Rethinking Orbital Imaging : Establishing Guidelines for Interpreting Orbital Imaging Studies and Evaluating Their Predictive Value in Patients with Orbital Tumors

PurposeTo establish guidelines for interpretation of orbital imaging by magnetic resonance imaging (MRI) and/or computed tomography (CT), and to apply these guidelines and examine their predictive value in 131 patients with biopsy-proven orbital tumors.DesignProspective evaluation of imaging studies.ParticipantsImaging studies (CT and/or MRI) from 131 cases with biopsy-proven orbital tumors.MethodsGuidelines for reviewing orbital imaging studies (MRI and/or CT) were established based on 5 major characteristics: (1) anatomic location, (2) bone and paranasal sinuses involvement, (3) content, (4) shape, and (5) associated features. In total, 84 features were established by an experienced orbital surgeon and a neuroradiologist. Applying these 84 features, imaging studies of 131 biopsy-proven orbital tumors were evaluated by 3 physicians.Main Outcome MeasuresImaging features: characteristics, sensitivity, specificity, and positive and negative predictive values in various groups of orbital tumors and κ values.ResultsOne hundred thirty-one cases of biopsy-proven orbital tumors were evaluated. Benign lesions were more likely to be smaller in size, round or oval in shape (29% of all benign tumors, 0% in malignant and inflammatory, P<0.001), and associated with hyperostosis (22% of all benign lesions, P<0.001). They were also more likely to be hyperdense or hypodense on CT imaging (15% and 11%, respectively; P<0.05 in comparison with inflammatory and malignant tumors). Inflammatory processes showed panorbital involvement (23% vs. 3%, and 0% in benign and malignant tumors, respectively; P<0.001). Orbital fat involvement and fat stranding were noticed only in inflammatory lesions (19% and 16%, respectively; P<0.001). None of the features occurred only in malignant tumors, but they tend to involve the anterior orbit more commonly (54% vs. 20%, and 29% in benign and malignant; P = 0.002), and were more likely to show bone erosion (31% vs. 6%, and 16% in benign and inflammatory tumors, respectively; P = 0.004) and molding around orbital structures (29% vs. 3% in benign, and 0% in inflammatory tumors, respectively; P<0.001). Features such as panorbital involvement, orbital fat, frontal sinus opacity, molding around orbital structures, perineural involvement, and fat stranding had specificity of 97% to 100%, but low sensitivity.ConclusionsGuidelines for analysis of orbital imaging studies (CT or MRI) are suggested. Based on these guidelines several imaging features showed significantly different occurrences in benign, malignant, and inflammatory processes; although this can help in differential diagnosis, tissue diagnosis may still be required.