Role of diffusion weighted MR-imaging in the evaluation of malignant mediastinal lesions

Conducted studies showed that the ADC (apparent diffusion coefficient) values of malignant mediastinal lesions are significantly lower than those of benign lesions. Investigators determined cut-off ADC values to differentiate the two; concluding that ADC value is a promising noninvasive, imaging parameter that helps assess and characterize mediastinal tumors. Taking this a step forward, the primary objective of our prospective study was to investigate the potential of DW-MRI (diffusion-weighted magnetic resonance imaging) to characterize malignant mediastinal lesions using their ADC values. Thirty-three patients that underwent MRI of the chest with DWI and latter pathologically diagnosed with a malignant mediastinal lesion were included in this study. Lesions’ ADC values were measured and correlated with the histopathological results. The statistical significance of differences between measurements was tested using the one-way ANOVA (analysis of variance) test; p values equal to or less than 0.05 were considered significant. There was no statistically significant difference between the ADCmean values of the histopathological groups of lesions assessed with the overlap of their ADCmean values. The average ADCmean value of NHL (non-Hodgkin lymphoma) was evidently lower than that of HD (Hodgkin disease) with no overlap between their ADCmean values. DWI failed at characterizing one lesion in this study as a malignant tumor, namely an immature teratoma (germ-cell tumor). Again DWI could not be used to evaluate a mass, latter pathologically diagnosed as an angiosarcoma, because of its overall hemorrhagic nature showing no definite non-hemorrhagic soft tissue components. The aforementioned results did not differ considerably when minimum ADC was used instead of mean ADC. There was no statistically significant difference between the ADC values of the malignant mediastinal lesions evaluated. However, regarding lymphoma subtypes, our limited sample study of lymphoma suggested a considerable difference between the ADC values of Hodgkin disease and non-Hodgkin lymphoma.


Background
The mediastinum comprises the thoracic compartment anatomically bounded by the thoracic inlet superiorly, the diaphragm inferiorly, the posterior sternal border anteriorly, and posteriorly by the vertebral column. It is an intricate segment of the thorax that contains vital intra-thoracic structures such as the heart and great vessels, trachea and main bronchi, esophagus, thymus, venous and lymphatic structures, and nerve tissue. Lesions of the mediastinum span a wide histopathological and radiological spectrum. Both anatomical information (location) and internal characteristics (composition) are essential to formulate the differential diagnosis of a mass in the mediastinum and therefore determine therapeutic options [1][2][3].
A cross-sectional imaging-based classification that divides the mediastinum into anterior (prevascular), middle (visceral), and posterior (paravertebral) compartments was recently published by the International Thymic Malignancy Interest Group (ITMIG). Contents of each mediastinal compartment, as well as the most commonly encountered lesions, are reviewed in Table 1 [4].
Computed tomography (CT) is considered the imaging modality of choice for evaluating most mediastinal masses; however, the role of magnetic resonance (MR) imaging continues to expand given its high soft-tissue contrast. It is superior to CT in differentiating between cystic and solid masses, identifying cystic and solid components within complex lesions, and distinguishing thymic hyperplasia and normal thymus from thymic epithelial neoplasms and other neoplasms-the result is added diagnostic specificity or virtual biopsy of the lesion. Assessment of preoperative relationships with the pericardium, heart cavities, spinal cord, and vascular involvement is a common indication of MR imaging of a mediastinal lesion [6].
With the recent advances in MR systems, DW-MRI (diffusion-weighted magnetic resonance imaging) of the thoracic cavity has become possible with fast imaging time that minimizes the effect of gross physiological motion from respiration and cardiac movement. DWI, a non-contrast functional MR imaging technique, allows the analysis of tissue characteristics based on the diffusivity of water molecules within tissues. Quantitative assessment of a tumor is possible by calculating its apparent diffusion coefficient (ADC) value which is inversely correlated with tissue cellularity. Hence, DWI has been proposed as a cancer biomarker with the diagnostic potential to distinguish benign from malignant tumors because of the tendency of the latter to show more restricted diffusion, i.e., lower ADC values (Table 2) [9][10][11][12].
Studies, conducted over the past few years, showed that the ADC values of malignant mediastinal lesions are significantly lower than those of benign lesions and determined cut-off ADC values to differentiate the two; concluding that ADC value is a promising noninvasive, imaging parameter that helps assess and characterize mediastinal tumors. Further studies were recommended [11,[13][14][15][16][17].
Thereby, the primary objective of this prospective study was to further investigate the potential of DW-MRI to characterize malignant mediastinal lesions using their ADC values; in an attempt to improve non-invasive approaches by which patients with mediastinal lesions are diagnosed/managed. On searching the literature, no studies, aiming primarily to differentiate between the various pathological types of malignant mediastinal lesions using DWI, were found.

Subjects
Institutional ethical clearance was taken before conducting this prospective study. Written consent was obtained from patients or their authorized representatives. Forty-eight patients (29 males and 19 females; age range: 2 to 73 years; mean age = 40 years) presenting Patients with contraindications to MRI including pacemakers, cochlear implants, cerebral aneurysm clips, ocular metallic foreign body, and bullets or shrapnel near great vessels or vital organs.
History of previous biopsy/treatment of the mediastinal mass.

MRI of the chest with DWI (image acquisition and evaluation)
1. Image acquisition: MRI of the chest with DWI was done for all patients with a 1.5 Tesla unit (Achieva; Philips Medical Systems, Best, The Netherlands) using a 16-channel phased array Pitfalls of DWI T2 -Shine through effect E.g. a cyst will demonstrate high signal intensity on T2-weighted images and on diffusion images obtained at high b values but also on the ADC map. Unlike a cyst, a region with truly restricted diffusion will demonstrate low signal intensity on the ADC map.
Slow-flowing blood within a region may demonstrate the signal intensity characteristics of a highly cellular lesion. This pitfall is readily observed on DWI of a hemangioma. In this situation, interpretation of the findings obtained with other MR sequences is very helpful. The following was recorded: Lesions' location (anterior, middle, and posterior mediastinum), morphological features (e.g., shape and margin) as well as the extent and relations to adjacent structures. Signal intensity on the T1, T2, and STIR WI relative to that of muscles in the same pulse sequence. Signal intensity on the high b value (b = 1000 s/ mm 2 ) DWI and corresponding ADC map. Associated MR imaging findings, e.g., pleural effusion.

Quantitative assessment:
ADC values were measured by placing a ROI (region of interest) within the lesion on the trace ADC maps. ROIs were positioned within areas visually judged to be the most restricted (excluding obviously cystic/necrotic areas). To avoid image selection bias, three ROIs were placed on at least three sections (cranial, middle, and caudal portions) within the lesion whenever possible and average mean and minimum ADC values (ADC mean and ADC min ) were calculated.

Histopathological diagnosis
Taking both lesion accessibility and patients' general condition into account, patients underwent U/S or CT-guided biopsies, bronchoscopy, mediastinoscopy, or thoracotomy.
Consequently, 33 patients pathologically diagnosed with a malignant mediastinal lesion were included in the study; 21 males (63.6%) and 12 females (36.4%) with age range 10 to 73 years (mean age 42.18 years).

Statistical/data analysis
Data was coded and entered using the statistical package SPSS version 21. It was summarized using descriptive statistics: number and percentage for qualitative values and mean and standard deviation and median and interquartile range for quantitative variables. One-way ANOVA (analysis of variance) test was used to compare means. p values equal to or less than 0.05 were considered significant. Data was presented by Box-Plot graphs, bar charts, histograms, pie charts as well as tables.
-Since lymphoma and bronchogenic carcinoma were the two most prevalent lesions in our study, the following was considered: -Average ADC mean of lymphoma was evidently lower than that of bronchogenic carcinoma; however, no statistically significant difference was found between their ADC mean values (p value > 0.05).   -Each histopathological group was studied distinctly and the following was recognized: -Among the nine patients, pathologically diagnosed as lymphoma; five had non-Hodgkin lymphoma (NHL) and four had Hodgkin lymphoma (HD) (Figs. 5 and 6). The average ADC mean value of NHL was evidently lower than that of HD with no overlap between their ADC mean values on the Box-and-Whisker plot (Fig. 7). However, the statistical significance of this important finding could not be assessed and a cut-off value between these two could not be determined because of the limited number of cases involved. -Similarly, the ten bronchogenic carcinoma cases in this study were histopathologically classified into non-small-cell lung cancer (NSCLC 60%) and small-cell lung cancer (SCLC 40%) (Fig. 8).
Average ADC mean value of SCLC (0.855 ± 0.324x10− 3 mm 2 /s) was lower than that of NSCLC (0.936 ± 0.380 × 10− 3 mm 2 /s) but their ADC mean values showed substantial overlap (Fig. 9). -DWI failed at characterizing one lesion in this study as a malignant tumor, namely an immature teratoma (germ cell tumor) (Fig. 10). Its soft tissue component showed facilitated diffusion with an ADC mean (2.44 × 10− 3 mm 2 /s) higher than previously reported cut-off ADC values, hence, it was misdiagnosed as a benign lesion. -Again DWI could not be used to evaluate the mass that was later pathologically diagnosed as an angiosarcoma because of its overall hemorrhagic nature showing no definite non-hemorrhagic soft tissue components (Fig. 11). -Minimum ADC of each lesion was taken into account when attempting to characterize the different malignant mediastinal lesions included in this study. However, the aforementioned results did not differ considerably when minimum ADC was used instead of mean ADC.

Discussion
Influenced by the constantly evolving approach to non-invasive patient management in the era of functional imaging, investigators demonstrated the diagnostic potential of DW-MR images and calculated ADC values to differentiate between benign and malignant tumors including those of the mediastinum. Hence, serving as a virtual biopsy and in some cases preventing unnecessary diagnostic intervention. Conducted studies showed that the ADC values of malignant mediastinal lesions are significantly lower than those of benign lesions and determined cut-off ADC values to differentiate the two.
Taking this a step forward, the main purpose of our prospective study was to investigate the potential of DW-MRI to characterize malignant mediastinal lesions using their ADC values.
The study included 33 patients with a mediastinal mass identified on CT that underwent MRI of the chest with DWI and were later histopathologically diagnosed with a malignant mediastinal lesion. Patients were grouped according to their histopathological diagnosis.
There was no statistically significant difference between the ADCmean values of the histopathological groups of lesions assessed. This is in concordance with the study carried out by Tondo et al., 2011 who found substantial overlap in ADCmean values of all examined malignant lesions (n = 30), namely bronchogenic carcinoma, thymic carcinoma, and malignant teratoma [16].
Moreover, the 32 malignant mediastinal lesions studied by Nasr et al., 2016 included lymphoma, bronchogenic carcinoma, invasive thymoma, and metastases. No significant difference between their ADCmean values was found as well [11].
Likewise, the 30 malignant mediastinal tumors in Abdel Razek et al., 2009s study were lymphoma, bronchogenic carcinoma, invasive thymoma, and angiosarcoma and again difference between their ADCmean values was insignificant with evident overlapping of the ADC values of lymphoma, bronchogenic carcinoma, and thymoma [13].
These findings may be attributed to a relatively small sample size and/or heterogeneity of the histopathological subtypes of malignant mediastinal lesions evaluated.
In agreement with Nasr et al., there was no statistically significant difference between the ADC mean values of lymphoma and the ADC mean values of the rest of the histopathological groups conjointly. Tondo et al., 2011 reported similar results to ours where no significant difference was found between bronchogenic carcinoma (adenocarcinomas) and malignant mediastinal lesions in their study [11,16].
On comparing the ADCmean values of lymphoma and bronchogenic carcinoma, lesions most frequently reported that the ADC of lymphoma is significantly lower than that of squamous cell carcinoma of the head and neck [14,[18][19][20][21][22].
Anterior mediastinal lesions "most importantly lymphoma, thymic, and germ cell tumors" in our study were entitled to be evaluated separately and again difference between their ADC values was statistically insignificant. Yabuuchi et al., 2015 conducted a retrospective study in the search of significant parameters to characterize anterior mediastinal tumors; one of the parameters evaluated was tumors' ADC value. Malignant anterior mediastinal lesions included in their study were thymic epithelial tumors (malignant thymoma and thymic carcinoma), lymphoma, and germ cell tumors. In agreement with our results, there was no significant difference between the ADC of these histopathological subtypes at both initial and validation studies [23].
Regarding the pathological subtypes of lymphoma; in the retrospective study of mediastinal lymphadenopathy in children by Abdel Razek et al., 2015, the average ADCmean of NHL was lower than that of HD-results similar to ours-but the difference between the two was insignificant. On searching the literature, no attempts at using DWI to differentiate between lymphoma subtypes, in particular, were found [24].
A meta-analysis of 34 studies (conducted in the years 2007 to 2014) involving 2086 patients with pulmonary lesions was recently performed by Shen et al., 2016. One of their objectives was to evaluate the role of ADC in characterizing subtypes of lung cancer. Pooled ADC values of SCLC were significantly lower than those of NSCLC, which to an extent agrees with our study (results more or less similar to ours). The potential histopathological rationale might be that SCLC has high tumoral cellularity, large nuclei, and almost no cytoplasm, all of which restrict the diffusion of water molecules thereby reducing ADC values [25].
Limited data are available on quantitative assessment of thymic epithelial tumors (TETs) by using DW-MRI from small cohorts of studies which considered various anterior mediastinal tumors. Not to mention that those studies did not attempt to differentiate thymomas based on WHO and Masaoka-Koga classifications by using ADC. Recently, a study by Abdel Razek et al. involving 30 patients with TETs has demonstrated the ability of ADC in differentiating low-risk from high-risk tumors, and early from advanced disease. Mean ADC values of high-risk thymomas and advanced stage TETs were significantly lower than those of lowrisk thymomas and early-stage TETs respectively. They determined cut-off mean ADC values of 1.22 mm 2 /s and 1.25 mm 2 /s below which a high-risk thymoma is indicated. In a relatively similar study by Priola et al., accurate cut-off mean ADC values of 1.309 × 10− 3 mm 2 /s and 1.243 × 10− 3 mm 2 /s were determined below which a high-risk thymoma and an advanced stage TET are indicated respectively. Moreover, mean ADC of B3 thymoma was significantly lower than that of B2 thymoma. ADC levels were significantly associated with disease-free survival of patients with a recurrence rate being higher for patients with ADC ≤ 1.299 × 10 −3 mm 2 / s. To conclude, ADC helps to differentiate high-risk from low-risk thymomas and may be used as a prognostic indicator of recurrence, yet further studies are needed to validate these results [26,27]. conducted prospective/retrospective studies to quantitatively assess and differentiate benign and malignant mediastinal lesions using DWI. They determined highly sensitive and specific cut-off ADC values below which a malignant mediastinal lesion is indicated; 1.15 × 10− 3 , 2.21 × 10− 3 , 1.25 × 10− 3 , 1.39 × 10− 3 , and 1.56 × 10− 3 mm 2 /s respectively. One lesion, namely a malignant teratoma had an ADC mean evidently higher than these cut-off ADC values (ADCmean = 2.44 × 10− 3 mm 2 /s) and was hence mistaken for a benign lesion. This may be explained by the fact that when the mass was surgically excised, it proved malignant only by the presence of microscopic foci of immature neuroepithelial component [11,13,14,16,17].
The one lesion pathologically diagnosed as an angiosarcoma was overall hemorrhagic in nature with no sizeable non-hemorrhagic soft tissue components seen on the conventional MR images rendering a qualitative assessment of the DWI and ADC measurement not possible. This was based on the fact stated by Qayyum, 2009: blood demonstrates signal characteristics of highly cellular soft tissue lesions; a pitfall readily observed on DWI [7]. A number of authors recommended that minimum ADC is used instead of mean ADC when quantitatively assessing a lesion since it theoretically reflects the area of highest tumor cellularity and unlike mean ADC, it would not underestimate tumor cellularity in the context of a largely necrotic tumor [28]. Accordingly, a minimum ADC of each lesion was taken into account when attempting to characterize the different malignant mediastinal lesions included in this study. However, the aforementioned results did not differ considerably when minimum ADC was used instead of mean ADC. Nevertheless, most published studies that evaluated mediastinal lesions using DWI referred to mean and not minimum ADC values.

Conclusion
Although highly sensitive and specific cut-off ADC values to distinguish benign and malignant mediastinal lesions were published, in this study, DWI could not be used to characterize the histopathological subtypes of malignant mediastinal lesions. However, regarding lymphoma subtypes, our limited sample study of lymphoma suggested a considerable difference between the ADC values of Hodgkin disease and non-Hodgkin lymphoma.