The Application Value Of Dual-energy CT-based Radiomics In Predicting The Response Of Chemotherapy In Advanced Gastric Adenocarcinoma

Objective:1.The aim of our study was to develop a baseline dual-energy computed tomography(DECT)based radiomics nomogram to predict the clinical response of advanced gastric cancer patients to systemic chemotherapy.2.To investigate the valued-added of multi-energy image based radiomic model for the prediction of clinical response by comparing with different single-energy image-based radiomics model and conventional clinical mode.Materials and methodsRetrospective analysis the patients who were confirmed to be gastric adenocarcinoma by pathology biopsy and met the inclusion criteria.A total of 69 patients were included and analyzed in the current study,including 48 males and 21 females,age range 23-84 years,median age 56 years.Patients were divided into a chemotherapy response group and a chemotherapy non-response group according to Response Evaluation Criteria in Solid Tumors(RECIST v.1.1).Collecting clinical information of patients:age,sex,clinical stage and so on.CT parameters of patients was analyzed and recorded:On the iodine water map automatically generated by GSI viewer software,the iodine concentration(IC)in venous phase of gastric cancer was measured,and normalized iodine concentration(NIC)was calculated.The maximum thickness of the lesions was measured,and Borrmann type was also assessed.The cN and cM stages of tumors were assessed according to the 8th edition of AJCC guidelines.Radiomics software(Radiomics 1.0.9a,Siemens Healthineers,Germany)on a research platform(Syngo.Via VB10,Research Frontier,Siemens Healthineers,Germany)was used to complete volumes of interest in 3D tumor segmentations and feature extraction on both 70keV energy images and were automatically matched to 40keV energy image and lOOkeV energy images.A total of 1691 radiomics features were extracted from each patient,including 17 shape features,324 first-order features and 1350 texture feature.Univariate logistic regression and least absolute shrinkage selection operator(LASSO)were used for important radiomics feature selection.Multivariate logistic regression analysis was used to establish the radiomics prediction model.Firstly,univariate and multivariate logistic regression analysis was performed clinical features to identify independent clinical predictors of response evaluation and then to construct a clinical prediction model.Then,three single-energy models and one multi-energy model(Full)were established and compared.Finally,clinical models were combined with full radiomics model to establish a mixed model and presented in the form of a nomogram.The predictive performance was validated by receiver operating characteristic curve(ROC)analysis,calibration curve,and decision curve analysis(DCA).Result1.A total of 69 patients were included in the current study.The ratio of male to female was 16:7.Most of the patients were in stage IV(n=47,68.1%),and most of the lesions were single.On CT,more than 80%of the patients with lymph node positivity,and more than 50%of the patients with distant metastases.Most of the patients were classified as type III or IV by Borrmann classification.Based on current research,we set up new response evaluation criteria especially for upcoming radiomic analysis as following:24 patients with CR or PR by RECIST(v.1.1)after six cycles of treatment were classified into effective groups(chemo-sensitive).However,54 patients with PD or SD after six cycles of chemotherapy were classified into non-effective groups(chemo-insensitive).2.After features screening,we obtained 8,6,3 and 11 most significant radiomics features for model building.The AUC of the 40-,70-,100-,Full radiomics model was 0.747(95%CI:0.628-0.866),0.793(95%CI:0.678-0.90),0.881(95%CI:0.791-0.971),0.914(95%CI:0.846-0.982)respectively.The multi-energy model has the highest prediction efficiency.3.The clinical-radiomics nomogram integrated the multi-energy radiomics model with IC value and clinical stage with an AUC of 0.934,and its predictive efficiency was higher than single-energy,multi-energy and clinical model.4.The point-0.48 was taken as the threshold value,the accuracy,sensitivity,specificity,positive and negative rate of the nomogram in data set was 88.4%,83.3%,91.1%,83.3%and 91.1%,respectively.5.The calibration curve showed that the nomogram had high accuracy,and good calibration in evaluating the efficacy of chemotherapy.On the other hand,the decision curve showed that both the nomogram and the multi-energy model had high clinical benefit and application value.Conclusion1.The current study firstly established a nomogram based on DECT multi-energy images for the prediction of clinical response to systemic chemotherapy in advanced gastric adenocarcinoma.2.The nomogram integrated clinical stages,DECT quantitative parameters and CT radiomics features based on multi-energy images,and performed well in predicting clinical response to systemic chemotherapy in patients with advanced gastric adenocarcinoma,which may be helpful in clinical decision-making and improve patient survival.