An Application Study Of Radiomic Features Of Magnetic Resonance In The Diagnosis And Prediction Of Extracapsular Extension In Prostate Cancer

Purpose:To explore the application value of magnetic resonance(MRI)based imaging features in the diagnosis and prediction of extracapsular extension in prostate cancer.Materials and Methods:459 patients who underwent pelvic Multiparametric MRI(MP-MRI)examination and prostate biopsy between January 2015 and December 2018 in the first affiliated hospital of SooChow University were involved in our study,among which 186 patients were confirmed with benign prostatic hyperplasia(BPH)by pathology and 273 patients undergoing radical prostatectomy were confirmed with prostate cancer.Complete T2WI sequence and ADC sequence images were exported from the Picture Archiving and Communication Systems(PACS).Region of interest(ROI)was identified and manually delineated slice by slice on both T2WI and ADC sequences by Medical Imaging Interaction Toolkit(MITK)software.Radiomic features were extracted from the ROIs on both T2WI and ADC sequences for each patient.The Spearman correlation analysis and Max-Relevance and Min-Redundancy(MRMR)method were used to select the features.The least absolute shrinkage and selection operator(LASSO)was performed to build radiomics signatures.Meanwhile,clinical data of patients were collected,including age,tPSA,f/tPSA,PI-RADS v2 score,percentage of positive cores,biopsy Gleason score,and postoperative pathology report.The independent clinical risk factors,which were identified by multivariable logistic regression analysis,were combined with the corresponding radiomics signatures to build integrated models.In the establishment of a prostate cancer diagnosis model,all patients were divided into a training set and a testing set by stratified sampling method according to the ratio of 7:3.In the establishment of extracapsular extension of prostate cancer prediction model,all prostate cancer patients were divided into training set and testing set in the same way.The radiomics signatures and integrated models were built based on the training sets and were tested on independent testing set.The areas under the receiver operating characteristic(ROC)curves(AUCs)along with 95%CI and accuracy were calculated to assess classification performance,and the cutoff value was selected according to the Youden index to determine the corresponding sensitivity and specificity.The efficiency of models were evaluated by comparing the models including different factors.Results:The prostate cancer diagnosis models based on T2WI sequence and ADC sequence in MRI performed respectively with the AUC of 0.775 and 0.863,sensitivity of 0.654 and 0.827,specificity of both 0.782,and accuracy of 0.699 and 0.809.The corresponding integrated models increased AUC to 0.851 and 0.912,sensitivity to 0.840 and 0.877,specificity to 0.727 and 0.873,and accuracy to 0.794 and 0.868,which were better than the models established by clinical risk factors.The models of extracapsular extension prediction based on T2WI sequence and ADC sequence in MRI performed respectively with the AUC of 0.599 and 0.625,sensitivity of 0.636 and 0.697,specificity of 0.625 and 0.521,accuracy of 0.617 and 0.580,which were worse than the models established by clinical risk factors.The corresponding integrated models increased the AUC to 0.726 and 0.728,sensitivity to 0.849 and 0.727,specificity to 0.583 and 0.688,and accuracy to both 0.691.Conclusion:Radiomic features based on MRI had a good diagnostic efficiency for distinguishing benign and malignant prostate lesions.While radiomic features were not effective in predicting extracapsular extension in patients with prostate cancer,but it still helped to improve the sensitivity of prediction.The performance of radiomics signatures based on ADC sequences were more efficient than those based on T2WI sequences.Model efficiency improved when combined radiomic features and clinical characteristics.Radiomic features based on MRI have a certain application value in the diagnosis and prediction of extracapsular extension in prostate cancer,which may become an important auxiliary diagnostic tool in the future with its great potential.