An integrated approach based on morphology, texture, and backscattering-statistics for distinguishing between benign and malignant breast

The B-mode image has become a popular diagnostic tool in breast ultrasound for describing the properties of breast tumors in morphology and texture. In order to better characterize scatterers in breast tumors, the Nakagami image based on the statistical distribution of raw ultrasound data has also been successfully proposed. However, since morphological analysis, texture analysis, and Nakagami imaging actually supply information on different physical characteristics of breast tumors, a functional complementation by combining the above methods may provide more clues for classifying breast tumors. For this reason, this study investigated a novel multiple-parameter-based analysis method for evaluating breast tumors, in which the tumor morphology, the scatterer echogenicity, and the scatterer arrangements and concentrations are described using the B-mode, texture, and Nakagami images, respectively. To verify the validity of the concept, raw data were obtained from 100 clinical cases. All patients were examined by an experienced radiologist, and the results were confirmed by surgery and pathological examinations or biopsy. The contours of the tumor were described manually by a breast surgeon familiar with breast ultrasound interpretation. Three morphological parameters, three texture features, and the Nakagami parameter of benignancy and malignancy were extracted. Fuzzy c-means clustering was applied to identify a tumor as benign or malignant based on combining the parameters. The results indicated that there would be a trade-off between sensitivity and specificity when combining the same physical characteristics. However, the combination of the standard deviation of the shortest distance (morphology), the variance (texture), and the Nakagami parameter concurrently allows both the sensitivity and specificity to exceed 85%, making the performance to diagnose breast tumors better.