Computer-aided diagnosis of breast elastography and B-mode ultrasound

Abstract

Ultrasound (US) elastography, a new technique that images the elasticity of tissues, is now into the course of breast cancer diagnosis. The purpose of this study was to assess the diagnostic performance of a neural network using a combination of US elastography technique and US B-mode. A back-propagation neural network (BPN) is used to classify the breast masses as benign cyst, benign solid mass, or malignant solid mass using texture, strain, and morphological features computed from the segmented lesions. Sixty-two breast lesions in US elastography and US B-scan images that are biopsy proved are examined. A classification accuracy using a combination of US elastography and B-scan images is 87.09%, sensitivity 89.29%, specificity 85.29%, positive predictive value 83.33%, and the negative predictive value is 90.63%. With statistically significant features, the classification accuracy using a combination of US elastography and B-scan images is reported to be 82.25% with sensitivity 92.86%, specificity 73.53%, positive predictive value 74.29%, and negative predictive value 92.59%. The classification results indicate that US elastography in combination with US B-mode improves both sensitivity and specificity.