| Part 1Objective:To explore a new type of nomogram based on magnetic resonance imaging(MRI)parameters combined with clinical pathological characteristics to predict the risk of positive surgical margin(PSM)after radical prostatectomy(RP),so as to guide clinical treatment decisions.Materials and methods:A total of 1055 male inpatients who underwent pelvic 3.0T magnetic resonance imaging(MRI)scans for suspected prostate cancer(PCa)from January 2016 to November 2021 were retrospectively collected.Inclusion criteria:(1)PCa was diagnosed as prostate cancer by tissue biopsy.(2)It has complete clinicopathological data,including age,prostate imaging report and data system v2.1(PI-RADS v2.1)score,total prostate specific antigen(TPSA),prostate volume,lesion diameter,preoperative biopsy Gleason score(GS)score,Percentage of positive cores(PPC),clinical tumor(c T)stage,postoperative GS grade,pathological tumor stage(p T)and extracapsular extension(ECE).(3)Complete MRI examination before biopsy.(4)Within 3 months after receiving MRI scan and prostate tissue puncture biopsy,the patients were completed laparoscopic RP by experienced urologists(more than 200 cases of laparoscopic RP surgery independently).Exclusion criteria:(1)Previous history of PCa treatment.(2)MRI scanning sequence is incomplete.(3)Prostate lesions with PI-RADS v2.1 score≤3.(4)The pathological tissue of the radical treatment showed that the volume of the lesion was small(diameter<1 cm),and the subsequent immunohistochemical staining of biomarkers could not be performed on the section.Finally,83 patients with PCa were included in this study.According to the status of surgical margins after surgery,the patients were divided into 37 patients in the negative surgical margin(NSM)group and 46 patients in the PSM group.MRI examination was performed with 3.0T magnetic resonance scanner(Signa,HDXT,GE Healthcare,United States)and abdominal 8-channel phased array coil.The scanning sequence includes: T1-weighted imaging(T1WI),T2-weighted imaging(T2WI),diffusion-weighted imaging(DWI),incoherent motion model(IVIM)and diffusion kurtosis imaging(DKI).IVIM uses single-shot echo plane(SS-EPI)imaging with b values of 0,20,50,100,150,200,400,800,1200,2000 and 3000 s/mm2.DKI uses SS-EPI imaging with 2 b values(0 and 1500 s/mm2)and 15 diffusion gradient directions.Then the image is reconstructed using GE Functool post-processing software to obtain 18 quantitative parameters,including DWI sequence(ADCME),IVIM sequence(ADCBE,Dmono,D * mono,fmono,Dbi,D * bi,fbi,DDC and α),DKI sequence(FA,MD,Da,Dr,FAk,MK,Ka and Kr).The region of interest(ROI)delineation is carried out at the level with the largest diameter of the lesion,along the edge of the lesion,and covers the entire lesion.R 4.0.3 is used for all statistical analysis(https://www.R-project.org).Shapiro-Wilk was used to test the normality of the test data.For the difference analysis between the two groups,independent sample t test was used for continuous variables of normal distribution,Mann-Whitney U test was used for continuous variables of non-normal distribution,and classification variables were used χ2 test or Fisher’s exact test.The intragroup correlation coefficient(ICC)function of Irr software package was used to analyze the inter-observer consistency.The stratified sampling method is used to divide the data into training set and verification set according to the ratio of 7:3.The least absolute shrinkage and selection operator(LASSO)algorithm is used to calculate and analyze the data,and the optimal prediction factor is selected from the quantitative parameters.Then,multivariate logistic regression was used to build a model to predict the PSM after RP surgery:(1)Model M based on quantitative parameters of DWIderived sequence.(2)The quantitative parameters of DWI-derived sequence combined with the preoperative clinicopathologic characteristic model MC with statistical differences.The effectiveness of the model was evaluated by the following methods:(1)The receiver operating characteristic curve(ROC)was used to evaluate the predictive ability of the model,and the area under the curve(AUC)was compared by Delong test.(2)Use calibration curve to evaluate the accuracy of the model.(3)The decision curve analysis(DCA)was used to evaluate the clinical net benefit of the model assisted decision-making.When the bilateral P value is less than 0.05,the difference is statistically significant.Results:A total of 83 patients with PCa were enrolled in this study,age(71±6.4)years.The incidence rate of PSM was 55.4%.The clinicopathological features with statistical differences between NSM and PSM included: PPC(0.417 in NSM group vs 0.625 in PSM group,P=0.012).Postoperative clinicopathologic features p T(p T2 in NSM group was 91.89%,p T3 a was 2.70%,p T3 b was 5.41% vs p T2 in PSM group was 60.87%,p T3 a was 4.35%,p T3 b was 34.78%,P=0.004)and pathological prostate ECE(91.89%,8.11% vs 60.87%,39.13%,P=0.001).The data were divided into training set(n=58)and validation set(n=25)according to the 7:3 ratio.There was no statistically significant difference in the clinicopathological baseline characteristics between the training set and the validation set(P>0.05).The ICC value of quantitative data measured between all observers was greater than 0.8,with good consistency.As the dependent variable,the risk coefficient of quantitative parameters of each DWI derivative sequence of PSM is estimated by using LASSO regression analysis.When λ= 0.074,the model has the best performance.At this time,three potential predictive factors of PSM,Da,Kr and Dbi,are screened.(1)By introducing Da,Kr and Dbi as independent predictors,a PSM prediction model M based on IVIM and DKI sequences is developed.(2)Da,Kr,Dbi and statistically different preoperative clinicopathological characteristics PPC were introduced as independent predictors to develop the PSM prediction model MC.The AUC of Da was 0.668(sensitivity 60.9%,specificity 73%,P=0.006).The AUC of Kr was 0.627(sensitivity 47.8%,specificity 81.1%,P=0.042).The AUC of Dbi was 0.689(sensitivity 67.4%,specificity 67.6%,P=0.001).The AUC of Da and Kr combined with Dbi was 0.730(sensitivity 76.1%,specificity 67.6%,P<0.001).However,there was no significant difference in AUC after comparing Da,Kr,Dbi and their combination(Da vs combined P=0.211,Kr vs combined P=0.140,Dbi vs combined P=0.242).The AUC of the M model training set was 0.716(sensitivity 69.2%,specificity 71.9%),and the AUC of the MC model training set was 0.737(sensitivity 65.4%,specificity 78.1%).The AUC of M model validation set was 0.805(sensitivity 90.9%,specificity 78.6%),and the AUC of MC model validation set was 0.831(sensitivity 90.9%,specificity 78.6%).There was no statistically significant difference in AUC between the two models(M model training set vs MC model training set P=0.391,M model validation set vs MC model validation set P=0.613).Comparing the calibration curves of the two models,the accuracy of the calibration curves of the training set and the verification set of the MC model is better than that of the M model.Comparing the DCA of the two models,in the current study,using MC model to predict the clinical net benefit of PSM risk in the range of 0.2-0.8 is higher.Conclusion:The nomogram based on multimodal MRI can predict the risk of PSM after RP surgery,and the combination of clinical and pathological features further improves the ability of the model to predict PSM before surgery.Part 2Objective:The differentially expressed genes(DEGs)between benign prostate tissue and PCa were analyzed by bioinformatics,and the risk factors related to the prognosis of PCa were screened in DEGs,and then the molecular biomarkers related to the risk factors before PSM surgery were further screened in combination with the clinical pathological information of patients.Materials and methods:By comparing and analyzing the gene expression profiles of benign prostate tissue and PCa tissue in the Genotype-Tissue Expression(GTEx)database and the Cancer Genome Atlas(TCGA)database,the PCa expression DEGs were obtained,and the intersection of the PCa expression DEGs obtained in the Gene Expression Omnibus(GEO)database was obtained.Then TCGA database was used to further determine the prognostic risk factors closely related to PCa overall survival(OS),disease-free survival(DFS)and progression-free survival(PFS).The risk factors were analyzed by gene ontology(GO)database function annotation and Kyoto Encyclopedia of Genes and Genomes(KEGG)database pathway enrichment.Finally,the correlation between the risk factors obtained by screening and the clinicopathological risk factors(age,GS grade of biopsy and c T stage)of PSM in TCGA database was analyzed.R 4.0.3 is used for the analysis of TCGA,GTEx and GEO databases(https://www.R-project.org)Conduct data conversion and statistical analysis.Firstly,the gene probe information matrix is extracted with R language,and the probe information is annotated with annotation file.Then all data are standardized and converted to log2 format.The limma package was used to analyze DEGs,set P < 0.05,log2FC>0 as DEGs,output the expression matrix of all differentially expressed genes,and use the ggplot package to draw gene expression hotspots and volcanic maps.Compare and analyze the two groups of DEGs obtained from TCGA and GEO databases,and use the Venn Diagram program package to draw the Wayne diagram for intersection.The survival analysis of the single-factor COX regression risk model was carried out using the surminer package,and the Wayne map was drawn to take the intersection of OS,DFS and PFS related risk factors,and the forest map was drawn.Subsequently,GO function annotation analysis and KEGG pathway enrichment analysis were performed on the selected risk factors,and a histogram was drawn.Use ggcor package to carry out Spearman correlation analysis and draw correlation matrix diagram.When the bilateral P value is less than 0.05,the difference is statistically significant.Results:The analysis of m RNA sequencing data of 493 PCa tissues in TCGA database and 100 benign prostate tissues in GTEx database showed that there were 16334 DEGs in total,of which 8067 were up-regulated and 8267 were down-regulated in PCa tissues.Through the analysis of m RNA sequencing data of 36 PCa tissues and 14 benign prostate tissues in GSE46602 data set in GEO database,a total of 5347 DEGs were found,of which 2907 were up-regulated and 2440 were down-regulated in PCa tissues.Compare and analyze the two database DEGs and take the intersection.After taking the intersection,there are 4862 remaining DEGs.In the TCGA database,with PCa patient OS,DFS and PFS as the end points,the single factor COX risk regression analysis was performed on 4862 DEGs and the Wayne diagram was drawn to obtain the intersection.There are 15 protection factors for OS,221 for DFI,and 272 for PFS.After taking the intersection,there are 0 remaining protection factors.There are 68 risk factors of OS,490 risk factors of DFI and 830 risk factors of PFS.After taking the intersection,there are 22 remaining risk factors.Finally,22 risk factors related to the prognosis of PCa were obtained.Combining 22 risk factors,GO biological function annotation and KEGG pathway enrichment results showed that MTA1 was involved in the regulation of molecular function-gene expression(P=0.03).TAF11 is involved in the biological process-thyroid hormone receptor binding(P=0.04),and TAF11 is enriched in the basal transcription factor(P=0.03).Further analysis showed that MTA1 was positively correlated with age(r=0.10,P=0.03),and MTA1 and TAF11 were positively correlated with biopsy GS grading(r=0.26,P<0.001,r=0.22,P<0.001).Conclusion:The key differential genes MTA1 and TAF11 obtained by using TCGA database and GEO database analysis are not only closely related to the prognosis of PCa patients,but also related to the clinicopathological risk factors before PSM surgery,which provides a direction for us to explore molecular biomarkers to predict PSM before surgery.Part 3Objective:To evaluate the predictive ability of specific molecular biomarkers MTA1 and TAF11 on PSM,and combine the quantitative parameters of DWI-derived sequence and clinicopathological features to build a PSM predictive model to provide preoperative personalized treatment guidance for PCa patients who are ready to receive RP.Materials and methods:In this part,83 patients with PCa were studied,and 30 patients with benign prostatic hyperplasia confirmed by biopsy pathology were selected as the benign control group with immunohistochemical staining.The MRI scanning scheme,image reconstruction and data measurement are the same as the first part.All pathological tissues(83 specimens of radical prostatectomy and 30 specimens of benign prostatic hyperplasia in prostate biopsy)were fixed with 10% neutral formalin and embedded in paraffin to make tissue sections.Two pathologists reviewed the diagnosis according to the WHO prostate histological classification and the relevant diagnostic criteria in the ISUP expert consensus.Olympus multifunctional biological microscope was used to take the image of immunohistochemistry reaction products.Under medium magnification(10 × 20),3 representative areas were selected for each slice and taken as digital photos.Use Image J1.52 V software to analyze the photos taken and measure the positive area ratio of protein immunoproducts.Immunohistochemical staining shows that the brownish yellow particles in the nucleus are positive.R 4.0.3 is used for all statistical analysis(https://www.R-project.org).ShapiroWilk test was used to verify the normality of the data.For the difference analysis between the two groups were tested by two-sided t test and Mann-Whitney U test.The correlation analysis adopts Spearman correlation test.The stratified sampling method is used to divide the data into training set and verification set according to the ratio of 7:3.Multivariate logistic regression was used to build a model to predict the postoperative PSM of RP before operation:(1)MC model based on quantitative parameters of DWI-derived sequence combined with statistically different preoperative clinicopathological characteristics.(2)Model BMC based on molecular biomarkers,quantitative parameters of DWI-derived sequences and statistically different preoperative clinicopathological characteristics.The effectiveness of the model is evaluated by the following methods:(1)ROC is used to evaluate the prediction ability of the model,and Delong test is used to compare AUC.(2)Use calibration curve to evaluate the accuracy of the model.(3)DCA evaluation model is used to assist the clinical net income of decision-making.When the bilateral P value is less than 0.05,the difference is statistically significant.Results:The clinicopathological characteristics of patients and the clinicopathological baseline characteristics between the training set and the validation set are the same as those in the first part.The positive area ratio of MTA1 in benign prostate tissue(0.012 [0.009,0.017])and PCa tissue(0.011 [0.01,0.013])had no statistically significant difference P=0.716.The positive area ratio of TAF11 was significantly different between benign prostate tissue(0.010 [0.007,0.012])and PCa tissue(0.040 [0.034,0.045])(P<0.001).The positive area ratio of MTA1 in NSM lesions(0.012 [0.011,0.015])and PSM lesions(0.010 [0.009,0.011])was significantly different(P=0.003).The positive area ratio of TAF11 in NSM lesions(0.037 ± 0.009)and PSM lesions(0.041 ± 0.006)was significantly different(P=0.023).MTA1 was negatively correlated with TPSA(r=-0.270,P=0.014),while TAF11 was not correlated with clinicopathological features.The expression and distribution of MTA1 in PCa tissue were statistically different in GSⅡ grading before and after operation P=0.05,while the expression and distribution of MTA1 and TAF11 in other grading PCa tissue were not statistically different P>0.05.MTA1 is negatively correlated with FAk(r=-0.223,P=0.043).TAF11 was negatively correlated with ADCME(r=-0.255,P=0.02),positively correlated with D * mono(r=0.224,P=0.042),and positively correlated with FAk(r=0.241,P=0.028).ROC curve analysis showed that MTA1 and TAF11 were risk predictors of PSM after RP,and the AUC of MTA1 was 0.693(sensitivity 73.9%,specificity 62.2%).The AUC of TAF11 was 0.643(sensitivity 84.5%,specificity 48.7%).However,there was no significant difference in AUC between MTA1 and TAF11(P=0.529).As in the first part of the study,Da,Kr,Dbi and statistically different preoperative clinicopathological characteristics PPC were introduced as independent predictors to develop a PSM prediction model MC based on quantitative parameters of DWI-derived sequences combined with preoperative clinicopathological characteristics PPC.Da,Kr,Dbi,PPC,MTA1 and TAF11 were introduced as independent predictors to develop a PSM prediction model BMC based on quantitative parameters of DWI-derived sequences,preoperative clinicopathological characteristics PPC and molecular biomarkers.AUC of MC model training set was 0.737(sensitivity 65.4%,specificity 78.1%),and AUC of BMC model training set was 0.811(sensitivity 88.5%,specificity 65.6%).The AUC of MC model validation set was 0.831(sensitivity 90.9%,specificity 78.6%),and the AUC of BMC model validation set was 0.883(sensitivity 100%,specificity 78.6%).There was no statistical difference in AUC between the two models(MC model training set vs BMC model training set P=0.177,MC model validation set vs BMC model validation set P=0.370).Comparing the calibration curves of the two models,the accuracy of the calibration curves of the training set and the validation set of the BMC model is better than that of the MC model.Comparing the DCA of the two models,in the current study,using BMC model to predict the clinical net benefit of PSM risk in the range of 0.2-0.8 is higher.Conclusion:(1)MTA1 and TAF11 can predict the occurrence of PSM after RP.(2)Compared with the PSM prediction model based on the quantitative parameters and pathological characteristics of DWI-derived sequences,the efficiency of the model was improved after the combination of molecular markers MTA1 and TAF11.(3)The expression of MTA1 and TAF11 in localized PCa tissue is correlated with the quantitative parameters of DWI and derivative sequence,which indicates that the quantitative parameters of DWI and derivative sequence can not only be used for the detection and grading of tumors,but also more likely to provide the biological characteristics of PCa before RP treatment,which is expected to become a potential biomarker of PCa biological image and provide a basis for the follow-up mechanism study. |
PDF Full Text Request