| Gastric cancer(GC)is the most common malignant tumor of digestive tract in China,which seriously threatens national health.The status and stage of lymph node metastasis(LNM)is one of the most important indicators reflecting the progression of GC,and its quantitative analysis is of great value for prognostic determination,clinical diagnosis and treatment.The specificity of 18F-fluorodeoxyglucose positron emission tomography/computed tomography(18F-FDG PET/CT)in determining N staging(judging LNM)is much better than that of CT and magnetic resonanceimaging(MRI),with a specificity of more than 90%.The preoperative nomogram for predicting LNM constructed by 18F-FDG PET/CT radiomics combined with clinical risk factors,which provides additional information to the existing traditional methods for preoperatively evaluating N staging of GC.However,there are still many doubts and difficulties in the application of 18F-FDG PET/CT metabolic and radiomics models based on the primary tumor of GC to predict LNM.For this reason,we intend to carry out the following study.PartⅠRelationship between metabolic parameters of primary lesions based on 18F-FDG PET/CT and lymph node metastasis in gastric cancerObjective:The incidence and mortality of GC are high in China.Preoperative determination of LNM is an important basis for diagnosis and treatment decision in GC patients,which is still challenging presently.18F-FDG PET/CT imaging has been used in staging,restaging,treatment response evaluation,and detection of recurrence in GC.Previous studies have shown that 18F-FDG PET/CT provided valuable information for the detection of LNM,especially for distant metastatic lymph nodes in GC patients.Traditional criteria for diagnosing LNM in GC are mainly based on the LNs,including the short-axis diameter of LNs≥10 mm on the CT images or the LNs with 18F-FDG uptake similar to or higher than that of the liver on the PET images.Although 18F-FDG PET/CT has shown high specificity(73-92.2%)in the detection of LNM in GC,the sensitivity(40-54.7%)is limited.Therefore the previously evaluated parameters of 18F-FDG PET/CT are not precise enough to confirm LNM in GC,which made is urgent to find an imaging marker to identify LN status from the primary lesions preoperative accurately and promote clinical decision-making.The study was to explore and evaluate the specific dose-response relationship between LNM and total lesion glycolysis(TLG)of primary lesions determined by 18F-FDG PET/CT in patients with GC.Methods:This retrospective study included 106 GC patients who were examined by preoperative 18F-FDG PET/CT imaging in the Third Affiliated Hospital of Soochow University between April 2016 and April 2020.According to postoperative pathology,patients were divided into LNM group and non-LNM(NLNM)group.We measured maximum standardized uptake value(SUVmax),mean standardized uptake value(SUVmean),metabolic total volume(MTV)and total lesion glycolysis(TLG)of the primary gastric lesions.Multivariate logistic regression,a two-piece wise linear regression and smooth fitting curve were performed to evaluate the relationship between TLG of primary lesions and LNM.Results:Of the 106 patients,75 cases(71%)had LNM and 31 cases(29%)did ont have LNM.Univariate analyses revealed that a per-standard deviation(SD)increase in TLG was independently associated with LNM[odds ratio(OR)=2.37;95%confidence interval(CI):1.42-3.98,P=0.0010].After full adjustment of confounding factors,multivariate analyses exhibited that TLG of primary lesions was still significantly associated with LNM(OR per-SD:2.20,95%CI:1.16-4.19,P=0.0164).Generalized additive model indicated a non-linear relationship and saturation effect between TLG of primary lesions and LNM.When TLG of primary lesions<23.2,TLG was significantly correlated with LNM(OR=1.26,95%CI:1.07-1.48,P=0.0053);Whereas TLG of primary lesions≥23.2,the probability of LNM was greater than 60%,gradually reached saturation effect,as high as 80%or more.Conclusions:In this preliminary study,there were saturation and segmentation effects between TLG of primary lesions determined by preoperative 18F-FDG PET/CT and LNM.TLG of primary lesions can be used as an independent risk factor for LNM in GC patients.The higher TLG of primary lesions the higher risk of LNM,which indicated that TLG of primary lesions is helpful in the preoperative diagnosis of LNM in patients with GC.Part Ⅱ Incremental value of preoperative 18F-FDG PET-based radiomics of the primary lesions for prediction of lymph node metastases in gastric cancerObjective:PET metabolic parameters from the primary gastric lesions have a certain value in predicting LNM,but it is still controversial.PET-based radiomics extracts a large number of quantifiable image features from PET images with high throughput,visualizes tumor heterogeneity and obtains information related to efficacy prediction or prognosis to guide clinical decision-making.However,there are few studies on the prediction of LNM in GC.The purpose of this study was to evaluate the incremental value of preoperative prediction of LNM in GC based on the primary tumor radiomics characteristics of 18F-FDG PET.Methods:A retrospective analysis was performed on 127 patients who underwent preoperative 18F-FDG PET/CT examination in the Third Affiliated Hospital of Soochow University from January 2014 to September 2020 and were pathologically diagnosed as GC,including 89 males(70.08%)and 38 females(29.92%).The mean age was 65.63±10.33 years.All patients underwent radical gastrectomy within 14 days after PET/CT examination.According to postoperative pathology,the patients were divided into LNM group and NLNM group.LIFEx software was used to analyze and obtain primary PET metabolic parameters(including:SUVmax,SUVmean,MTV and TLG)and PET radiomics features of the primary gastric lesions.Radiomics signature was developed by least absolute shrinkage and selection operator(LASSO)and Radiomics score(Rad-score)was calculated.To further evaluate the incremental value of PET imaging radiomics features in predicting LNM in GC based on traditional risk factors and PET metabolic parameters from the primary gastric lesions,three models were constructed.Model 1 was composed of traditional risk factors,including gender,age,body mass index(BMI),fasting blood glucose,carcinoembryonic antigen(CEA),carbohydrate antigen 199(CA199)and C-reactive protein(CRP);Model 2 was model 1+primary PET metabolic parameters;Model 3 was model 2+Rad-score of the primary lesions based on PET.Receiver operating characteristic(ROC)curve analysis was performed.To evaluate the incremental value of primary Rad-score for predicting LNM by calculating integrated discriminant improvement(IDI)and net reclassification improvement(NRI)in GC.Results:Among the 127 GCpatients,95(74.8%)cases were diagnosed as LNM by postoperative pathology,and 32(25.2%)cases were NLNM.Compared with the NLNM group,the primary tumor SUVmean[5.30(3.88-8.61)vs 3.69(2.15-6.17)],SUVmax[9.38(6.66-15.03)vs 7.03(3.74-10.56)],MTV[14.89(8.47-20.22)cm3 vs 9.22(6.49-14.86)cm3],TLG[81.88(48.05-117.61)g vs 41.17(19.10-74.92)g]in the LNM group were significantly higher(z value:-3.823-0.053,all P<0.05).The Rad-score of the primary tumor in the LNM group was significantly higher than that in the NLNM group[0.35(-0.13-0.85)vs-0.61(-1.92--0.18)],and the difference was statistically significant(z=-5.842,P<0.001).Multivariate logistic regression showed that the Rad-score of the primary tumor was an independent factor for LNM[odds ratio(OR):6.38,95%CI:2.73-14.91,P<0.001].Model 1 predicted the area under the curve(AUC)of LNM to be 0.651.After adding the conventional metabolic parameters of PET,the predicted AUC of LNM was slightly improved,but not statistically significant(AUC 0.751 vs 0.651,P>0.05).Adding Rad-score on the basis of model 2,the discrimination of predicting LNM was significantly improved(AUC 0.882 vs 0.750,P=0.006),the best cut-off value of model 3 predicting LNM was 0.664,and the sensitivity,specificity,accuracy,negative predictive value(NPV)and positive predictive value(PPV)of predicting LNM were 0.895,0.813,0.874,0.722 and 0.935,respectively.Compared with Model 2,the calibration of Model 3 was significantly improved,with NRI of 0.293(P=0.004)and IDI of 0.293(P=0.005).Conclusions:The PET Rad-score of primary GC is an independent influencing factor of LNM.Adding PET-score on the basis of traditional risk factors and conventional PET metabolic parameters has a beneficial incremental value in predicting LNM,which is helpful for accurate preoperative staging and treatment decisions.Part Ⅲ Clinical study of radiomics nomogram based on preoperative 18F-FDG PET/CT in predicting lymph node metastasis in gastric cancerObjective:GC is the most common malignant tumor of the digestive system in China.LNM is not only one of the important factors affecting the prognosis of GC,but also an important basis for selecting personalized treatment plans.The status of LN has guiding significance on LN dissection,selection of neoadjuvant chemotherapy and other treatment strategies.However,the existing imaging diagnostic methods such as CT,endoscopic ultrasonography(EUS),MRI,18F-FDG PET/CT in the diagnosis of LNM in GC still need to be improved.It is necessary to find a non-invasive and accurate method for preoperative assessment of LNM risk.The purpose of this study was to investigate the value of the radiomics nomogram established by preoperative 18F-FDG PET/CT primary gastric lesions in predicting LNM based on the radiomics feature set combined with clinical risk factors.Methods:A total of 224 cases from two centers[The Third Affiliated Hospital of Soochow University(Research Center 1)and the First People’s Hospital of Yancheng City(Research Center 2)]were analyzed retrospectively.All the subjects performed preoperative 18F-FDG PET/CT examination,and the postoperative pathological diagnosis were GC.Of these,193 cases were from research center 1 and 31 cases were from research center 2.Cases from research center 1 were randomly divided into a training set(n=134)and an internal validation set(n=59)in a 7:3 ratio;Cases from research center 2 were used as an external validation set(n=31).The LIFEx software was used to segment and delineate the lesions from the 18F-FDG PET/CT images,and the radiomics features were extracted.The LASSO regression was used to screen out the effective features,and the Rad-score was calculated and established a radiomics prediction model.Multivariate logistic regression analysis was used to screen independent risk factors for LNM,and the Akaike’s information criterion(AIC)was used to select the optimal model parameters to construct a radiomics nomogram.ROC and decision curve analysis(DCA)were used to evaluate and validate the performance of nomograms for predicting LNM in training set,internal validation set and external validation set.Results:There were no significant differences between the two centers of the clinical data of patients(all P>0.05).The AUCs for predicting LNM based on the 18F-FDG PET/CTscore in the training set,internal validation set and external validation set were 0.792(95%CI:0.712-0.870),0.803(95%CI:0.681-0.924)and 0.762(95%CI:0.579-0.945),respectively.Multivariate logistic regression showed that CA199[(OR(95%CI):10.180(1.267-81.831)],PET/CT diagnosis of LNM[(OR(95%CI):6.370(2.256-17.984)],PET/CT Rad-score[(OR(95%CI):16.536(5.506-49.660)]were independent influencing factors for LNM(all P<0.05)in GC,and a radiomics nomogram was established based on this.The radiomics nomogram was trained during the training set.The AUCs for predicting LNM were 0.861(95%CI:0.799-0.924),0.889(95%CI:0.800-0.976)and 0.897(95%CI:0.683-0.948)in the training set,internal validation set and external validation set,respectively.The sensitivity,specificity and accuracy of the radiomics nomogram for predicting LNM were 85.4%,71.1%,and 80.6%in the training set,82.1%,85.0%,and 83.1%in the internal validation set,and 95.2%,80.0%,90.3%in the external validation set,respectively.DC A indicated that the nomogram based on 18F-FDG PET/CT radiomics has good clinical utility.Conclusions:Radiomics nomogram based on 18F-FDG PET/CT primary lesions can be effectively applied to preoperative LNM assessment of GC,which is helpful for risk stratification of GC patients and promotes individualized treatment.Part Ⅳ Radiomics model based on pre-operative 18F-FDG PET/CT predicts N2-3b stage lymph node metastasis in gastric cancer patientsObjective:According to the 8th edition of the American Joint Committee on Cancer(AJCC)TNM staging system,N staging of GC is divided into NO(no LNM),N1(1-2 LNMs),N2(3-6 LNMs),N3a(7-15 LNMs)and N3b(7-15 LNMs).The higher N stage of GC patients,the higher the mortality rate.In addition,different N stages have different treatment options.Patients with stage NO can choose less invasive procedures,such as endoscopic mucosal resection and endoscopic submucosal dissection,and do not require LN dissection.N1-2 stage patients need standard D2 LN dissection,and neoadjuvant chemotherapy is given priority for patients with stage N2 or higher.All GC patients underwent at least D1 LN dissection,so the N1 status did not affect the choice of surgical treatment,and it is more practical to determine the N2 or N3 status.The purpose of this study was to construct and verify the value of a radiomics nomogram model based on 18F-FDG PET/CT in predicting N2-3b stage LNM in GC.Methods:A total of 180 patients from the Third Affiliated Hospital of Soochow University who underwent 18F-FDG PET/CT before surgery and were pathologically confirmed to be GC were retrospectively collected for modeling and internal verification.According to the ratio of 7:3,127 patients from January 2014 to September 2020 were set as the training set for building the model,and 53 patients from October 2020 to August 2021 were set as the internal validation set.Thirty-one patients admitted to Yancheng First People’s Hospital who underwent 18F-FDG PET/CT examination before surgery and were pathologically confirmed to be GC after surgery were selected as the external validation set to verify the external prediction accuracy of the model.The LIFEx software was used to extract the PET and CT radiomics features,respectively.And the LASSO algorithm was used to select the radiomic features,and the Rad-score was calculated.Multivariate logistic regression analysis was used to screen independent risk factors for N2-3b stage LNM in GC,and the best predictors were screened based on AIC to construct a radiomics nomogram that combined Rad-score and clinical risk factors.Draw the ROC,calculate AUC to evaluate the predictive performance of each prediction model in the training set,internal validation set and external validation set,and use DCA to evaluate clinical utility of various nomogram prediction models.Results:There was no significant difference between the training set,internal verification set and external verification set of the clinical data of patients(all P>0.05).From the training set based on the radiomics features of 18F-FDG PET/CT,four features related to pathological N2-3b stage LNM were screened out,all of which were PET features,CONVENTIONAL SUVSkewness,GLCM_Correlation,GLZLM_LGZE and GLZLM_LZHGE.A prediction model of N2-3b stage LNM including Rad-score and CEA was constructed based on AIC criteria and presented in the form of a nomogram.The radiomics nomogram was more effective in predicting N2-3b metastasis in the training set[AUC:0.814(95%CI:0.740-0.888),P<0.001),the accuracy was 74.8%],and the nomogram in the internal validation set and the external validation set to predict N2-3b metastasis with AUCs of 0.810 and 0.779,and accuracies of 83.0%and 74.2%,respectively.DCA showed that the nomogram has strong clinical utility.Conclusions:In this study,we constructed and verified the radiomics nomogram based on 18F-FDG PET/CT images and CEA to predict N2-3b stage LNM,thus basis is provided for effectively predicting N2-3b stage LNM in GC before surgery. |
PDF Full Text Request