Soft-tissue lipomas: accuracy of sonography in diagnosis with pathologic correlation.

To retrospectively determine the accuracy of sonography in helping to distinguish soft-tissue lipomas from other soft-tissue masses by using histologic proof as the reference standard. Institutional review board approval was obtained, and informed consent was waived. Thirty-nine patients who underwent sonographic evaluation of a soft-tissue mass followed by biopsy or resection were retrospectively evaluated. Two musculoskeletal radiologists (readers 1 and 2) reviewed the sonographic images, characterized the masses, and rated the level of confidence in the diagnosis of lipoma by using a five-point scale. A level of confidence was also rated for the prospective sonographic report, which was reviewed and designated as reader 3. Receiver operating characteristic (ROC) curves, including 95% confidence intervals, were generated, and the area under the ROC curve (A(z)) was calculated for each reader. Sensitivity, specificity, and accuracy for each reader were calculated by using a confidence rating of 4 or 5 as positive for lipoma. Weighted kappa analysis was also performed to assess for interobserver variability. Histologic examination yielded 25 lipomas and 14 nonlipomas. The echogenicity of lipomas ranged from hypoechoic to hyperechoic relative to muscle, although most were isoechoic or hyperechoic. A(z) values were 0.79 for reader 1, 0.56 for reader 2, and 0.77 for reader 3. There was no significant difference between the A(z) for each reader and for chance. Interobserver agreement was fair, with a kappa value of 0.35 among the three readers. Sensitivities were 52%, 40%, and 52%, and accuracies were 64%, 49%, and 64% for readers 1, 2, and 3, respectively. Sonography demonstrates low accuracy in the diagnosis of soft-tissue lipomas. (c) RSNA, 2004.