ABSTRACT
Objective:To investigate the value of texture analysis based on T 2WI and apparent diffusion coefficient (ADC) maps in discriminating low grade from high grade prostate cancer (PCa). Methods:Retrospective analysis was performed on patients who were confirmed to be PCa by pathology after surgery and underwent MRI examination in the department of radiology,Tongji Hospital,Tongji Medical College, Huazhong University of Science and Technology before radical surgery, including routine T 1WI, T 2WI and diffusion weighted imaging (DWI) sequences. 3D data analysis module of the MaZda software was used to manually draw region of interest (ROIs) slice by slice on T 2WI and ADC images, and generate volume of interest (VOI) of the entire tumor. MaZda software was also used to extract texture features. The independent sample t test or Mann-Whitney U test were used to identify the texture features with statistically significant differences between low and high grade PCa groups. Lasso regression model was used to select the best combination of texture features for identifying low and high grade PCa, and then the model was built. The receiver operating characteristic (ROC) curve was used to evaluate the diagnostic performance of the model in both training cohort and test cohort. Results:The best texture feature combination selected by Lasso regression model were the S (1, 0, 0) correlation of T 2WI and the S (1, 0, 0) correlation, S (1, -1, 0) sum entropy and vertical-run length nonuniformity of ADC maps. The area under the ROC curve (AUC) of the model in training cohort was 0.823, and the sensitivity and specificity were 70.4% and 80.8%, respectively, which were better than the single texture feature. The AUC of the model in test cohort was 0.714, which was worse than training cohort. Conclusion:The texture analysis of T 2WI and ADC maps is valuable for the identification of low and high grade PCa.
ABSTRACT
Objective@#To investigate the value of texture analysis based on MR ADC map of prostate in differentiating between low-grade and high-grade prostate cancer (PCa).@*Methods@#PCa confirmed by pathology after radical prostatectomy were analyzed retrospectively, all patients underwent multiparametric MRI before radical prostatectomy, including T1WI,T2WI and DWI. On the ADC map, ROI was drawn manually to encompass the whole tumor by ITK-SNAP software. The python-based pyradiomics package was used to extract 105 texture features. The intraclass correlation coefficient was used to evaluate the repeatability of the texture features. The independent sample t test or Mann-Whitney U test was used to exclude features that had no significant difference between low grade and high grade PCa. Lasso regression model and 5 fold cross validation method were used to obtain texture feature combination of the highest performance and develop a classification modelfor discriminating low from high grade PCa. ROC curve was used to evaluate the diagnostic efficiency of the model.@*Result@#Ninety patients with PCa confirmed by pathology after radical prostatectomywere analyzed retrospectively,including 36 patients with low-level PCa (GS≤3+4) and 54 patients with high-level PCa (GS≥4+3). The area under curve of the model was 0.841, with sensitivity 69.6% and specificity 91.2%, which was significantly higher than single texture feature or traditional mean ADC value.@*Conclusion@#Texture analysis based on MRI-ADC map of prostate could be used to discriminate low grade PCa from high grade PCa.
ABSTRACT
Objective To investigate the value of texture analysis based on MR ADC map of prostate in differentiating between low?grade and high?grade prostate cancer (PCa). Methods PCa confirmed by pathology after radical prostatectomy were analyzed retrospectively, all patients underwent multiparametric MRI before radical prostatectomy, including T1WI,T2WI and DWI. On the ADC map, ROI was drawn manually to encompass the whole tumor by ITK?SNAP software. The python?based pyradiomics package was used to extract 105 texture features. The intraclass correlation coefficient was used to evaluate the repeatability of the texture features. The independent sample t test or Mann?Whitney U test was used to exclude features that had no significant difference between low grade and high grade PCa. Lasso regression model and 5 fold cross validation method were used to obtain texture feature combination of the highest performance and develop a classification modelfor discriminating low from high grade PCa. ROC curve was used to evaluate the diagnostic efficiency of the model. Result Ninety patients with PCa confirmed by pathology after radical prostatectomywere analyzed retrospectively,including 36 patients with low?level PCa (GS≤3+4) and 54 patients with high?level PCa (GS≥4+3). The area under curve of the model was 0.841, with sensitivity 69.6% and specificity 91.2%, which was significantly higher than single texture feature or traditional mean ADC value. Conclusion Texture analysis based on MRI?ADC map of prostate could be used to discriminate low grade PCa from high grade PCa.