Discrimination of Benign and Malignant Suspicious Breast Tumors Based on Semi-Quantitative DCE-MRI Parameters Employing Support Vector Machine

Abstract

Purpose: Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) is an effective tool for detection and characterization of breast lesions. Qualitative assessment of suspicious breast DCE-MRI is problematic and operator dependent. The purpose of this study is to evaluate diagnostic efficacy of the representative characteristic parameters, extracted from kinetic curves of DCE-MRI, for discrimination between benign from malignant suspicious breast tumors.
Methods: Pre-operative DCE-MR images of twenty-six histopathological approved breast lesions were analyzed. The images were reviewed by an expert radiologist and the regions of interests (ROI)s were selected on the most solid part of the lesion.Semi-quantitative kinetic parameters, namely: maximum signal enhancement (SI), max 60 initial area under the curve (IAUC), time to peak (TTP), wash in rate (WIR), wash out rate (WOR) and signal enhancement ratio (SER), were calculated within each ROI. Mean values of the calculated features among benign and malignant groups were compared using student’s t-test. Finally, a classification was performed employing support vector machines (SVM) using each of the parameters and their combinations in order to investigate the efficacy of the parameters in distinguishing between benign from malignant tumors.
Results: The performance of the classification procedure employing the combination of semi-quantitative features with (p-value< 0.001) was evaluated by means of several measures, including accuracy, sensitivity, specificity, positive predictive value and negative predictive value which returned amounts of 97.5%, 96.49%,100%, 100% and 95.61% respectively.
Conclusion: In conclusion, semi-quantitative analysis of the characteristic kinetic curves of suspicious breast lesions derived from SVM classifier provides an effective lesion classification in breast DCE-MR images.