MRI‐Based Radiomics Nomogram for Preoperative Differentiation Between Ocular Adnexal Lymphoma and Idiopathic Orbital Inflammation

Abstract

Background: Ocular adnexal lymphoma (OAL) and idiopathic orbital inflammation (IOI) are malignant and benign lesions for which radiotherapy and corticosteroids are indicated, but similar clinical manifestations make their differentiation difficult. Purpose: To develop and validate an MRI-based radiomics nomogram for individual diagnosis of OAL vs. IOI. Study Type: Retrospective. Population: A total of 103 patients (46.6% female) with mean age of 56.4 AE 16.3 years having OAL (n = 58) or IOI (n = 45) were divided into an independent training (n = 82) and a testing dataset (n = 21).Field Strength/Sequence: A 3-T, precontrast T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), and postcontrast T1WI (T1 + C). Assessment: Radiomics features were extracted and selected from segmented tumors and peritumoral regions in MRI before-and-after filtering. These features, alone or combined with clinical characteristics, were used to construct a radiomics or joint signature to differentiate OAL from IOI, respectively. A joint nomogram was built to show the impact of the radiomics signature and clinical characteristics on individual risk of developing OAL. Statistical Tests: Area under the curve (AUC) and accuracy (ACC) were used for performance evaluation. Mann-Whitney U and Chi-square tests were used to analyze continuous and categorical variables. Decision curve analysis, kappa statistics, DeLong and Hosmer-Lemeshow tests were also conducted. P < 0.05 was considered statistically significant. Results: The joint signature achieved an AUC of 0.833 (95% confidence interval [CI]: 0.806-0.870), slightly better than the radiomics signature with an AUC of 0.806 (95% CI: 0.767-0.838) (P = 0.778). The joint and radiomics signatures were comparable to experienced radiologists referencing to clinical characteristics (ACC = 0.810 vs. 0.796-0.806, P > 0.05) or not (AUC = 0.806 vs. 0.753-0.791, P > 0.05), respectively. The joint nomogram gained more net benefits than the radiomics nomogram, despite both showing good calibration and discriminatory efficiency (P > 0.05). Data Conclusion: The developed radiomics-based analysis might help to improve the diagnostic performance and reveal the association between radiomics features and individual risk of developing OAL.