Rationale and objectives: This novel study aims to investigate texture parameters in distinguishing malignant and benign breast lesions classified as Breast Imaging Reporting and Data System 4 in dynamic contrast-enhanced magnetic resonance imaging (MRI).
Materials and methods: This retrospective study included 203 patients with 136 breast cancer and 67 benign lesions who underwent breast MRI between November 23, 2016, and August 27, 2018. Co-occurrence matrix-based texture features were extracted from each lesion on T1-weighted contrast-enhanced MRI using MatLab software. The association between texture parameters and breast lesions was analyzed, and the diagnostic model for breast cancer was created. Classification performance was evaluated by the area under the receiver operating characteristic curve, sensitivity, and specificity.
Results: Significant differences were seen between malignant and benign lesions for a number of textural features, including contrast, correlation, autocorrelation, dissimilarity, cluster shade, and cluster performance (P < 0.05). After the analysis of the multicollinearity, 5 texture features (contrast, correlation, dissimilarity, cluster shade, and cluster performance) were included for the next principal component analysis. The differentiation accuracy of breast cancer based on the diagnostic model was 0.948 (95% confidence interval, 0.908-0.974).
Conclusions: Texture features that measure randomness, heterogeneity, or homogeneity may reflect underlying growth patterns of breast lesions and show great difference in malignant and benign lesions. Therefore, texture analysis may be a valuable assisted tool for diagnostic analysis on breast.