The purpose of this study was to investigate the value and diagnostic efficiency of DWI,T1WI and T2WI using texture analysis for discriminating the gleason scores of prostate cancer. The results of this study indicate that texture analysis may provide a new method for Gleason classification of prostate cancer. A radiomics model of textural features from T2WI and ADC maps have a good diagnostic accuracy in patients of a prostate cancer. Quantitative textural analysis may help distinguish low cancers form high- or intermediate-grade cancer with high sensitivity and moderate specificity.
Purpose: To investigate the value and diagnostic efficiency of DWI,T1WI and T2WI using texture analysis for discriminating the gleason scores of prostate cancer.
Materials and methods: This study was approved by the Ethics Committee of Shanxi People’s Hospital, Xi’an, China. This retrospective study of prospectively collected data included 61 patients of 93 lesions suspected of having prostate cancer. There were 18 lesions for low-risk and 75 cases for intermediate/high risk prostate cancer. All patients had pre-treatment MRI scans comprised of T2WI and DWI with ADC based on a 3T scanner (Achieva TX, Philips Healthcare, the Netherlands). The 3D volumes of interest (VOIs) of all lesions based on T2WI using Imaging J software with the relative signal difference >two standard deviations. Forty three texture features were extracted from the 3D whole nodules region for T1WI, T2WI image and ADC map respectively. The features were reordered by using fisher score. And finally a SVM classifier was adopted for the classification of prostate cancer from hyperplasia with hypointensity nodule by using the first leading features of T1WI,T2WI, ADC and the combination of T2WI and ADC maps.VOIs were drawn by two independent readers in peripheral zone lesions and normal-appearing peripheral zone (NPZ) tissue identified on clinical images. Histopathologic correlation was based on systematic transrectal biopsy or cognitively targeted biopsy results, used to assess T1, T2, and ADC in the differentiation of (a) cancer versus NPZ, and (b )high- or intermediate-grade tumors versus low-grade tumors. Discriminating ability was evaluated by using the area under the receiver operating characteristic curve (AUC).
Results:T1WI, T2WI, ADC together and ADC, T2 together produced the same best separation between these two groups (AUC = 0.94), by the sensitivity of 0.967and specificity of 0.846. The SVM classifier using the optimal feature achieved the best performance in prostate cancer grading, with AUC, accuracy, sensitivity, and specificity of 0.861, 0.901, 0.980, and 0.500, respectively.
Discussion:The results of this study indicate that texture analysis may provide a new method for Gleason classification of prostate cancer. There is a certain features overlap between low cancers and high- or intermediate-grade cancer in conventional MR analysis, and texture analysis can be used to further evaluate the biological characteristics of tumor sites from the distribution of parameters.
Conclusion: A radiomics model of textural features from T2WI and ADC maps have a good diagnostic accuracy in patients of a prostate cancer. Quantitative textural analysis may help distinguish low cancers form high- or intermediate-grade cancer with high sensitivity and moderate specificity. A radiomics model approach may increase diagnostic confidence of abdominal radiologists on MR.