Abstract
BACKGROUND: Breast cancer is one of the most common cancers affecting women worldwide. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is a highly sensitive technique for the diagnosis of breast cancer, providing comprehensive kinetic analysis of lesions. OBJECTIVE: This study aimed to develop an automated method for classifying kinetic curves derived from DCE-MRI into three types - persistent (Type I), plateau (Type II), and washout (Type III) - using a support vector machine (SVM) classifier. METHODS: DCE-MRI scans from 41 histopathologically confirmed breast lesions were analyzed. Region of interest were manually selected by an expert radiologist on the most enhancing solid areas. Kinetic features, including initial enhancement (E_Initial), early signal enhancement ratio, peak enhancement (E_Peak), and four gradient-based slope features, were extracted. A SVM classifier was trained on these features, and performance was evaluated using accuracy, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC) to investigate the efficacy in distinguishing between the kinetic curve for each suspicious breast lesion. RESULTS: The performance of the classification procedure employing the kinetic features with (P < 0.001) was evaluated by means of several measures, including accuracy, sensitivity, specificity, 97.56%, 96.49%, 100%, 100%, and 97.62%, respectively. The results achieved a higher area under the ROC curve (AUC) of 100%. CONCLUSION: To diagnose breast lesions, DCE-MRI scans offer an important information, such as kinetic analysis, which is a useful and irreplaceable component of breast diagnostics. This approach may reduce unnecessary biopsies and improve diagnostic efficiency.