Differentiating Benign and Malignant Soft Tissue Masses by Magnetic Resonance Imaging: Role of Tissue Component Analysis

BackgroundThere is a variable degree of accuracy in discriminating benign from malignant soft tissue masses based on signal intensity and morphologic characteristics by magnetic resonance imaging (MRI). The aim of this study was to determine the utility of detailed component pattern assessment, in addition to morphologic study, for differentiating benign from malignant soft tissue masses by MRI.MethodsThe imaging features of 118 histologically proven soft tissue masses were analyzed according to: (1) signal characteristics: (a) high T1 matrix; (b) low T2 matrix; (c) fibrous tissue signal; (d) calcification; (e) myxoid signal tissue; (f) fatty signal tissue; (g) cystic signal; (h) necrotic signal; (i) septations; (j) vascular signal void signal; (k) fat rim; and (l) hemorrhage; and according to (2) morphologic assessment: (a) lesion size (maximal diameter) in centimeters (cm); (b) lesion depth in cm; (c) margins; (d) peritumoral edema; (e) bone involvement; (f) marginal capsule or pseudocapsule; and (g) neurovascular bundle involvement. Univariate and multivariate analyses followed by stepwise logistic regression of combination of imaging features were performed. The predictive value of each imaging feature and various combinations of imaging features were determined.ResultsIn univariate analysis, T2 low signal matrix, fibrous tissue, calcification, necrosis, septum, fat rim sign, peritumoral edema, and hemorrhage showed statistically significant differences between benign and malignant masses (p < 0.05). The positive predictive value of necrosis for malignancy was 84.8%, and its specificity was 90.9%. In multivariate analysis, the best model for predicting malignant masses was the combination of necrosis, maximal mass diameter, peritumoral edema, and absent fibrosis, absent calcification, and lack of fat rim. The combination of these parameters resulted in the most correct diagnoses of malignancy, with a sensitivity of 84.2%, specificity of 64.0%, and accuracy of 74.8%, whereas the accuracy of models consisting of component character and morphologic feature were 74.3% and 70.9%, respectively.ConclusionMRI is useful in determining whether a soft tissue mass is malignant or not. Traditional morphologic assessment was reinforced by detailed component characterization analysis. The parameters favoring malignancy were large lesion size, peritumoral edema, necrosis, and absent calcification, absent fibrosis, and lack of fat rim