Computer aided differentiation of benign and malignant breast tumors by Ultrasound Images

Objective: To evaluate and compare the diagnostic potential of computerized texture analysis methods for classifying correctly individual ultrasound (US) images of the breast cancers. Methods and materials: In general, 32 breast US images (20 benign and 12 malignant tumors) were analyzed by MaZda software. More than 91 non overlapping region of interest (ROI) consisting 56 benign and 35 malignant tumors were selected for automatic differential diagnosis. Gray level intensity rages of each selected ROI normalized by Three normalization schemes. More than 270 texture features parameters computed per ROIs per normalization schemes. Two feature reduction methods algorithmswere applied to find the most effective features to differentiate benign from malignant breast tumors. Obtained features parameters under two standardization states: standard (S) and nonstandard (NS) were used for texture analysis with Linear discriminant analysis (LDA) and Nonlinear discriminant analysis (NDA). Finally, Receiver Operating Characteristic (ROC) curve analysis was used via calculating sensitivity, specificity accuracy and Az value (area under the ROC curve) to examine the discrimination performance of applied texture analysis methods. Results: In comparison with LDA, NDA represent high sensitivity, specificity and accuracy (94%, 100% and 98% respectively) in differential diagnosis of breast cancer. Conclusion: Our results indicate that computerized texture analysis is a reliable method and has a potential to be used by radiologist for detection and discrimination breast cancer by ultrasound imaging

Keywords:

: Ultrasound , breast tumors , texture analysis , sensitivity , specificity