Models Based on Multiple-parameters Magnetic Resonance Imaging Radiomics to Predict Prostate Cancer of Clinically Significant
Objective This study explored the value of the models based on the preoperative multi-parameter magnetic resonance images in identifying clinically significant prostate cancer.Methods Our study was a retrospective study on 332 patients who underwent mp-MRI examination in Affiliated Hospital of Nantong University.After a collection of their complete preoperative clinical factors and pathological outcomes,these patients were randomly classified at a ratio of 7:3 into a training cohort and a validation cohort.We applied T-test or Chi-square test to assess whether the differences of clinical risk factors between the two cohorts were statistically significant.Applying PyRadiomics were for radiomics features extraction from T2WI,DWI and DCE sequences,we retained the radiomics features with the highest correlation with CsPCa after selection.Therefore,we got the radiomics score formula based on the summary of corresponding coefficients of the radiomics features.We evaluated the performance of the radiomics models via accuracy,sensitivity,specificity,Area Under Curves,Receiver Operating Characteristic Curves,Decision Curve Analyses and calibration curves.Results 1.We randomly divided all the patients into the training cohort(n=232)and the validation cohort(n=100)at a ratio of 7:3,and all the variables had no statistical difference between the two cohorts.2.After assessing the efficacy of the five radiomics models,we found that multimodal model B showed the best predictive efficacy,with AUC values reaching 0.90 and 0.86 respectively in the training and validation groups.The calibration curve illustrated good agreement between the predicted and pathological results,and the DCA curve also showed the greatest clinical benefit.The accuracy and sensitivity of model A reached the highest among the five models,0.86 and 0.86 in the training group and 0.81 and 0.78 in the validation group,respectively.The specificity of DWI model reached 0.91 and 0.80 in the two groups,which was the highest.Conclusion The mp-MRI radiomics models could be applied for predicting CsPCa.It demonstrated its excellent capacity for facilitating preoperative predication and clinical decisions.
NETL
NSTL
万方数据
维普
