<![CDATA[Introduction: Prostate cancer is a well-recognized medical problem accounting for the most diagnosed type of cancer in men. The importance of early detection and its improved survival rate have motivated research on the best cancer detection method. Consequently, computer-aided diagnosis was introduced; however, more datasets are needed, and more testing and trials are required to reach a feasible and reliable diagnostic method. In this study, we use MRI T2 WI and ADC-map sequences to build a classifier to differentiate between clinically significant and insignificant prostate lesions. Material and Methods: Haralick’s first and second order statistical features were extracted from pathologically proven prostate lesions found in The Cancer Imaging Archive open data source. We used the WEKA platform for data analysis, including 152 lesions divided into 70% training set and 30% testing set. Results: The proposed classifier showed sensitivity, specificity, F-measure, and AUROC of 82.6%, 87%, 84.4%, and 92.6%, respectively. Conclusions: The proposed classifier does not require a high-end computer, outperforms many previous classifiers, and has the potential to discriminate clinically significant from insignificant prostate lesions.]]>
Copyrights © 2023
