Quantitative breast mass classification based on the integration of B-mode features and strain features in elastography. (1st September 2015)

• Record Type: Journal Article
• Authors: Lo, Chung-Ming; Chang, Yeun-Chung; Yang, Ya-Wen; Huang, Chiun-Sheng; Chang, Ruey-Feng
• Is Part Of: Computers in biology and medicine. Volume 64(2015)
• Page Start: 91
• Page End: 100
• Publication Date: 2015-09-01
• Subjects: Breast cancer -- Elastography -- B-mode -- Computer-aided diagnosis -- Fuzzy c-means clustering; Medicine -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; 610.285
• DOI: 10.1016/j.compbiomed.2015.06.013
• Languages: English
• ISSNs: 0010-4825

Abstract

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