| Part one.Study on the value of the radiomics model based on T2W-MRI image texture in the diagnosis of clinically significant prostate cancerObjective:Based on T2W-MRI image texture,a radiomics model was built to explore the preoperative diagnostic efficacy of this model for clinically significant PCa.Materials and Methods:Patients with clinically symptoms of urinary tract or elevated PSA levels who underwent a prebiopsy prostate 3.0 T MRI examination(conventional MRI,DWI and DCE-MRI sequences)and then underwent TRUS guidance systematic prostate biopsy at the second affiliated hospital of soochow university were enrolled in this retrospective study between December 2014 and March 2017.All the included cases were treated with the same MRI equipment,and the scanning sequence parameters were consistent.The obtained T2WI images of each prostate case were imported into A.K software.Three pre-processing technologies were used to improve the recognition degree of image texture,and the pre-processed images were quantitatively analyzed.According to the final pathological diagnosis results,the included cases were divided into clinically significant PCa group and non-clinically significant PCa group(including non-clinically significant cancer and benign lesions).After each case was preprocessed,T2WI images were used to sketch the three-dimensional ROI of the lesion.The location and size of the lesion was determined by DWI,ADC images and DCE-MRI images as reference and manually sketched layer by layer along the contour of the lesion,and then multilayer fusion was used to obtain the three-dimensional ROI of the lesion.In this study,40 patients were randomly selected.The reproducibility of the observer segmentation was evaluated by evaluating the reproducibility of the feature extraction from the T2W-MRI.Through the ICC evaluation the feature extraction of observer A(twice),observer A(first)and observer B,the ICC is greater than 0.75 for good reproducibility.First,all cases were randomly divided into the training set and test set in a ratio of 6:4.Each lesion ROI voxel of training set based on T2W-MRI was extracted the computer derivative characteristics by methods of gray level co-occurrence matrix(GLCM),gray-level run-length matrix(RLM),gray histogram feature and form feature.Then using ANOVA,Mann-Whitney U test,Spearman correlation test and LASSO algorithm selected features to determine the best feature set.The most valuable features of clinically significant PCa were analyzed by logistic regression analysis,and a logistic regression classifier was trained and tested to classify clinically significant PCa group and non-clinically significant PCa group,so as to obtain a radiomics model for the diagnosis of clinically significant PCa.The diagnosis efficiency of the radiomics model was evaluated using ROC curve for identifying the clinically significant PCa group and non significant PCa.The diagnosis of clinically significant PCa about the sensitivity,specificity and accuracy were calculated at the maximum condition of the Jordan index.Results:All 381 cases of the group of prostate disease,199 cases were benign,and 182 cases were PCa(including 40 cases with GS<7 and 142 cases with GS>7).In response to the random selection of 40 cases,the ICC range of the twice images feature extraction of the observer A was 0.83?0.96.The ICC range of the observer A(the first)and the observer B was 0.76?0.93.The reproducibility between the inter-observer and intra-observer is good.A total of 396 radiomics features were obtained through the above multiple feature extraction methods,and finally the dimensions were reduced to 10 non-zero coefficient optimal diagnostic features through the multi-step feature selection methods.The optimal diagnostic feature analyzed by the logistic regression analysis obtained a radiomics model for diagnosing the clinically significant PCa,with an AUC of 0.957 in the training group and 0.908 in the test group.The sensitivity,specificity and accuracy of the training group were 0.872,0.933 and 0.908,respectively.The sensitivity,specificity and accuracy of the test group were 0.854,0.827 and 0.836,respectively.Conclusion:This study show that there was a good reproducibility between intra-observer and inter-observer about feature extraction based on T2W-MRI.the training and test a radiomics model was constructed by the diagnosis of clinically significance of PCa with high efficiency and can be used as a clinically preoperative diagnosis clinically significant auxiliary method for PCa.The trained and tested radiomics model has a high diagnostic effectiveness for clinically significant PCa and can be used as an auxiliary method in clinically application.Part two.Study on the value of the combined model based on radiomics and clinically factors in the diagnosis of clinically significant prostate cancerObjective:The first part of the study have obtained the radiomics model for the diagnosis of clinically significant PCa based on the T2W-MRI image of the prostate.The second part of the study will establish the clinically factor model and the combined model based on the clinically factors and radiomics model for the diagnosis of clinically significant PCa.To study on the value of the combined model based on radiomics and clinically factors in the diagnosis of clinically significant PCa,we will compare the differences of diagnostic efficacy and clinically net benefit between the combined model,radiomics model and the clinically factor model.Materials and Methods:Patients with clinically symptoms of urinary tract or elevated PSA levels who underwent a prebiopsy prostate 3.0 T MRI examination(conventional MRI,DWI and DCE-MRI sequences)and then underwent TRUS guidance systematic prostate biopsy at the second affiliated hospital of soochow university were enrolled in this retrospective study between December 2014 and March 2017.According to the final pathological diagnosis results,the included cases were divided into clinically significant PCa group and non-clinically significant PCa group(including non-clinically significant cancer and benign lesions).In this part of the study,the following clinically factors related to the diagnosis of PCa were used as the observation indicators,including:age,prostate volume(PV),serum PSA(including tPSA and fPSA),f/t PSA,PSAD.First,all cases were randomly divided into the training set and test set in a ratio of 6:4.The clinically factors which were statistically significant in predicting the clinically significance of PCa were obtained by univariate and multivariate logistic regression analysis,and then the clinically factor model was obtained by logistic regression analysis.The first part of this study obtained the radiomics model for the diagnosis of clinically significant PCa based on T2W-MRI images.The logistic regression analysis was used to combine the statistically significant clinically factors with the radiomics model,and the radiomics-clinically combined model was obtained.The above three models were all constructed in the training set and tested in the corresponding test set.The ROC curve was used to evaluate the diagnostic efficacy of the clinically factor,radiomics and radiomics-clinically combined models in the differential diagnosis of clinically significant PCa group and non-clinically significant PCa group.The maximum effectiveness of the three models in the diagnosis of clinically significant PCa was analyzed by AUC.The critical value was taken under the condition of the maximum Johden index,and the sensitivity,specificity and accuracy of each model to predict the clinically significant PCa were calculated.The nomogram of clinically factor model and radiomics-clinically combined model were constructed to assist physicians in making clinically decision using the above characteristics.In order to determine the clinically practicability of the the clinically factor,radiomics and radiomics-clinically combined models,the decision curves of the three models were plotted,and the clinically net benefit under different threshold probability was quantitatively analyzed in the test set.Results:This study collected clinically factors related to the diagnosis of PCa,including:age,PV,serum PSA(including tPSA and fPSA),f/t PSA,PSAD.Univariate logistic regression analysis showed that age PV,tPSA,fPSA and PSAD were significant predictors of clinically significant PCa.Multivariate logistic regression analysis showed that the age,tPSA and fPSA was statistically significant(p<0.05).These three clinically factors could be used as independent predictors factors to diagnose the clinically significance of PCa.In the test set,the AUC,accuracy,sensitivity and specificity of the clinically factor model were 0.842,0.790,0.646 and 0.856,respectively.The AUC,accuracy,sensitivity and specificity of the radiomics model were 0.908,0.836,0.854 and 0.827,respectively.The accuracy,sensitivity and specificity of the radiomics-clinically combined model were 0.932,0.875,0.885 and 0.885,respectively.In the test set,the AUC value of the radiomics-clinically combined model was slightly higher than that of the radiomics model,but the p value was 0.47,which was higher than that of 0.05,there was no statistical difference.The decision curves of the clinically factor,radiomics and radiomics-clinically combined models show that if the threshold probability of a patient is 10-70%,using either the radiomics or combined model to diagnose clinically significant PCa adds more benefit than using the clinically factor model,with the optimal performance being 10-30%.Conclusions:This study shows that radiomics features can increase the value of clinically risk factors associated with PCa in diagnosing clinically significant PCa.The nomogram of the radiomics-clinically combined model provide a quantitative and intuitive method for clinicians to predict clinically significant PCa.The radiomics-clinically combined model has better diagnostic efficacy and clinically net benefit in predicting clinically significant PCa.Therefore,the radiomics-clinically combined model is helpful for the preoperative diagnosis and evaluation of clinically significant PCa,and it is expected to become an effective and chnicallyly applicable new diagnostic method with the deepening of research. |
PDF Full Text Request