To Predict Treatment Efficacy After Chemoradiotherapy Among Esophageal Squamous Cell Carcinoma Patients By CT Radiomics Features

Objective: Radiomics features extremly related to treatment efficacy were extracted and screened from CT images of patients with esophageal squamous cell carcinoma before chemoradiotherapy.Statistical model was constructed and validated to predict treatment effficacy.Method: Patients with esophageal squamous cell carcinoma who received chemoradiotherapy in two domestic hospitals were retrospectively collected and divided into training group(n=318)and verification group(n=77).RECIST 1.1(response evaluation criteria for solid tumor treatment)was used to evaluated treatment efficacy based on enhanced CT after 3 months of review.The collected image datas were determined the Volume of Interest(VOI)with 3D Slicer by two experienced CT physicians.Radiomics features were extracted with IBEX toolkit.Lasso-Logistic regression was used to filtrate the radiomics features and set the radiomics label(Rad＿score).Multivariate analysis was used to establish and verify the efficacy prediction model.The prediction model was valuated in the end of this research.Results: With effective relief and no relief,there were 259 cases and 136 cases respectively.With concurrent chemoradiotherapy and radiotherapy alone,there were 237 cases and 158 cases,respectively.Lasso-Logistic regression screened out 4 radiomics features were set as radiomics labels that could predict the efficacy of chemoradiotherapy.Multivariate Logistic regression in training group showed that the statistical model that clinical stages combined with radiomics labels could better predict the treatment efficacy.The AUC values were 0.773(95%CI: 0.655-0.890)and0.693(95%CI: 0.631-0.755),respectively,in the training group and the verification group.It was better than that of separate radiomics label model and separate clinical stages model.P value of Nomogram calibration curve was 0.990(P >0.05).The AUC values were 0.823 and 0.714 in the PSM and IPTW among patients of concurrent chemoradiotherapy.The AUC values of radiotherapy alone were 0.817 and 0.708.Conclusion: CT radiomics features before chemoradiotherapy can predict the efficacy of esophageal squamous cell carcinoma patients who recieved Concurrent chemoradiotherapy and radiotherapy alone.