Diffusion-Weighted Imaging of the Lung Cancers: Preliminary Evaluation of Capability for Detection and Subtype Classification in Pulmonary Adenocarcinomas on Comparison with STIR Turbo SE Imaging

Introduction: Recently, image quality and diagnostic capability of chest MR imaging have been improving according to the advancement of MR system and sequence and the utilization of contrast media, and potential advantages of chest MR imaging have also been addressed in many literatures (1,2). In addition, it has been suggested that diffusion-weighted imaging (DWI) could be useful for assessment of primary malignancy, lymph node and/or distant metastases, as well as detection of additional benign and/or malignant tumors (3-5). In addition, the apparent diffusion coefficient (ADC) values add important information to findings obtained with conventional MR imaging. In chest MR imaging, DWI and ADC values are also suggested as useful for prediction of tumor activity such as subtype classification in pulmonary adenocarcinoma (4). On the other hand, short TI inversion recovery (STIR) turbo spin-echo (SE) imaging sequence is also useful for detection and clinical stage assessment in non-small cell lung cancer and other malignancies, and for screening of malignant tumors compared with T1 weighted imaging and T2 weighted imaging (6,7). In addition, there was a report that the STIR sequence was useful detection and diagnosis of lung cancer as well as thin-section MDCT (7). For these reasons, the STIR is an important sequence on chest MR imaging, and the sequence is also widely used in whole-body MR imaging. Under these circumstances, we hypothesized that both DWI and STIR were useful sequence for the detection of lung cancers and the subtype classification of pulmonary adenocarcinomas. The purpose of this study is to evaluate the capabilities of DWI for detection and subtype classification in patients with pulmonary adenocarcinoma, and to directly compare the capabilities for detection and subtype classification with STIR. Materials and Methods: Thirty-two patients (14 males, 18 females; mean age 65.3 years, range age 45-81 years) diagnosed as pulmonary adenocarcinoma by surgical and pathological examinations were enrolled in this study. MR imaging was performed with a 1.5 T superconducting magnet (Gyroscan Intera; Phillips Medical Systems) using a four-channel sensitivity encoding (SENSE) body coil. STIR was obtained by using a centrically-reordered multishot blackblood STIR turbo SE sequence with SENSE (TR = 2–3 msec, TEeff = 8 msec, TI = 165 msec, echo train length = 27, slice thickness = 5 mm, slice gap = 1.5 mm, NEX=2, matrix size = 256 × 256, reconstruction matrix size = 512 × 512, field of view = 320 mm, reduction factor = 4) and DWI was by using the sequentially-reordered half-Fourier single-shot STIR spin-echo echo-planar imaging sequence (TR = 5000 ms, TE = 70 ms, TI = 180 ms, echo train length = 41, slice thickness = 5 mm, slice gap = 1.5 mm, NEX=5, b-values 0 and 1,000 sec/mm2, matrix size = 96×96, reconstruction matrix = 256×256). For the analysis of the cancer detection capability of DWI and STIR, two chest radiologists without access to any patient information independently assessed DWI and STIR for the presence or absence of pulmonary cancers in random order by means of a 5-point visual scoring system. The final score for each cancer was then decided by consensus of the two readers. For quantitative assessment of DWI and STIR on detectable cancers, regions of interest (ROIs) were placed over each pulmonary cancer detected on DWI and STIR, and acquired ADC values on DWI and the contrast ratio for each cancer and the muscle (CR) on STIR. To compare the pulmonary cancer detection capability of DWI and STIR, both interobserver agreements were evaluated with the Kappa coefficient (κ) and the detection rates of the both sequences based on final scores were compared with each other by means of McNemar’s test. To compare ADC values and CRs among histological types of pulmonary adenocarcinomas for detectable cancers on each sequence, the analysis of variance (ANOVA), followed by Fisher’s protected least significance difference (PLSD) test was used. Finally, to compare the capability of quantitatively assessed DWI and STIR for differentiate BAC from others and adenocarcinoma with mixed subtypes from that except BAC component, ROC based positive test were performed. When ADC values and CRs were not able to be measured because of no visible cancers, the values were assumed to be 0. And the results were tested by means of McNemar’s test with the feasible threshold values of both sequences adapted for the highest accuracy value of quantitative assessments. Results: According to the results of pathological examination, 33 adenocarcinomas (mean diameter, 27.6 mm; range 5-69 mm) were diagnosed. These adenocarcinomas comprised ten BACs, 17 adenocarcinomas with mixed subtypes, and six adenocarcinomas except BAC component. Interobserver agreements for DWI and STIR were substantial (DWI: κ = 0.78; STIR: κ = 0.72) and detection rate of DWI and STIR were 85% (28/33) and 100% (33/33), respectively. Detection rate of DWI was significantly lower than that of STIR. The results of comparison of ADC values and CRs were shown in Table 1. Mean ADC values of BAC, adenocarcinoma with mixed subtypes and adenocarcinoma except BAC component were 0.0012, 0.0014, and 0.0012, respectively. There were no significant differences between ADC values and pathological subtype classifications (p>0.05). On the other hand, mean CRs were 0.63, 1.17 and 1.29, respectively. CRs of BAC were significant lower than those of adenocarcinoma with mixed subtypes and those of adenocarcinoma except BAC component (p<0.05). Based on the results of ROC based positive test of quantitative distinguishing BAC from others, the feasible threshold values of qualitatively assessed DWI and STIR were determined as 0.0012 and 0.8, respectively. When threshold values for differentiate BAC from others were adapted, the diagnostic capabilities are shown in Table 2. The sensitivity (56.5 [13/23] %, p<0.05) and accuracy (63.6 [21/23] %, p<0.05) of DWI were significantly lower than those of STIR (sensitivity: 95.7 [22/23] % and accuracy: 90.9 [30/33] %). On the other hand, based on the results of the results of ROC based positive test of quantitative differentiate BAC with mixed subtypes from that except BAC component, the feasible threshold values of qualitatively assessed DWI and STIR were determined as 0.0014 and 1.2, respectively. When adapted these threshold values (ADC=0.0014, CR=1.2), there were no significant difference between the diagnostic capabilities of two sequences (p>0.05). Representative cases are shown in Figure 1. Conclusion: The detection rate and the differentiation capabilities of pathological subtype classification for pulmonary adenocarcinomas of DWI were lower than those of STIR, and therefore the application of chest DWI is necessary to consider its’ limited capability for detection of malignant nodule in routine clinical practice.