Quantitative ultrasound examination of peritumoral tissue improves classification of breast lesions

Abstract

Quantitative Ultrasound (QUS) methods showed high suitability for classifying malignant and benign tumors based on ultrasound data from suspicious breast lesions. Apart from differences in internal structure, malignant and benign tumors have been also shown to have different effects on neighboring tissues. In our previous work we investigated the usefulness of QUS methods based on ultrasound data from surroundings of breast tumors. The present study is an attempt to answer the question of the optimal area of the surroundings to be used. The study included 116 tumors whose malignancy was determined by histopathological examination of biopsy samples. The parameters used in tumor classification were the shape parameter of the Nakagami distribution and ten texture parameters. The Linear Discriminant Analysis and the Leave-One-Out cross-validation were used to classify tumors. Classification results were assessed based on the area under the ROC curve (AUC). The best multi-parametric classifier for intra-tumor data has reached AUC = 0.82. In case of the data from the tumor surrounding area the best classification result was AUC = 0.89 and it was obtained for the surroundings range of 5 mm.