Cox regression model for predicting surgical necrotizing enterocolitis based on abdominal radiographs radiomics features

Abstract Objectives To evaluate the ability of abdominal radiographs-based radiomics features to predict surgical risk and timing in neonates with necrotizing enterocolitis (NEC), providing an objective basis for clinical decision-making. Methods A retrospective cohort of 262 NEC patients (2016-2022) was divided into surgery and non-surgery groups. Radiomics features were extracted from abdominal radiographs at NEC diagnosis using LASSO and Cox regression to generate a radiomics risk score (Riskscore). The cohort was split into training (n = 184, 42 surgical cases) and test sets (n = 78, 18 surgical cases). Kaplan-Meier survival curves and Cox proportional hazards models were used to assess surgical risk. A nomogram with AUC was developed to predict surgical exemption over 1-7 weeks. Results The surgical group had lower gestational age and birth weight (P 0.001). Significant differences in survival curves were seen between high- and low-risk groups in both sets (P 0.001, P = 0.0047). Riskscore was a significant factor for surgical interventions (P 0.05). The nomogram model showed good performance, with AUCs of 0.716-0.744 in the training set and 0.668-0.733 in the test set. Conclusions Abdominal radiographs-based radiomics features analysis shows potential as a valuable tool for predicting surgical risk and timing in NEC patients. Advances in knowledge This study is the first to assess abdominal radiographs' value in predicting surgical risk and timing for NEC.