Speckle reduction imaging of breast ultrasound does not improve the diagnostic performance of morphology‐based CAD System

Abstract

Speckle reduction imaging (SRI) is a newly developed technique in ultrasound examination. This study aimed to compare the diagnostic performance of SRI and non-SRI breast ultrasound examinations by using a morphology-based computer-aided diagnostic system. One hundred ten patients with pathologically proven breast lesions were enrolled consecutively from April 2008 to October 2008. SRI and non-SRI ultrasound images were both obtained at the same examination for each patient. The regions of interest were manually sketched by an experienced physician without histological information. Nineteen practical morphologic features from the extracted contour were calculated and a support vector machine classifier identified the breast tumor as benign or malignant. Conventional binomial receiver operating characteristics curve analysis was used to represent the diagnostic performance of both SRI and non-SRI. Between SRI and non-SRI methods, there were no significant differences in the area under the receiver operating characteristics curve (Az value: 0.82 versus 0.81), the sensitivity (78.9% versus 84.2%), and the specificity (73.6% versus 70.8%). Based on the morphology study, the performance of breast ultrasound in characterizing the solid breast mass as benign or malignant was not significantly improved with SRI.