Semi-Quantitative Analysis of DCE-MRI for Classification of the Prostate with and without Cancer

In our study, we aim to evaluate the diagnostic performance of semi-quantitative parameters on prostate tumors. Our dataset includes 47 Asian cases with 26 cases showed no signs of prostate cancer and 21 diagnosed with prostate cancer. Each individual has a ROI (region of interest) labeled by the radiologists. We established a model and extract five semi-quantitative parameters from the image information on the ROI regions. After that, linear discriminant analysis (LDA) and support vector machine (SVM) were used for classification, and area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of the classifier. When three semiquantitative parameters were studied, the AUC of LDA classifier was 0.815, and that of SVM was 0.840. When put 5 parameters to the 2 classifiers, after LDA training, the AUC is 0.837 (accuracy of 80.9%, sensitivity 88.5%, specificity 76.2
^%
). SVM yields an AUC of 0.896 (accuracy 85.1%, sensitivity 95.2%, specificity 75.1%). The results show that the trained classifier can accurately distinguish between patients with prostate cancer and those with no evidence of prostate cancer, and the results are at the same level of accuracy as the classification between benign and malignant areas in diagnosed patients.