with MR imaging could not be proved. In two patients, we were not able to delineate the cavitations in nodules. Characterization of the lesions with MR imaging, as well as with CT, seems to be a problem.
Our study population was highly selected. We included only neutropenic patients with a high probability of pneumonia. It is possible that the rate of false-positive findings that occurred as a result of blurring is higher if there is a lower prevalence of pneumonia. In an older and otherwise more severely ill population, it can be expected that motion and breathing artifacts will result in a higher rate of image degradation. In our study population, we did not observe severe artifacts that would result in image degradation on the MR images.
In summary, pulmonary MR imaging can depict and help exclude pulmonary abnormalities that are suggestive of pneumonia in the immunocompromised patient. Therefore, this modality might be considered for diagnostic work-up in young patients for whom multiple CT examinations might be harmful. In the detection of ground-glass opacity areas and consolidation, MR imaging seems to be equal to thin-section CT. Imaging and motion artifacts are a problem in MR imaging and also in CT. Nodules smaller than 10 mm were detected with higher sensitivity on CT images. Larger nodules were depicted equally well with both modalities. The lower sensitivity for small nodules is likely the result of the different section thicknesses: 1 mm with CT and 6 mm with MR imaging. Therefore, it can be expected that the detection rate with MR data sets can be improved when thinner sections become available.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: GRAPPA = generalized autocalibrating partially parallel acquisition • TE = echo time
Author contributions: Guarantors of integrity of entire study, all authors; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, R.E., P.H., C.T.R., S.O.S.; clinical studies, all authors; statistical analysis, R.E., O.D., H.O., S.O.S.; and manuscript editing, all authors
Authors stated no financial relationship to disclose.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. Kumar, S. Basu, D. Torigian, V. Anand, H. Zhuang, and A. Alavi Role of Modern Imaging Techniques for Diagnosis of Infection in the Era of 18F-Fluorodeoxyglucose Positron Emission Tomography Clin. Microbiol. Rev., January 1, 2008; 21(1): 209 - 224. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. A. Altes, M. Eichinger, and M. Puderbach Magnetic Resonance Imaging of the Lung in Cystic Fibrosis Proceedings of the ATS, August 1, 2007; 4(4): 321 - 327. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
