Quantitative breast mass classification based on the integration of B-mode features and strain features in elastography

摘要: BackgroundElastography is a new sonographic imaging technique to acquire the strain information of tissues and transform into images. Radiologists have observe gray-scale distribution on elastographic image interpreted as reciprocal Young's modulus evaluate pathological changes such scirrhous carcinoma. In this study, computer-aided diagnosis (CAD) system was developed extract quantitative features from images reduce operator-dependence provide an automatic procedure for breast mass classification. MethodThe collected database composed 45 malignant benign masses. For each case, tumor segmentation performed B-mode obtain contour which then mapped define corresponding area. The pixels around area were classified white, gray, black by fuzzy c-means clustering highlight stiff with darker values. Quantitative extracted cluster compared in classification ResultsThe performance proposed achieved accuracy 80% (72/90), sensitivity (36/45), specificity normalized under receiver operating characteristic curve, Az=0.84. Combining obtained significantly better Az=0.93, p-value<0.05. ConclusionsSummarily, quantified can be combined promising suggestion distinguishing tumors. wereextracted fromelastographic express tissue elasticity.A based classify masses.Combining malignancy evaluation.