A Nomogram Based on Pretreatment Radiomics and Dosiomics Features for Predicting Overall Survival for Esophageal Squamous Cell Cancer: Multi-Institutional Study.

Abstract

The current study aims to propose a nomogram-based 2- and 3-years survival prediction model for esophageal squamous cell carcinoma treated by definitive radiotherapy using pretreatment computed tomography (CT), positron emission tomography (FDG PET) radiomic features and dosiomics features in addition to the common clinical factors using multi-institution data. Data of 112 patients from one institution and 28 patients from the other institution were retrospectively collected. Radiomics and dosiomics features were extracted using five segmentations on CT and PET images and dose distribution. The least absolute shrinkage and selection operator (LASSO) with logistic regression was used to select radiomics and dosiomics features by calculating the radiomics and dosiomics scores (Rad-score and Dos-score), respectively, in the training model. The predictive clinical factors, Rad-score, and Dos-score were identified to develop a nomogram model. We extracted 15219 features from the radiomics and dosiomics analysis. By LASSO Cox regression analysis, 13 CT-based radiomics features, 11 PET-based radiomics features, and 19 dosiomics features were selected. Clinical factors of T-stage, N-stage, and clinical stage were selected as significant prognostic factors by univariate Cox regression analysis. A predictive nomogram for prognosis in was established using these factors. In the external validation cohort, the C-index of the combined model of CT-based radiomics, PET-based radiomics, and dosiomics features with clinical factors were 0.74, 0.82, and 0.92, respectively. Moreover, we divided the cohort into high-risk and low-risk groups using the median nomogram score. Significant differences in overall survival (OS) in the combine model of CT-based radiomics, PET-based radiomics, and dosiomics features with clinical factors were observed between the high-risk and low-risk groups (P = 0.019, P = 0.038, and 0.014, respectively). The current study established and validated 2- and 3-year survival prediction models based on radiomics and dosiomics features with clinical factors. The prediction model with dosiomics analysis could better predict OS than CT- and PET-based radiomics analysis in esophageal cancer patients treated with radiotherapy.