MR-IMAGING OF SOFT-TISSUE MASSES - DIAGNOSTIC EFFICACY AND VALUE OF DISTINGUISHING BETWEEN BENIGN AND MALIGNANT LESIONS

OBJECTIVE. The purpose of this study was to evaluate the efficacy of M R imaging in predicting the pathologic diagnosis of soft-tissue masses , both neoplastic and nonneoplastic, and in distinguishing benign from malignant lesions. MATERIALS AND METHODS. The imaging features of 225 soft-tissue tumors (179 benign, 46 malignant) in 222 patients were an alyzed. Univariate analysis of multiple individual imaging features wa s done, along with stepwise logistic regression analysis of combinatio ns of imaging features, to determine how useful these are for predicti ng malignancy or benignity, A subjective (group consensus) analysis of each case was done prospectively, and each tumor was placed into one of three diagnostic categories: (1) benign, diagnostic of a specific e ntity; (2) nonspecific, most likely benign; or (3) nonspecific, most l ikely malignant. Results were compared with the final diagnosis establ ished by pathologic examination (n = 184) or imaging/clinical data (n = 41). RESULTS. By quantitative analysis, no single imaging feature or combination of features could reliably be used to distinguish benign from malignant lesions, For the subjective analysis, a correct and spe cific benign diagnosis could be made on the basis of MR imaging findin gs in 100 (44%) of the 225 tumors. For the entire cohort, the sensitiv ity was 78%, the specificity was 89%, the positive predictive value wa s 65%, and the negative predictive value was 94% for a malignant diagn osis, When the diagnostic benign tumors were excluded, the specificity and negative predictive value decreased to 76% and 86%, respectively, whereas the sensitivity and positive predictive value remained the sa me. CONCLUSION. Many benign soft-tissue masses can be correctly and co nfidently diagnosed with MR imaging. The prevalence of benign lesions among soft-tissue masses accounts for the relatively high specificity and negative predictive value that can be achieved with MR imaging for tissue characterization. However, the accuracy of MR imaging declines when these characteristic benign tumors are excluded from analysis, A significant percentage of malignant lesions may appear deceptively '' benign'' with the currently used criteria. For lesions whose imaging a ppearance is nonspecific, MR imaging is not reliable for distinguishin g benign from malignant tumors, and these lesions warrant biopsy in mo st cases.