An Application Study On Preoperative Predict The Aggressiveness And Bone Metastasis Of Prostate Cancer Based On Multi-parameter MRI Radiomics

PART Ⅰ A study on the Preoperative Evaluating the Aggressiveness of Prostate Cancer based on Multi-parameter MRI RadiomicsPurpose:To explore the significance and value of multi-parameter MRI(Mp-MRI)radiomics in preoperative evaluation of the aggressiveness of prostate cancer(PCa)patients,and establish a predictive model combining multi-parametric MRI radiomics with clinically independent risk factors.Methods:One-hundred and sixteen patients who underwent prostate Mp-MRI and confirmed by pathology with diagnosed PCa from January 2014 to January 2019 were enrolled,and they were divided into training set and validation set according to the random stratified sampling method.Sketching region of interest and extracting radiomic features were from diffusion-weighted imaging(DWI),axial T2-weighted imaging fat suppression(FS-T2WI),and dynamic contrast enhanced(DCE)magnetic resonance imaging of each patient.Dimension reduction,feature selection,and radiomics signature construction were performed using the minimum Redundancy Maximum Relevancemand the Least Absolute Shrinkage and Selection Operator regression.The established radiomics signatures included three single-sequence and three-sequence combined Mp-MRI signature,and evaluated their diagnostic efficacy respectively.Meanwhile,independent clinical predictors were constructed based on age,total prostate-specific antigen(TPSA)level and clinical stage.The radiomics nomogram was established combined the radiomics signature with the clinical predictors.The above models were validated by the validation set.Nomogram calibration and discrimination were evaluated in training set and verified in the validation set using calibration curve and relative operating characteristic curve(ROC)curve.Finally,the clinical usefulness of the nomogram was estimated through decision curve analysis.Results:A total of one clinical risk factor(TPSA)was screened for a significant correlation with Gleason grade.T2WI,DWI,DCE,and Mp-MRI combined with three sequences screened 3,8,5,and 7 radiomics features that correlated with Gleason grade,respectively.Among the 4 sets of radiomics signatures constructed,the area under the ROC curve of the Mp-MRI group was the largest,which was 0.79(95%CI,0.69～0.89)in the training set and 0.75(95%CI,0.58～0.91)in the validation set.Therefore,the constructed radiomics nomogram combined Mp-MRI radiomics score(rad-score)and a clinical risk factor(TPSA level).The area under ROC curve of the nomogram was 0.83(95%CI,0.74～0.92)in the training set,and 0.81(95%CI,0.65～0.96)in the validation set.The calibration curve showed that the model had good calibration performance in both the training set and the validation set.Meanwhile,decision curve analysis also demonstrated that the radiomics nomogram had good clinical application value.Conclusion:The radiomics method based on multi-parametric MRI can be used to evaluate the aggressiveness of prostate cancer patients before surgery,and the diagnostic effectiveness of radiomics signature based on multi-parameter MRI(including T2WI,DWI and DOE sequences)was higher than any single sequence alone.Meanwhile,the established radiomics nomogram based on multi-parameter MRI radiomics signature and clinically independent risk factor(TPSA)has further improved diagnostic efficiency,and can be considered for clinical practice.PART Ⅱ A Multi-parameter MRI Radiomics Nomogram for Predicting Bone Metastases in Newly Diagnosed Prostate Cancer PatientsPurpose:To establish and validate a nomogram modal based on combining multi-parameter MRI(Mp-MRI)radiomics with clinically independent risk factors to predict the risk of bone metastasis(BM)in patients with newly diagnosed prostate cancer(PCa).Methods:One-hundred and sixteen patients(bone metastasis group n=55;non-bone metastasis group n=61)who underwent prostate Mp-MR imaging and confirmed by pathology with newly diagnosed PCa from January 2014 to January 2019 were enrolled.Sketching region of interest and extracting radiomics features were from diffusion-weighted imaging(DWI),axial T2-weighted imaging fat suppression(FS-T2WI),and dynamic contrast enhanced(DCE)magnetic resonance imaging of each patient.The samples were randomly divided into the training set and the validation set at a ratio of 7:3.Dimension reduction,feature selection,and radiomics feature construction were performed using the Least Absolute Shrinkage and Selection Operator regression in the training set,and the radiomics signature was established.At the same time,a clinical predictive model was established based on clinical risk factors including age,total prostate-specific antigen(TPSA)level,clinical stage,lymph node stage,Gleason score.In order to facilitate clinical application,a radiomics nomogram was established which combined the radiomics signature and clinical independent risk factors.The data of the validation set was substituted into the model to validate the nomogram.Nomogram calibration and discrimination were evaluated in training set and verified in the validation set using calibration curve and relative operating characteristic curve(ROC)curve.Finally,the clinical usefulness of the nomogram was estimated through decision curve analysis.Results:Radiomics signature consisting of 12 selected features and a clinical risk factor(TPSA)were significantly correlated with bone status(P≤0.05).The radiomics nomogram combined a radiomics signature from Multiparametric MR images with independent clinic risk factors.The area under ROC curve of the model was 0.93(95%Cl,0.86～0.99)in the training set,and 0.92(95%CI,0.84～0.99)in the validation set.The calibration curve showed that the model had good calibration performance in both the training set and the validation set.Meanwhile,decision curve analysis also demonstrated that the radiomics nomogram had good clinical application value.Conclusion:The radiomics nomogram,which incorporates the multi-parametric MRI-based radiomics signature and clinically independent risk factor(TPSA),can be used to promote individualized prediction of BM in patients with newly diagnosed PCa,and has good diagnostic efficiency,and can be considered for clinical application.