ABSTRACT
The primary attraction of ultrasound mammography is its ability to image breast repeatedly with no deleterious effects. In addition, papable breast masses can be conclusively determined to be fluid-filled or solid. Benign and malignant lesions can be further diagnosed. The major limitations of breast echography are the dependency of the technique on the sonographer as well as the complexity of the qualitative visual criteria for differentiation. A quantitative approach may thus be appreciated for eliminating these factors, enhancing the reproducibility and objectivity of a given breast echogram. A real-time data acquisition and analysis system has been developed employing image texture analysis system. A diagnostic database is used in classification of images according to the specific statistical parameters for each pathology in the database. The accuracy of the results is estimated by relating the case to the nearest statistical clusters representing the different pathologies in the database. Correction of the statistical parameters used in the analysis against shape, area, depth and angle of the selected region from transducer is also considered. Standardized frequency and time gain control are used for all cases. The overall accuracies for benign cases is 84 percent by normal echography and 89 percent by additional tissue analysis methods, while that for different malignancies is 86 percent by normal echography and 90.3 percent by tissue analysis