¹Department of Radiodiagnosis and Imaging, PGIMER, India
²Department of Opthalmology, PGIMER, India
³Department of Histopathology, PGIMER, India

^*Corresponding author: Kumar A, Department of Radiodiagnosis and Imaging, PGIMER, Chandigarh - 160012, India, Phone: +91-7087009385; E-mail: ajay2509@gmail.com

Copyright: © 2021 Bansal A, et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Aim: The purpose of this study was to evaluate the utility of advanced MR techniques to improve the diagnostic ability to differentiate between these entities.

Material and Methods: A prospective evaluation of 26 patients (12 men and 14 women with a mean age of 34.8 years) presenting with orbital masses to our institute was done. MRI was performed in all patients with advanced techniques DWI, MRS and dynamic (DCE) on a 1.5 T scanner.

Results: ADC cut off value of 0.809 x 10-3 mm2/sec was determined using the ROC curve to differentiate malignant from benign lesions resulting in a sensitivity of 72.2% specificity of 75%, positive predictive value of 86.6% and Negative predictive value of 54.5%. Lesions with Tp < 141.5s showed a sensitivity of ~94.4% and specificity of ~87.5%, positive predictive value of 94.4% and Negative predictive value of 87.5%for malignancy. The lesions with Slope > 0.47 showed a sensitivity of ~100% and specificity of ~78%, positive predictive value of 66.6% and Negative predictive value of 100% for malignancy. Significant difference was seen between type I and type III curves with a p value ~0.002. Presence of choline peak had a sensitivity of ~ 62.5%, specificity of ~94.4%, positive predictive value of 83.3% and negative predictive value of 85% for malignancy.

Conclusion: Use of advanced MR imaging with inclusion of DW, MRS and perfusion parameters can significantly increase our diagnostic confidence in differentiating benign from malignant orbital lesions.

Lesions in the orbit can be classified either according to the location of the mass or its histological subtype [1,2]. Conventional imaging techniques i.e. Multidetector computed tomography (MDCT) and magnetic resonance imaging (MRI) do not have sufficient sensitivity and specificity to help in differentiating malignant masses from benign lesions, with a high chance of misdiagnosis in case of unexpected rare entities [3,4]. Addition of new MR imaging techniques, like dynamic contrast enhanced (DCE) MR sequences, diffusion weighted imaging (DWI) and magnetic resonance spectroscopy (MRS), to the original sequences can provide different forms of tissue contrast and may help in better characterization of orbital lesions [5].

This was a single institution prospective study which was approved by the Ethics Committee of the institute. A total of 26 patients with orbital mass (12 men and 14 women with a mean age of 34.8 years) were enrolled over a period of one year in the study. These patients had come to our institute with complaints of proptosis and/ or reduced visual acuity. The exclusion criteria for this study were contraindication for performing MRI study i.e. metallic implants, claustrophobia, and refusal to give consent and presence of large amounts of necrosis or calcification in orbital mass.

All studies were performed at 1.5T MRI unit (Siemens, Magnetom Aera; Siemens Healthcare, Erlangen, Germany) equipped with high performance gradient system with a head coil with the following protocol (Table 1). All the imaging data was systematically evaluated by two radiologists (one fellow resident and a second neuroradiologist with over 10 years of experience).

For DWI, all regions of interests (ROIs) were drawn manually. ADC values were calculated by adjusting the size of the ROI according to the size of the lesion. In lesions less than 2 cm in maximum diameter, one ROI was taken of .014 sq cm areas. In lesions that were larger than 2 cm, at least two or more ROIs were placed, avoiding the necrotic component of the lesion, and their mean (ADC_mean) values calculated. The ADC values were expressed in 10^-3 mm²/s. For MRS, Chemical shift imaging (CSI), using multiple voxels, or single voxel spectroscopy (SVS) was used according to size of the lesion and morphological appearance. Water and fat suppression was done automatically prior to all spectroscopic examinations using CHESS and other volume fat suppression techniques. For DCE, a coronal dataset including the lesion was acquired just before injection of the contrast and subsequently 6 datasets were acquired in a time period of 5 minutes after contrast injection. Multiple ROIs were applied and one ROI was selected which showed maximum enhancement pattern. The size of the lesion determined the size of the ROIs. The signal intensity (SI) of each slice of the dynamic sequence was determined from its mean pixel value. The representative ROI and corresponding time intensity curve (TIC) were obtained for each mass. -Three types of enhancement curves were obtained [4]: (i) The persistent pattern (type-I) - The lesion shows continuous progressive enhancement (ii) The plateau pattern (type-II) - In this pattern, there is a sharp rise in the enhancement, which then attains a plateau (iii) The washout pattern (type-III) - a rapid rise in enhancement followed by rapid decline with the final signal intensity <90% of the peak signal intensity. Each TIC was used to derive values for the baseline signal intensity (SI_pre), maximum signal intensity at the peak of enhancement (SI_max), and times corresponding to these signal intensities (T_pre, T_peak).Dependent parameters like peak enhancement (PH), enhancement ratio (ER_max) and Slopemax, were calculated using the above mentioned four independent variables with the help of the following formulas

Final diagnosis was achieved in all cases through histopathological examination after the MR imaging.

Statistical analysis: Discrete categorical data (sex, MRS, Biopsy) was represented in the form of either a number or a percentage (%); Continuous data that were not normally distributed were reported as median and Inter quartile ranges. The distribution of the variables was tested with the Shapiro-Wilk test. For medians of ADC _mean, group comparisons (groups were Biopsy positive/negative) were made with the help of Mann Whitney test. Comparison of means of 2 groups of SI_max, Tp,PH, ER, Slope of the subjects was done using Independent t-test. To assess the performance of MRS in predicting the disease, sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV) and diagnostic accuracy, keeping Biopsy as gold standard, were calculated using kappa test. Receiver operating characteristic (ROC) curves was calculated for SI (max), Tp, PH, ER, Slope and for ADC separately. Chi square or Fishers exact test was used to compare proportions which were dependent on their applicability. McNemar’s test was used to see differences in diagnosis between two modalities. Concordance and discordance was calculated by method of Kappa test of agreement. All the statistical tests were two-sided performed at a significance level of α=.05. Analysis was conducted using IBM SPSS STATISTICS (version 22.0).

The location of the masses is summarized in Table 2: Of the 26 cases which were included in the study, 18/26 (69.2%) orbital masses were histopathologically proven to be benign and 8/26 (30.7%) malignant. The histopathological distribution is given in Table 3. The means and standard deviations of ADC values of benign and malignant lesions were calculated separately and are shown in Table 4. ROC was drawn and area under the curve (AUC) was determined as ~0.854 (Figure 1). A cut off value of 0.809 x 10^-3 mm²/sec was determined using the ROC curve to differentiate malignant from benign lesions resulting in a sensitivity of72.2% specificity of 75%, positive predictive value of 86.6% and Negative predictive value of 54.5%.

The DCE parameters of the benign and malignant lesions are summarized in Table 5.

Out of these values Tp and Slope were found to have a significant difference between benign and malignant lesions. Tp had a p value of <0.001. ROC curve was drawn with the area under the curve 0.96 (Figure 2). A cut off value of 141.5 seconds was determined through the ROC curve and the lesions with Tp< 141.5s showed a sensitivity of ~94.4% and specificity of ~87.5% (Figure 2), Positive predictive value of 94.4% and Negative predictive value of 87.5% for malignancy. Slope had a p value of <0.001. ROC curve was drawn with the area under the curve 0.889 and cut off value of 0.47 (Figure 3). Lesions with Slope > 0.47 showed a sensitivity of ~100% and specificity of ~78%, positive predictive value of 66.6% and Negative predictive value of 100% for malignancy.

Time intensity curves were plotted for perfusion parameters (Table 6). All the lesions which demonstrated type I pattern were benign and all lesions demonstrating type III pattern were malignant [5,11]. Of the 10 lesions demonstrating type II curve, seven were benign and three malignant. The malignant lesions demonstrating type II curve were lymphoma, carcinoma NOS and one MPNST. A p value of ~0.002 was seen among type I and III curves. Significant difference between benign and malignant lesions was not seen using the values of SI(_pre), SI(_max), PH and ER.

MRS data of the patients was analyzed. The kappa value generated was 0.877 which implies almost perfect agreement of choline peak with malignancy (Table 7). With use of choline peak a sensitivity of ~ 62.5%, specificity of ~94.4%, positive predictive value of 83.3% and Negative predictive value of 85% for malignancy was obtained. Representative cases showing histopathological proven benign and malignant orbital mass evaluated with above parameters are shown in Figure 4-7.

There is a need to differentiate between benign and malignant orbital masses for better treatment planning of individual cases [6,7,8]. A study conducted by Ben Simon et al., showed poor sensitivity of both CT and MR features in differentiation of malignant from benignlesions [9].

In our study, we applied newer MRI sequences DWI, DCE-MRI and MRS to evaluate few objective criteria to differentiate the nature of the lesions. On histopathological evaluation, 69.2% (18) lesions in our study were benign and 30.7 (eight) malignant, which roughly corresponds to the ratio in the general population [2].

Malignant tumors have increased cellularity with large nuclei and reduced extra-cellular space, which leads to restriction of free water movement and a dark signal on ADC map [10]. Sepahdari et al. found that benign lesions had a higher ADC value than malignant A cut off value of 1.0 x 10^-3 mm²/sec was determined [5], which resulted in a sensitivity of 63%, specificity of 84% and accuracy of 71% for differentiating benign vs malignant lesion. Similar studies were done by Razek et al. and Fatima et al. [11,12], demonstrating the usefulness of ADC mapping in differentiating benign from malignant lesions.

In our study, cut off value of 0.809 x 10^-3 mm²/sec was generated having a sensitivity of 72.2% and specificity of 75%. The PPV generated was 86.6% and NPV 54.4%. Therefore, although the sensitivity and specificity of ADC values is relatively low, this test has a high positive predictive value for benign lesions having ADC value of > 0.809 x 10^-3 mm²/sec. A significantly low ADC value (0.656 x 10^-3 mm²/ sec) was seen in a benign lesion which was proven to be meningioma on histopathology, which can be explained by the high cellularity of meningiomas.

DCE-MRI assesses the nature of the enhancement of the lesion against time which helps in evaluating the microcirculation of the lesion [6,13]. Xian et al. performed DCE-MRI on orbital lesions and concluded that a washout type TIC was the most predictive for malignancy [3]. A Study was done by Yuan et al. that all the lesions with a persistent pattern of enhancement (type I) curve turned out to be benign and 10 out of 14 lesions were showing type III curve were malignant [13]. Malignant lesions were found to have statistically higher Out of all the parameters, highest sensitivity (93.8%) was provided by Slope when the cut off value used was 1.10 while highest specificity (87.5%). was provided by a cut off value of 35.14 secs for Tp

We found significant difference between type I and type III curves (p value ~0.002). All lesions demonstrating type I curve were benign (eight) and seven benign lesions also demonstrated type II curve. This can be attributed to the fact that benign lesions generally have an intact microvasculature which has limited permeability.

Tp was the most optimal with the value of 0.96 with the highest specificity of 87.5% (with a cut off value of 141.5 secs). The corresponding sensitivity generated was 94.4%, PPV was 94.7% and NPV 87.5%. This signifies a high diagnostic accuracy of this test to differentiate benign and malignant orbital masses. The maximum sensitivity (100%) was obtained from slope with a cut off value of 0.47, however the corresponding specificity was relatively low (78%). The corresponding PPV and NPV were 66.6% and 100%. Values may also differ from previous studies as the values are calculated by drawing ROIs, the method of which can be different for different studies.

MRS has been shown to be useful to distinguish pure tumors from spectroscopically normal tissues [14,15]. Roshdy et al. showed choline peak in all (100%) patients with malignancy, and also in 14% of benign lesions, both of which were found to be optic nerve glioma [16]. In our study kappa test of agreement was applied to correlate MRS with histopathological diagnosis and the K value generated for choline peak was 0.612 which denotes a good level of agreement between spectroscopy and histopathology. The sensitivity of this test was low (62.5%) however specificity was 94.4% and the test had high PPV and NPV which suggests that presence of a choline peak was a significant sign of malignancy. The major limitation of MRS in orbit is the presence of blood, bone and fat surrounding the lesion which may lead to artefacts which was mitigated by taking single voxel spectroscopy in smaller lesions and chemical shift imaging in larger lesions. For a diagnostic test to accurately diagnose malignant lesions, the specificity should be 100% which was only attained by MRS. (Tp had the next highest specificity of 87.5%). All the three modalities can be used complementary to each other in addition to the conventional sequences to increase the confidence of diagnosing malignant lesions and help to plan treatment.

One of the major limitations of our study is the small sample size which did not allow us to differentiate between the specific subtypes of malignant lesion. Another limitation is the use of ROIs in all the three sequences which are user dependent. However this limitation was mitigated in our study by taking multiple ROIs of fixed areas and taking mean values as the standard value. Further studies using a larger population group may be needed to establish the usefulness of these sequences in orbital masses and to differentiate different subtypes of masses.

Diffusion-weighted MRI using ADC (mean) can be used to differentiate between benign and malignant orbital masses. Out of the DCE parameters, only Time intensity curve, time to peak and Slope are useful in differentiating benign and malignant masses with Tp having high specificity and Slope having a high sensitivity. Choline peak in spectroscopy shows high agreement with malignancy on histopathology. Combining these three sequences increases the confidence of predicting malignancy in an orbital lesion