Synopsis

Keywords: Radiomics, Diffusion/other diffusion imaging techniques

The study explored more promising radiomics extracted from apparent diffusion coefficient maps for diagnosing Sjogren’s syndrome (SS) without head＆neck MR morphology changes. A total of 119 consecutive SS participants and 95 healthy volunteers were prospectively analyzed by 3.0 T MR including diffusion weighted imaging. Forty-five radiomic parameters were selected and twenty-two radiomic parameters showed significant difference between SS and controls, in which 11 parameters had an area under the ROC curve (AUC) greater than 0.700. The SVM classification model differentiated SS from healthy controls with an AUC of 0.932 and 0.911 in the training and testing sets, respectively.

Introduction Sjogren’s Syndrome (SS) is a chronic systemic autoimmune disease that is characterized by direct injury of the exocrine glands1, 2. So, early and accurate diagnosis of SS is vital for treatment planning3,4. Early diagnosis of SS patients without the accrual of MRI morphological changes has become a hot research point in the investigation of many functional MRIs as early diagnosis of SS allows early intervention, avoiding treatment delay5-7. In the diagnosis of SS, diffusion-weighted imaging (DWI) has been proved to have high diagnostic value5, 8, 9. According to the previous studies, only the mean apparent diffusion coefficient (ADC) values were obtained from the region of interest (ROI), but were insufficient for assessing the changes in the intra-glandular microenvironment and the spatial heterogeneity of the gland. Radiomics analysis assists in evaluating the gray-level intensity and the position of pixels within an image, and directly measures the heterogeneity of ROI10. Radiomics analysis has been widely used in various types of tumors11-13, and is also used to evaluate inflammatory lesions, such as multiple sclerosis and Crohn’s disease through texture features that reflect the heterogeneity14, 15. To the best of our knowledge, there were no reports till date that applied radiomics analysis based on ADC maps for diagnosing SS. This study aimed to explore more promising radiomic features extracted from the ADC maps of patients without any obvious morphological changes to accomplish early diagnosis of SS. Methods A total of 119 consecutive patients with SS (with 238 parotid glands) and 95 healthy volunteers (with 190 parotid glands) were enrolled in our study. All participants were scanned on a 3.0 T scanner (Ingenia, Philips Medical Systems, Best, the Netherlands) including DWI.. The ROIs were drawn manually on DWI (b=1000 s/mm2) to cover the largest slice of each parotid gland and then were copied to ADC maps. Feature selection was performed using R software, version 3.4.4. 45 feature parameters were selected for further processing of the study. Radiomics features with statistical significance in univariate analysis (p< 0.05) entered into a multivariate logistic regression analysis. Backward stepwise selection based on the Akaike information criterion (AIC) was applied using “MASS” package (ver. 7.3-50). The AIC value and the Hosmer-Lemeshow test were used as the measure of goodness of fit. After feature selection, the support vector machine (SVM) model with a radial basis function kernel was performed by repeated 10-fold cross-validation with 100 trials. The differences in continuous variables were analyzed by Mann-Whitney U test, and the differences in categorical variables were analyzed by chi-square test. The diagnostic performance of 45 selected radiomic features or multivariate models was evaluated using the receiver operating characteristic (ROC) analysis and the area under the ROC curve (AUC). All statistical analyses were performed with SPSS 22.0 software (SPSS Inc., Chicago, IL). A two-tailed p value of less than 0.05 was considered to be statistically significant. Results The mean value of ADC in patients with SS was significantly higher than that of healthy controls (p=0.001), with an AUC of 0.607. Among the 45 selected radiomic feature parameters, the value of 15 feature parameters was significantly higher in SS patient group than those in healthy volunteers group. The value of 7 feature parameters was significantly lower in the SS patient group than those in the healthy group. There were 11 parameters with AUC values greater than 0.700. The SVM classification model differentiates patients with no morphological changes from healthy volunteers with an AUC of 0.932 in the training set and an AUC of 0.911 in the testing set. Discussion We found that the ADC mean values of patients with SS were significantly higher than that of healthy volunteers, which might be due to parotid lymphocyte infiltration, glandular edema and increased capillary permeability, which in turn causes expanded extracellular space and increased water molecular diffusion, but still had a relatively low diagnostic value (AUC=0.607). Radiomics analysis is useful for the detection of different patterns of signal intensities that are not easily quantifiable by the human eye. We speculated that texture analysis of ADC maps can be used to reflect the microenvironment changes and the heterogeneity inside the parotid gland in SS patients without morphological changes in MR. Among the 11 positive parameters, 8 parameters belonged to GLCM classification, which is widely applied in texture description, and the results from the co-occurrence of matrices are better than those of the other texture discrimination methods16,17. There are two feature parameters that belonged to the Shape classification, which can distinguish the SS patients and healthy volunteers well (AUCs were 0.709 and 0.878, respectively). We speculated that the microenvironment and heterogeneity of the parotid gland have been changed in patients with early-stage SS and who were negative by conventional MRI.Our study included larger sample size for SS imaging studies and obtained 45 independent and significant feature parameters through ICC and ACC. Therefore, a diagnostic model with wide clinical application value was constructed by SVM. Conclusion We suggest that SVM modeling based on ADC texture analysis can be used for diagnosing patients with SS who were negative by conventional MRI.

Acknowledgements

No acknowledgement found.

References

1. Vitali C, Monti P, Giuggioli C, et al., Parotid sialography and lip biopsy in the evaluation of oral component in Sjogren's syndrome, Clin Exp Rheumatol. 1989;7(2):131-135. 2. Vitali C, Bombardieri S, Jonsson R, et al., Classification criteria for Sjogren's syndrome: a revised version of the European criteria proposed by the American-European Consensus Group, Ann Rheum Dis. 2002;61(6):554-558. 3. Fox RI, Sjogren's syndrome, Lancet. 2005;366(9482):321-331. 4. Shiboski SC, Shiboski CH, Criswell L, et al., American College of Rheumatology classification criteria for Sjogren's syndrome: a data-driven, expert consensus approach in the Sjogren's International Collaborative Clinical Alliance cohort, Arthritis Care Res (Hoboken). 2012;64(4):475-487. 5. Regier M, Ries T, Arndt C, et al., Sjogren's syndrome of the parotid gland: value of diffusion-weighted echo-planar MRI for diagnosis at an early stage based on MR sialography grading in comparison with healthy volunteers, Rofo. 2009; 181(3):242-248. 6. Su GY, Xu XQ, Wang YY, et al., Feasibility study of using intravoxel incoherent motion mri to detect parotid gland abnormalities in early-stage Sjogren syndrome patients, J Magn Reson Imaging. 2016;43(6):1455-1461. 7. Chu C, Zhou N, Zhang H, et al., Use of T1rhoMR imaging in Sjogren's syndrome with normal appearing parotid glands: Initial findings, J Magn Reson Imaging. 2017;45(4):1005-1012. 8. Ding C, Xing X, Guo Q, et al., Diffusion-weighted MRI findings in Sjogren's syndrome: a preliminary study, Acta Radiol. 2016;57(6):691-700. 9. Xu X, Su G, Hu H, et al., Effects of regions of interest methods on apparent coefficient measurement of the parotid gland in early Sjogren's syndrome at 3T MRI, Acta Radiol. 2017;58(1):27-33. 10. Davnall F, Yip CS, Ljungqvist G, et al., Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice?, Insights Imaging. 2012;3(6):573-589. 11. Li H, Zhu Y, Burnside ES, et al., MR Imaging Radiomics Signatures for Predicting the Risk of Breast Cancer Recurrence as Given by Research Versions of MammaPrint, Oncotype DX, and PAM50 Gene Assays, Radiology. 2016;281 (2):382-391. 12. Huang Y, Liu Z, He L, et al., Radiomics Signature: A Potential Biomarker for the Prediction of Disease-Free Survival in Early-Stage (I or II) Non-Small Cell Lung Cancer, Radiology. 2016;281(3):947-957. 13. Ren J, Tian J, Yuan Y, et al., Magnetic resonance imaging based radiomics signature for the preoperative discrimination of stage I-II and III-IV head and neck squamous cell carcinoma. Eur J Radiol. 2018;106:1-6. 14. Zhang Y, Moore GR, Laule C, et al., Pathological correlates of magnetic resonance imaging texture heterogeneity in 5. 15. Makanyanga J, Ganeshan B, Rodriguez-Justo M, et al., MRI texture analysis (MRTA) of T2-weighted images in Crohn's disease may provide information on histological and MRI disease activity in patients undergoing ileal resection. Eur Radiol. 2017;27(2):589-597. 16. Haralick RM, Shanmugam K, Dinstein I, TEXTURAL FEATURES FOR IMAGE CLASSIFICATION, Ieee Transactions on Systems Man And Cybernetics. 1973; SMC3 (6):610-621. 17. Zhang X, Cui J, Wang W, Lin C, A Study for Texture Feature Extraction of High-Resolution Satellite Images Based on a Direction Measure and Gray Level Co-Occurrence Matrix Fusion Algorithm, Sensors (Basel). 2017;17(7).