Breast disease: clinical application of US elastography for diagnosis.

PURPOSE

To evaluate the diagnostic performance of real-time freehand elastography by using the extended combined autocorrelation method (CAM) to differentiate benign from malignant breast lesions, with pathologic diagnosis as the reference standard.

MATERIALS AND METHODS

This study was approved by the University of Tsukuba Human Subjects Institutional Review Board; all patients gave informed consent. Conventional ultrasonography (US) and real-time US elastography with CAM were performed in 111 women (mean age, 49.4 years; age range, 27-91 years) who had breast lesions (59 benign, 52 malignant). Elasticity images were assigned an elasticity score according to the degree and distribution of strain induced by light compression. The area under the curve and cutoff point, both of which were obtained by using a receiver operating characteristic curve analysis, were used to assess diagnostic performance. Mean scores were examined by using a Student t test. Sensitivity, specificity, and accuracy were compared by using the standard proportion difference test or the Delta-equivalent test.

RESULTS

For elasticity score, the mean +/- standard deviation was 4.2 +/- 0.9 for malignant lesions and 2.1 +/- 1.0 for benign lesions (P < .001). When a cutoff point of between 3 and 4 was used, elastography had 86.5% sensitivity, 89.8% specificity, and 88.3% accuracy. When a best cutoff point of between 4 and 5 was used, conventional US had 71.2% sensitivity, 96.6% specificity, and 84.7% accuracy. Elastography had higher sensitivity than conventional US (P < .05). By using equivalence bands for noninferiority or equivalence, it was shown that the specificity of elastography was not inferior to that of conventional US and that the accuracy of elastography was equivalent to that of conventional US.

CONCLUSION

For assessing breast lesions, US elastography with the proposed imaging classification, which was simple compared with that of the Breast Imaging Recording and Data System classification, had almost the same diagnostic performance as conventional US.