Patients enrolment and characteristics
Our prospective study received institutional review board approval (INMI 3/150207) by the local ethical committee and all patients provided written informed consent.
The 50 patients (17 women and 33 men, ranging from 21-63 years of age with a median age of 47) who met our study criteria were referred to undergo HRCT and MRI imaging. The entry criteria for patients were as follows: (a) a chest X-ray with pulmonary abnormalities, (b) culture-proven pulmonary tuberculosis in culture from sputum (also induced) or bronchoalveolar lavage (c) absence of contraindication to MR examinations (ex. cardiac pacemakers, cochlear implants), (d) MRI obtained within 24 hours of the CT examination, to avoid divergent results during therapy.
All patients who did not have AIDS or additional concomitant infectious diseases were undergoing TB treatment. Altogether we performed 60 CT and MRI examinations, because 10 patients were also examined by CT and MRI at follow- up. However we analyzed the latter examinations to increment the number in our series, not to evaluate the effectiveness of therapy
Low dose HRCT was performed on all subjects using a helical four-channel MDCT scanner (Light Speed QX/i General Electric Medical System, Milwaukee, Wis). Unenhanced HRCT was obtained from the apex to the base of the lung at the end of inspiration, with 1-mm collimation, a high resolution algorithm, and 10 mm spacing. A specific mediastinum reconstruction algorithm was employed, and the images were obtained on lung and mediastinal settings.
To minimize the radiation required for obtaining diagnostic scans, the following scanning parameters were selected: tube current 70 mA, with 100 kV .
Parallel imaging MRI was performed with a 1, 5-T system (Signa Excite, General Electric Medical System, Milwaukee, Wis, USA), a maximum gradient strength of 33 mT/m, and a slew rate of 120 mT/m/s, using a six-channel body phased array coil system.
The examinations were performed with expiratory respiratory and diastolic gating. When pulsation was less vigorous, pulsation artifacts were reduced [1, 6, 8]. In agreement with the literature data , we preferred to use respiratory gating which allowed for continuous breathing instead of multiple breath hold acquisitions, also because the shifts of the lung parenchyma relative to the slice level are reduced. We performed MRI in expiration because the expiration phase is longer than the inspiration phase and signal intensity increases with deflation . Even when MRI was performed in expiratory respiration and CT at the end of inspiration, there was no significant discrepancy between the breathing positions of the images.
An axial T2-weighted Fast Recovery Fast Spin-Echo (FR FSE T2) FAT SAT was used with the following parameters: Echo Time, 90 msec; Repetition Time, 2500 msec; Echo Train Length, 14; bandwidth, 50; slice thickness, 5 mm; slice gap, 2 mm; field of view, 42 cm; matrix size, 288 × 224, reconstructed to 512.
Fat saturation sequences are very effective because the attenuated fat signal of the thoracic subcutaneous tissue reduces the ghosting artifacts of the ventral chest wall  and also increases conspicuity of fluids .
This sequence provides good image quality, and with imaging times of about 120 seconds we obtained enough slices to assess the entire lung.
The in-room time, including positioning the patient on the examination table, was approximately 15 minutes.
Both CT and MR images, all made anonymous, were directly displayed on the monitors of a picture archiving and communication system (PACS 5.1 Kodak, Rochester, NY, USA) with a window setting appropriate for lung parenchyma and mediastinum (pixel 2048 × 2560, display gradation 1021 (10-bit), maximum brightness 750cd/m2, LCD display device 54 cm). The readers were asked to assess presence, location and extension of pulmonary TB.
According to the standardized nomenclature for parenchymal findings on CT, consolidation was defined as a homogenous increase in lung parenchyma attenuation that obscures the margins of the vessels and airway walls (an air bronchogram may be present); a nodule was defined as a round lesion with a diameter of 3 cm or smaller; ground glass pattern was defined as a homogenous, hazy area of increased attenuation without obscuration of bronchovascular margins (an air bronchogram may be present); cavitation was defined as a gas-filled space, contained or not contained within a pulmonary consolidation, mass, or nodule. Tree in bud appearance was defined as a linear branching structure with more than one contiguous branching site. Furthermore, we assessed interstitial changes, in particular miliary, bronchial wall and peribronchial tissue thickening. Pleural and mediastinal lymph node involvement was also assessed.
Pleural effusion was defined as free-flowing pleural fluid producing sickle-shaped opacity (in most cases posteriorly) and loculated fluid collections as lenticular opacities in a fixed position. Pleural effusions with a volume of 15 ml or more can be detected with CT, however pleuritis sicca is not visible on CT scans. Lymph nodes were considered enlarged when they were greater than one centimeter on the short axis. Since there are no established MRI criteria to define parenchymal lung findings, we adopted the CT criteria. Regarding the pleura, MRI can detect subtle signal abnormalities that might be consistent pleuritis sicca. Concerning lymph nodes, we also assessed nodal signal intensity compared with thoracic wall muscle. Previous reports correlated histological data with MRI features in tuberculous lymphadenopathy , including signal intensity on unenhanced MR. Based on MRI findings, lymph node types could be defined according to the presence and degree of granuloma formation, caseation/liquefaction necrosis, fibrosis and calcifications. Signal intensity may differ depending on the stage of evolution: i) slight hyper-intensity may reflect lymphoid hyperplasia related to inflammation, ii) high hyper-intensity is suggestive of liquefactive necrosis, and iii) central isointensity associated with peripheral hyper-intensity may reflect caseosis.
The MRI findings to be assessed were previously established by consensus to avoid bias in individual interpretation.
All MR images were independently analyzed by two board-certified radiologists (VS, EBR, both with 10 years of experience in clinical MR imaging and 25 years of experience in chest imaging). The observers were unaware of CT results to avoid interpretation bias. Since CT is considered the gold standard technique, all CT images were considered as reference scans and were analyzed in a randomised order by the same radiologist in consensus, two months after analyzing the MR images.
Then, they directly compared MRI with CT examinations in consensus to verify the presence, distribution and characteristics of pathological features. In divergent cases, MRI and CT were re-examined to determine which imaging technique was correct. Disagreements in image scorings were resolved by consensus. MRI artifacts were graded as minimal (barely visible), moderate (clearly visible, but not interfering with evaluation) and severe (compromising evaluation). Particular attention was given to determining whether these artifacts interfered with the diagnostic value.
For CT and MR images, each parenchymal finding was scored on a scoring sheet using the following sliding scale of relative certainty: 0 = definitely negative; 1 = probably negative; 2 = indeterminate; 3 = probably positive; 4 = definitely positive. To calculate the MRI detection rate for each finding, we only considered those that were scored as 0 (definitely negative) and 4 (definitely positive).
Furthermore, for both imaging techniques, we classified each lung by zone: upper, middle, and lower, resulting in a total of six zones per patient. The upper zones were defined as areas of the lung above the level of the carina; the middle zones as areas between the level of the carina and the origin of the inferior pulmonary veins; and the lower zones as areas below the origin of the inferior pulmonary veins. Each zone had approximately the same number of sections. We scored lung zone involvement by using a four-point scale: 0 = no involvement, 1 = < 25%, 2 = 25%-50%, 3 = > 50%.
Statistical analyses were performed using the SPSS/PC+ version 11 (SPSS, Chicago, Ill). A p value lower than 0.05 indicated a statistically significant difference.
The degree of agreement between observers interpreting chest MRI was determined by using pair-wise kappa statistics as follows: very good, k value > 0.81; good, k value 0.80-061; moderate, k value 0.60-041. A separate kappa value was calculated for each sign that was reviewed.
K statistics were also calculated to analyze the agreement between MRI and CT and a detection rate for each finding; in this analysis, results were dichotomized as definitely positive or not.
The chi- square test was used to compare the proportion of images demonstrating different scores of involvement for each zone of the lungs depicted by MRI and CT.