OBJECTIVE. The purpose of our study was to evaluate the utility of 3-T MRI of the lung for differentiating inflammation- and fibrosis-predominant lesions in the usual and nonspecific types of interstitial pneumonia.
SUBJECTS AND METHODS. The subjects were 26 patients (10 men, 16 women; mean age, 57 ± 9 [SD] years; 16 with nonspecific interstitial pneumonia; 10 with usual interstitial pneumonia) who underwent 3-T MRI of the lung and surgical biopsy. A total of 54 biopsy sites were classified histopathologically into two groups: inflammation predominant and fibrosis predominant. After a T2-weighted triple-inversion black blood turbo spin-echo (TSE) sequence, dynamic MRI was performed with a T1-weighted 3D turbo field-echo sequence (coronal images with 2.5-mm slice thickness) before and 1, 3, 5, and 10 minutes after IV contrast injection. The chi-square test was used to compare differences in signal intensity on T2-weighted triple-inversion black blood TSE MR images and visually assessed enhancement patterns at dynamic MRI for the inflammation- and fibrosis-predominant sites.
RESULTS. Inflammation-predominant specimens were obtained from 31% (17 of 54) of the biopsy sites. Inflammation-predominant biopsy sites had an early enhancement pattern (82%, 14 of 17 sites, p < 0.001) on dynamic studies and high signal intensity (53%, nine of 17 sites, p = 0.001) on T2-weighted triple-inversion black blood TSE images.
CONCLUSION. Multiphase dynamic enhancement studies with a turbo field-echo sequence and T2-weighted triple-inversion black blood TSE images on 3-T MRI appear to be useful for differentiating inflammation- and fibrosis-predominant lesions.
Idiopathic interstitial pneumonia (IIP) is a heterogeneous group of nonneoplastic disorders resulting from lung parenchymal damage by varying combinations of inflammation and fibrosis. An international standard was established in 2002 in a consensus statement by the American Thoracic Society and European Respiratory Society [1, 2]. In the statement, the primary role of high-resolution CT was confined to differentiating patients with typical findings of idiopathic pulmonary fibrosis (histologically usual interstitial pneumonia [UIP]) from those with the less-specific findings associated with other types of IIP [3, 4]. Surgical lung biopsy is advised for patients with suspected IIP who do not have classic high-resolution CT features of idiopathic pulmonary fibrosis in the absence of a contraindication . Histopathologic distinction between inflammation- and fibrosis-predominant lesions is considered a major determinant of clinical characterization and prognosis of IIP [6, 7]. On high-resolution CT, disease patterns sometimes overlap between sites of active inflammation and fibrotic lesions; thus similar patterns of diseased segments on CT can have different responses to treatment .
An inherent feature of MRI is high tissue contrast; thus MRI is generally expected to be useful for differentiation of active inflammatory and fibrotic lung lesions. In an initial report , the presence of high-signal-intensity lesions was considered a useful predictor of treatment response and clinical outcome. Subsequent studies focused on the enhancement pattern of IIP lesions on MRI. Lesions with active alveolitis had prominent enhancement, whereas fibrotic lesions did not [10, 11]. Despite the advantages of MRI for differentiating inflammation- and fibrosis-predominant lesions, the merits of MRI over CT have been questioned in terms of utility in the evaluation of parenchymal lung lesions because MRI has inferior anatomic resolution and is compromised by inevitable artifacts due to motion and susceptibility effects [12, 13].
Technical developments in MRI and refinement of pulse sequences have improved the quality and speed of imaging [14–19]. In addition, high-field-strength MRI has increased signal-to-noise ratios during thoracic imaging . High-field-strength MRI for lung imaging has been found to have sensitivity comparable with that of CT in the detection of diffuse lung disease . These properties have facilitated the use of MRI to evaluate the dynamic enhancement patterns of lung lesions and to depict in detail the anatomic features of interstitial lung disease [22–25]. The purpose of our study was to evaluate the utility of 3-T lung MRI for differentiating inflammation- and fibrosis-predominant lesions in UIP and nonspecific interstitial pneumonia.
Subjects and Methods
Our institutional review board approved this prospective study, and informed consent for MRI was obtained from all patients. From November 2005 to July 2006, 59 patients with clinically suspected IIP were seen in an outpatient clinic by an experienced chest physician (11 years of experience in the diagnosis and management of IIP). From these 59 patients, we excluded 26 patients who had typical clinical and high-resolution CT features of UIP. Thus we included 33 consecutively registered patients who fulfilled the clinical and radiologic criteria for a diagnosis of IIP [1, 2] and who were candidates for surgical lung biopsy and underwent thoracic high-resolution CT and 3-T MRI. No patient in this group had a contraindication to MRI. These patients had no history of previous diagnosis of diffuse interstitial lung disease, known connective tissue disease, or corticosteroid use.
Biopsy specimens were obtained from multiple lung foci by use of video-assisted thoracoscopic surgery. Biopsy sites were chosen in consideration of high-resolution CT findings (areas containing ground-glass opacity irrespective of the presence of reticulation on coronal images). After pathologic evaluation of the 33 patients, seven patients with other than UIP or nonspecific interstitial pneumonia were excluded from data analysis. Three of the seven patients had cryptogenic organizing pneumonia, two had respiratory bronchiolitis–interstitial lung disease, and one each had desquamative interstitial pneumonia and lymphoid interstitial pneumonia. Thus we included 26 patients (10 men, 16 women; mean age, 57 ± 9 [SD] years; range, 37–70 years; mean body weight, 60 ± 8 kg; range, 44–79 kg) with UIP (n = 10) or nonspecific interstitial pneumonia (n = 16) in whom the expected pathologic findings were fibrosis and inflammation. Nine men and one woman were smokers (mean, 22 pack-years; range, 9–38 pack-years).
Lung MRI at 3 T
All MRI studies were performed with a 3-T system (Achieva, Philips Medical Systems) equipped with a gradient system capable of ensuring a maximum gradient amplitude of 80 mT/m, a rise time of 0.2 milliseconds, and a slew rate of 200 T/m/s. A cardiac coil (SENSE, Philips Medical Systems) with a six-coil element was used for coronal imaging of the thorax. After acquisition of T2-weighted triple-inversion black blood turbo spin-echo (TSE) images of the thorax, dynamic MRI was performed with a T1-weighted 3D turbo field-echo sequence before and after IV injection of contrast medium.
A breath-hold T2-weighted triple-inversion black blood TSE sequence with cardiac gating was performed with the following parameters: TR/TE, 1,200/1,800 (two R-R intervals); TEeff, 60 milliseconds; echo-train length, 21; field of view, 420 mm; acquisition matrix size, 256 reconstructed to 512; slice thickness, 3 mm; interslice gap, 10 mm; coronal orientation. A double-inversion blood-nulling preparation pulse was applied at the R-wave trigger to suppress blood signal with an inversion delay of 697.6 milliseconds. The preparation pulse was followed by a third inversion pulse to generate spectral presaturation by inversion recovery contrast enhancement for fat suppression. This fat suppression was based on selective excitation of the lipid-bound proton by application of a frequency-selective inversion pulse. One slice acquisition per breath-hold was continued for 15 slices to cover the whole lung from back to front. The imaging time per breath-hold was 12 seconds, giving rise to a total acquisition time of 3 minutes. A sensitivity-encoding factor of 2 with one signal averaged was applied in the phase-encoding direction to increase acquisition speed.
Dynamic enhancement study of the lung parenchyma was performed with a T1-weighted 3D multishot turbo field-echo sequence. The sequence was optimized with a turbo field-echo factor of 60, turbo field-echo shots of 62, and a radial turbo direction with the following acquisition parameters: 2.7/1.34; flip angle, 10°; number of signals averaged, 1; field of view, 420 mm; acquisition matrix size, 192. Fat suppression was achieved with a spectral presaturation attenuated by inversion recovery method with an inversion delay of 90 milliseconds. Before IV injection of contrast medium, a slab in the coronal orientation covering the whole thorax was obtained in the back-to-front direction in one breath-hold. The slab thickness was 175 mm, and the slab was divided into 70 partitions to produce 70 images with 2.5-mm in section thickness. The imaging time per breath-hold was 22 seconds. To increase acquisition speed, a sensitivity-encoding factor of 2 was applied in the phase-encoding direction. An additional four slabs of images were obtained 1, 3, 5, and 10 minutes after manual IV bolus injection of gadopentetate dimeglumine (0.2 mL/kg of Magnevist, Bayer HealthCare). The parameters were the same as those used for the initial unenhanced series. All MR image data were directly interfaced to our PACS system (Pathspeed or Centricity 2.0, GE Healthcare), which displayed all image data on four monitors (1,536 × 2,048 image matrices, 8-bit viewable gray-scale, 60-ft-lambert luminescence). Dynamic enhanced images were viewed on the monitors.
Two chest radiologists unaware of clinical and pathologic information assessed all images together and made decisions on findings by consensus. For comparative study with pathologic results, biopsy sites on MR images, where the signal intensity of MR images was analyzed, were chosen in consideration of the biopsy sites on coronal CT scans (areas containing ground-glass opacity irrespective of the presence of reticulation), where biopsy had been recommended. The signal intensity (high, similar, or low in comparison with chest wall muscle) of parenchymal lung lesions on T2-weighted triple-inversion black blood TSE images was evaluated for each biopsy site.
Dynamic images were evaluated qualitatively and quantitatively. For qualitative analysis, enhancement pattern was visually assessed prospectively and classified into three categories at review of dynamic images: pattern 1, early enhancement and washout with discernible enhancement of peak enhancement at 1 or 3 minutes in dynamic study; pattern 2, slight enhancement with no discernible enhancement at specific time point throughout dynamic phases; pattern 3, delayed persistent enhancement with discernible peak enhancement at 5 or 10 minutes in dynamic study.
For the quantitative assessment of dynamic enhanced MR images, the mean signal intensity of the lesions was measured on PACS monitors by one of the two chest radiologists. A circular region of interest with an area of 100 mm2 was placed at two or three surgical biopsy sites in the peripheral portion of the lungs. The areas were chosen in consideration of the high-resolution CT findings before biopsy and contained ground-glass opacity irrespective of the presence of reticulation. Each observer recorded the mean signal intensity. The percentage signal intensity of parenchymal lesions at a given time point (t) was determined with the following formula: percentage signal intensity = ([SIt–SIt0]/SIt0) × 100%, where SI is signal intensity. The following parameters were evaluated: Maximum peak enhancement was determined as the maximum percentage signal intensity throughout the time points. Time to peak was defined as tmax – t0 in minutes. Slope of enhancement was calculated as maximum percentage signal intensity/(tmax – t0). Extent of washout was defined as maximum percentage signal intensity minus percentage signal intensity at t10.
An experienced lung pathologist evaluated all pathologic specimens. Pathologic diagnoses were made according to the classification suggested in the 2002 consensus statement of the American Thoracic Society and European Respiratory Society . Pathologic findings were divided into two groups, inflammation predominant and fibrosis predominant, by integration of the following semiquantitative evaluation results. Although there was no standardized scoring system for quantification of fibrosis and inflammation in IIP, after discussion with our experienced lung pathologist, we developed this system by modifying an existing system . Semiquantitative evaluation included three types of histologic features: interstitial mononuclear cell infiltration, intra alveolar macrophages and exudates, and established fibrosis. Interstitial mononuclear cell infiltration was graded on a 5-point scale: 0, scanty; 1, scattered mononuclear cells; 2, patchy clustered mononuclear cells; 3, mononuclear cells with germinal center; 4, diffuse mononuclear cell infiltra tion. Intraalveolar macrophages and exudates was graded on a 3-point scale: 0, absent; 1, focal (< 10%) intraalveolar macrophage and exudates; 2, multiple patchy (> 10%) intra alveolar macrophage and exudates. Established fibrosis was quantified as percentage of dense fibrosis in a diseased area of a biopsy specimen (Figs. 1 and 2). If on histopathologic specimens, the biopsy site had a score of 2 or greater for interstitial mononuclear cell infiltration or intraalveolar macrophages and exudates and less than 10% fibrosis for established fibrosis, the biopsy site was classified inflammation predominant.
The chi-square test was used to compare the differences in signal intensity on T2-weighted triple-inversion black blood TSE MR images for the active inflammation and fibrosis-predominant groups. The chi-square test also was used to compare the frequency of the early enhancement pattern (pattern 1) on dynamic MRI for the active inflammation- and fibrosis-predominant histopathologic groups. The McNemar test was used to calculate and compare the diagnostic efficacies of the qualitative criteria of the early enhancement pattern (pattern 1) on dynamic MRI and of high signal intensity on T2-weighted triple-inversion black blood TSE MR images for differentiating inflammation- from fibrosis-predominant sites.
The Student's t test or Mann-Whitney U test, depending on the results of a normality (Kolmogorov-Smirnov) test, was used to compare quantitative parameters from time–intensity curves on dy namic MRI and semiquantitative histo patho logic scores between the inflammation-predo minant and fibrosis-predominant groups. In semi quant itative histopathologic scoring, this stati stical comparison was done to support the face validity of the scoring system. Spearman's rank correlation coefficients were used to estimate the relations between the quantitative parameters from dynamic MRI and semi quantitative histo pathologic scores. For all statistical analyses, p < 0.05 was considered significant.
The 26 patients had the following final histopathologic diagnoses: UIP (n = 10; 22 biopsy specimens; two inflammation- and 20 fibrosis-predominant lesions), group 1 nonspecific interstitial pneumonia (n = 5; 11 biopsy specimens; 10 inflammation- and one fibrosis-predominant lesions); group 2 nonspecific interstitial pneumonia (n = 2; five biopsy specimens; two inflammation- and three fibrosis-predominant lesions), and group 3 nonspecific interstitial pneumonia (n = 9; 16 biopsy specimens; three inflammation- and 13 fibrosis-predominant lesions). After acquisition and analysis of histopathologic specimens from multiple biopsy sites in these 26 patients, a total of 54 lesion sites were designated inflammation predominant (n = 17, 31%) or fibrosis predominant (n = 37, 69%).
Qualitative MRI Analyses
Nine (53%) of 17 biopsy sites with an inflammation-predominant lesion had high signal intensity on T2-weighted triple-inversion black blood TSE images (Figs. 3 and 4), whereas four (11%) of 37 biopsy sites with a fibrosis-predominant lesion had high signal intensity (p = 0.001). The frequencies of lesion sites with high signal intensity on T2-weighted triple-inversion black blood TSE images were significantly different for inflammation- and fibrosis-predominant sites.
Visual assessment of dynamic enhancement patterns showed that 14 (82%) of 17 inflammation-predominant lesion sites exhibited early enhancement (pattern 1) and that 34 (92%) of 37 fibrosis-predominant lesion sites exhibited slight enhancement (pattern 2, n = 19) or delayed persistent enhancement (pattern 3, n = 15) (Figs. 5A, 5B, 5C, 5D, 5E, 5F, 6A, 6B, 6C, 6D, 6E, 6F, 7A, 7B, 7C, 7D, 7E, and 7F). The frequency of the early enhancement pattern (pattern 1) was significantly higher in the inflammation-predominant group (p = < 0.001) (Table 1). The three lesion sites, which did not exhibit early enhancement on dynamic MRI but had the histologic finding of inflammation predominance, were two lesion sites of group 1 nonspecific interstitial pneumonia with a slight enhancement pattern and one lesion site of group 1 nonspecific interstitial pneumonia with a delayed enhancement pattern.
TABLE 1: Results of Qualitative Analysis of Dynamic MRI Enhancement and Washout Patterns (n = 54)
Inflammation Predominant (n = 17)
Fibrosis Predominant (n = 37)
3, Delayed persistent
The sensitivity, specificity, and accuracy of diagnostic criteria of the early enhancement pattern (pattern 1) on dynamic MRI were higher than those of high-signal-intensity lesions on T2-weighted triple-inversion black blood TSE MR images for differentiating inflammation-predominant and fibrosis-predominant sites, but the differences were not statistically significant (Table 2).
TABLE 2: Diagnostic Characteristics for Predicting Disease Activity According to Dynamic MRI Qualitative Criteria (n = 54)
The results of analysis of time–intensity curve parameters for dynamic MRI are summarized in Table 3. Inflammation-predominant lesion sites had higher percentage signal intensity at 1 minute (p = 0.033), shorter time to peak (p = 0.032), and faster slope of enhancement (p = 0.010) than fibrosis-predominant sites. The extent of delayed enhancement was positively higher in the fibrosis-predominant group, but the difference between the two groups was not statistically significant (p = 0.067).
Semiquantitative Pathologic Scores, Dynamic MRI Parameters, Clinical Outcome, and Correlation
The semiquantitative scores of the three histologic features for interstitial mononuclear cell infiltration, intraalveolar macrophages and exudates, and established fibrosis are summarized in Table 4. Scores for interstitial mononuclear cell infiltration and intraalveolar macrophages and exudates were significantly higher for active inflammation sites (p = 0.003 and p < 0.001, respectively). Established fibrosis scores were significantly higher for fibrosis sites (p < 0.001). Slope of enhancement had a positive correlation with the inflammatory histologic features for interstitial mononuclear cell infiltration (r = 0.314) and intraalveolar macrophages and exudates (r = 0.306). Established fibrosis had a negative correlation with slope of enhancement (r = –0.389) and a positive correlation with time to peak (r = 0.436) and delayed enhancement (r = 0.361) (Table 5).
TABLE 4: Pathology Scores (n = 54) in Inflammation- and Fibrosis-Predominant Biopsy Sites
TABLE 5: Correlation Between Histopathologic Score and Dynamic MRI Parameters
Percentage Signal Intensity at 1 Minute
Extent of Washout
Time to Peak
Slope of Enhancement
Interstitial mononuclear cell infiltration
Alveolar macrophages and exudates
Note—Numbers in parentheses are p calculated with Spearman rank correlation coefficients
It is important to discriminate between inflammation- and fibrosis-predominant lesions of IIP because treatment response and long-term survival rate can be predicted with histologic patterns (inflammation vs fibrosis) [6, 7]. Although patients with IIP often are treated with high-dose corticosteroids, such therapies are not universally effective and are associated with risk. Thus noninvasive imaging methods, which facilitate differentiation of inflammation- and fibrosis-predominant lesions of IIP, are needed for treatment decision-making and for prediction of treatment response.
In this study, qualitative analysis of dynamic T1-weighted 3D turbo field-echo MR images obtained at 3 T proved helpful for differentiating inflammation- and fibrosis-predominant lesions. Most (82%, 14 of 17 sites) of the inflammation-predominant lesions had an early enhancement pattern. Three inflammation-predominant lesion sites did not have early enhancement on dynamic MRI. We presume that those lesions were too subtle in terms of density or too small in extent to be detected as enhancing lesions during visual assessment of lesions on MR images.
From the practical point of view, visual assessment of dynamic enhancement pattern is easier than time–intensity curve analysis. Overall, the presence of a visually noticeable early-enhancing lesion was an accurate indicator of active disease with high positive and negative predictive values, 82% and 92%, respectively. At time–intensity curve analysis, inflammation-predominant lesions had a higher percentage signal intensity in the initial dynamic phase at 1 minute with faster time to peak and slope enhancement. Fibrosis-predominant lesions had delayed enhancement (positive percentage signal intensity difference in subtraction of percentage signal intensity at 1 minute from that at 10 minutes). Therefore, absence of early enhancement significantly decreases the probability that treatable or reversible lung disease is present and that corticosteroid therapy probably is not necessary for these patients.
Gaeta et al.  reported that pulmonary lesions associated with inactive chronic infiltrative lung disease do not exhibit enhancement but that most (82%, 14 of 17) active lesions exhibit noticeable enhancement on MR images. This finding was corroborated in our study. At quantitative analysis we identified early enhancement with a higher percentage signal intensity at 1 minute and faster slope of enhancement in the inflammation-predominant group. In our study, however, the fibrosis-predominant lesions (41%, 15 of 37 sites) exhibited enhancement in the delayed phase. Owing to the short period (3 minutes) of dynamic range covered in the previous study, the investigators might have not noticed the delayed enhancement of fibrotic lung lesions during the late redistribution phase after IV administration of MRI contrast medium.
The early enhancement pattern of inflammation-predominant lesions on dynamic MRI may be due to an increased extravascular interstitial space with abundant inflammatory cell infiltrates or promotion of neovascularization with increased angiogenesis in those lesions . Diffusible hydrophilic low-molecular-weight gadopentetate dimeglumine can freely cross the interendothelial pores of the microvasculature by diffusing into extravascular lung spaces and reentering the intravascular space [28, 29]. If active inflammation is associated with an increased volume of extravascular interstitial space and angiogenic activity, net movement of gadopentetate dimeglumine would remain greater in the extravascular space of the lung interstitium during the early equilibrium period approximately 3 minutes after contrast injection. In our study, at pathologic examination, this expected phenomenon of early enhancement of inflammation-predominant lesions was reflected by the presence of high scores for interstitial mononuclear cell infiltrates and intraalveolar macrophages and exudates at the corresponding lesion sites.
Another important pathologic feature was the presence of established fibrosis in the fibrosis-predominant group. This presence of fibrous tissue replacing interstitium might have contributed to the destruction of capillaries and thus impaired washout, resulting in delayed rather than early enhancement at fibrosis sites on dynamic MRI. Therefore, in our study, the most reliable MRI finding for predicting the presence of inflammation-predominant lesions was the presence of early and rapid enhancement. Moreover, the enhancement rate was found to have a positive correlation with inflammatory histopathologic score (interstitial mononuclear cell infiltration, intraalveolar macrophages and exudates) and a negative correlation with established fibrosis. The extent of enhancement per se did not contribute measurably to differentiation of inflammation- and fibrosis-predominant lung lesions because the latter lesions did exhibit delayed enhancement.
T2-weighted triple-inversion black blood TSE images had fairly good resolution and contrast for lung imaging with less motion artifact due to breath-holding, ECG gating, and decreased imaging time. High-signal-intensity lesions on T2-weighted triple-inversion black blood TSE images may represent increased water content due to the presence of inflammatory cells and exudates in the interstitium and alveolar air-spaces. It has been reported  that MRI at 3 T can be used to detect diffuse pulmonary disease with sensitivity close to that of helical CT. In our experience  with lung MRI at 3 T, the characteristics and the extent of parenchymal lesions were discerned with anatomic detail and without substantial intrinsic or extrinsic motion artifact.
Our study had several limitations. First, the analyses were executed per biopsy site. In clinical practice, the unit of disease treatment is the patient, not the individual lesion. However, our study started with the idea that in a given patient, pathologic specimens can have heterogeneous results. Second, the number of lesion sites with active inflammation was relatively small compared with the number of sites with fibrosis. Third, the patients who underwent surgical biopsy for IIP were mostly those who had somewhat atypical presentations of the disease. Patients with the typical clinical presentation of UIP usually do not undergo surgical biopsy. This fact might have contributed to selection bias. Fourth, the MRI protocols were designed with a patient breath-hold. Several patients might have had difficulty holding their breath, but those difficulties did not cause a crucial artifact.
Assessment of signal intensity on T2-weighted triple-inversion black blood TSE images and of multiphase dynamic enhancement studies performed with fast imaging technique (T1-weighted turbo field-echo sequence) at 3-T MRI is useful for evaluating IIP activity and enables prediction of the histopathologic features of predominant inflammation and fibrosis.
Presented at the 2006 scientific assembly of the Radiological Society of North America, Chicago, IL (SSQ05-03).
Supported by Samsung Medical Center Clinical Research Development Program grant CRDP CRS106-40-2.
American Thoracic Society, European Respiratory Society. American Thoracic Society/European Respiratory Society International Multidisciplinary Consensus Classification of the Idiopathic Interstitial Pneumonias. This joint statement of the American Thoracic Society (ATS), and the European Respiratory Society (ERS) was adopted by the ATS board of directors, June 2001 and by the ERS Executive Committee, June 2001. Am J Respir Crit Care Med 2002; 165:277-304
Nicholson AG, Colby TV, du Bois RM, Hansell DM, Wells AU. The prognostic significance of the histologic pattern of interstitial pneumonia in patients presenting with the clinical entity of cryptogenic fibrosing alveolitis. Am J Respir Crit Care Med 2000; 162:2213-2217
Ohno Y, Adachi S, Motoyama A, et al. Multiphase ECG-triggered 3D contrast-enhanced MR angiography: utility for evaluation of hilar and mediastinal invasion of bronchogenic carcinoma. J Magn Reson Imaging 2001; 13:215-224
Yamashita Y, Yokoyama T, Tomiguchi S, Takahashi M, Ando M. MR imaging of focal lung lesions: elimination of flow and motion artifact by breath-hold ECG-gated and black-blood techniques on T2-weighted turbo SE and STIR sequences. J Magn Reson Imaging 1999; 9:691-698
Lutterbey G, Gieseke J, von Falkenhausen M, Morakkabati N, Schild H. Lung MRI at 3.0 T: a comparison of helical CT and high-field MRI in the detection of diffuse lung disease. Eur Radiol 2005; 15:324-328
Nicholson AG, Fulford LG, Colby TV, du Bois RM, Hansell DM, Wells AU. The relationship between individual histologic features and disease progression in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 2002; 166:173-177