MR imaging of silicone breast implants: evaluation of prospective and retrospective interpretations and interobserver agreement.

MR imaging was used to evaluate the integrity of silicone breast implants in 54 women with 108 implants. MR images were interpreted by relatively inexperienced readers who tried to reproduce the experiences reported in the literature. The study examines the interobserver agreement using different diagnostic signs and the influence of experience on interpretation errors. Prospective and retrospective interpretations were compared with surgical findings at the time of explanation. Diagnostic indicators, including the linguine sign, the inverted tear drop sign, the C sign, water droplets mixed with silicone, and extracapsular globules of silicone, were evaluated for diagnostic efficacy and interobserver agreement. The prospective sensitivity and specificity were 87% and 78%, respectively. With the retrospective interpretations, the sensitivity and specificity increased to 93% and 92%, respectively. Most of the prospective false-positive interpretations were due to misinterpreting radial folds as signs of implant rupture. Six implants interpreted retrospectively as false positives had gross amounts of silicone around the implants at surgery but there were no obvious rents in the implant shells. There was fair to excellent interobserver agreement with the individual diagnostic signs except for extracapsular globules of silicone. All of the signs had specificities of greater than 90%. The sensitivities of the individual signs were less than the overall retrospective sensitivity. With experience, the sensitivity improved from 87% to 93% and the specificity improved from 78% to 92%. This study helps substantiate the use of diagnostic signs used by other authors to detect silicone loss from breast implants by MR imaging; however, questions remain as to the clinical role of MR imaging in evaluating implants for silicone loss.