Differentiating Between Malignant and Benign Breast Masses: Factors Limiting Sonoelastographic Strain Ratio

Abstract

We compared strain ratio vs. qualitative elastography for the further differentiation of focal breast lesions, with special focus on limiting factors. 215 patients with 224 histologically proven breast masses (116 malignant, 108 benign) were prospectively examined using a high-end ultrasound system (Philips iU22) with serial elastography function. B-mode scans and available mammograms were reviewed according to the BIRADS classification, raw elastogram data was analyzed qualitatively using the Tsukuba score and semiquantitatively by calculating the strain ratio (fat to lesion ratio). For diagnostic performance, Receiver Operating Characteristic (ROC) curve analysis was obtained. A sub-group analysis regarding breast density, lesion size, lesion depth and histological subtypes was performed. Mean strain ratio values were 3.04 ± 0.9 for malignant and 1.91 ± 0.75 for benign lesions (p < 0,001). The areas under the ROC curve values were 0.832 (95 % CI 0.777; 0.888) for strain ratio, 0.869 (95 % CI 0.822; 0.917) for Tsukuba score, 0.822 (95 % CI 0.768; 0.876) for B-mode ultrasound and 0.853 (95 % CI 0.799; 0.907) for mammography. Sensitivity, specificity, positive predictive value and negative predictive value of the strain ratio were 90.7 %, 58.2 %, 70.3 % and 85.1 %, when a cutoff point of 2.0 was used. Only lesion depth ≤ 4 mm was associated with diagnostic failure in the multivariate analysis of factors influencing accuracy, whereas no significant correlation between breast density and lesion size and the accuracy of the strain ratio could be found. The addition of strain ratio to B-mode ultrasound increases specificity without loss of sensitivity in differentiating between malignant and benign breast tumors. Strain ratio measurements should not be carried out on tumors with a lesion depth ≤ 4 mm.