The Value Of MRI-based Radiomics In The Differential Diagnosis Of Prostate Cancer And Prostatic Hyperplasia Of Central Gland

Part Ⅰ The Value of Histogram Texture Analysis in the Differential Diagnosis of Prostate Cancer and Prostatic Hyperplasia of Central GlandObjective:To explore the value of histogram texture analysis(TA)based on T2WI-MRI and DWI-MRI images in the differential diagnosis of central gland(CG)prostate cancer(PCa)and benign prostatic hyperplasia(BPH).Methods:The clinical data of 21 patients with CG-PCa and 30 patients with CG-BPH confirmed by pathological biopsy under TRUS-guided biopsy from January 2015 to June 2019 in our hospital were retrospectively analyzed.From the AW4.6 workstation,the T2WI-MRI and DWI-MRI cross-sectional original images of all patients were collected and sequentially imported into ITK-SNAP software.Two radiologists participated in the reading together,combined with T2WI-MRI,DWI-MRI images and pathological results to confirm the image and range of the largest level of the lesion.One doctor manually sketched and saved the ROI,and the other doctor supervised and confirmed that if there were different opinions,consensus was reached through negotiation.GE’s AI-Kit software was used to standardize the original image of the lesion and the ROI-marked image,and matched one by one to extract the histogram feature parameters in the ROI.Statistical analysis,the ROC of statistically significant texture parameters,and then evaluate the efficacy of various features in differential diagnosis of benign and malignant lesions in prostate CG.Results:In T2WI-MRI images,the variance,energy,entropy,kurtosis,and skewness of the BPH group were significantly higher than those of the PCa group(P<0.05);in DWI-MRI images,the variance,mean,and skewness of the BPH group were significantly lower PCa group(P<0.05).On ROC,the differential diagnosis efficiency of variance in T2WI-MRI images is the highest,AUC is 0.77(95%CI,0.63-0.87;P<0.001).When the critical value is 2807.17,its sensitivity and specificity are 80.95%and 66.67%;DWI-MRI images have the highest diagnostic efficiency,with an AUC of 0.92(95%CI,0.81-0.98;P<0.001).When the critical value is 349.07,the sensitivity and specificity are 85.70%and 90.00%,respectively.In addition,the diagnostic power of the variance in the DWI-MRI images(AUC=0.87)is significantly higher than the diagnostic power of all histogram parameters under the T2WI sequence.Conclusion:The histogram analysis based on T2WI and DWI images is conducive to the differential diagnosis of PCa and BPH in the CG area.Although the diagnostic efficiency of DWI is high,the clinical diagnostic value of T2WI cannot be ignored.Part Ⅱ The Value of Radiomic Model in the Prediction of Prostate Cancer and Benign Prostatic HyperplasiaObjective:To explore the value of radiomic model based on different MRI sequences in the prediction of prostate cancer(PCa)and benign prostatic hyperplasia(BPH)of central glandMethods:The clinical data of 21 patients with CG-PCa and 30 patients with CG-BPH who were confirmed by pathological biopsy under TRUS-guided puncture biopsy from January 2015 to June 2019 in our hospital were retrospectively analyzed.From the AW4.6 workstation,the original T2WI-MRI and DWI-MRI cross-sectional images of all patients were collected and sequentially imported into the ITK-SNAP software.Two radiologists participated in the reading.Combining T2WI-MRI,DWI-MRI images and pathological results,confirm the image and range of the largest level of the lesion.One doctor manually sketched and saved the ROI,and the other doctor supervised and confirmed that if there were different opinions,consensus was reached through negotiation.Using GE’s AI-Kit software,the original image of the lesion and the image with ROI markers were standardized and preprocessed,and matched one by one.Collect information on the characteristics of lesions in each sequence.Dimensionality reduction through ANOVA+MW,correlation analysis and LASSO regression.According to the ratio of 7:3,the data is divided into training group and verification group.In the T2WI sequence,36 cases were used as the training group and 15 cases as the verification group;in the DWI sequence,36 cases were used as the training group and 15 cases were used as the verification group.The random forest method is used to construct the prediction model,and the effectiveness of the prediction model of each sequence is evaluated by drawing ROC.Results:Eight best features were selected on T2WI and DWI sequences.The AUC of the prediction model of the T2WI sequence in the training group was 0.87,and the sensitivity,specificity,accuracy,positive predictive value and negative predictive value were 90%,80%,86%,86%,and 86%,respectively;The AUC was 0.87,and the sensitivity,specificity,accuracy,positive predictive value and negative predictive value were 67%,100%,80%,100%and 67%,respectively.The prediction model of the DWI sequence has an AUC of 0.99 in the training group,and the sensitivity,specificity,accuracy,positive predictive value and negative predictive value are 95%,100%,97%,100%and 93%;It was 0.98,and the sensitivity,specificity,accuracy,positive predictive value and negative predictive value were 89%,100%,93%,100%and 86%,respectively.Conclusion:The radiomicmodel based on T2WI and DWI has certain value for the prediction of PCa and BPH nodules in the CG area,and the prediction effect based on the DWI sequence is better.