Construction of a Contrast Agent Extravasation Prediction Model for Lung Cancer Patients Based on Decision Tree Algorithm

doi:10.12104/j.issn.1674-4748.2025.09.014

Abstract：Objective: To explore the influencing factors of contrast agent extravasation in lung cancer patients during CT examination and to construct a decision tree algorithm prediction model. Methods: A total of 205 lung cancer patients who underwent CT examination at the hospital from January 2023 to August 2024 were selected as the study subjects, and were divided into the extravasation group and the normal group according to whether contrast agent extravasation occurred. Univariate analysis and logistic regression analysis were performed, and a decision tree model was constructed. The predictive performance was evaluated using the receiver operating characteristic (ROC) curve. Results: Among the 205 patients, 23 experienced contrast agent extravasation, with an incidence rate of 11.22%. There were statistically significant differences between the two groups in terms of age, presence of diabetes, chemotherapy status, injection rate, injection site, and contrast agent temperature (P < 0.05). Logistic regression analysis showed that age, diabetes, chemotherapy, injection rate, and injection site (forearm small veins) were independent risk factors for contrast agent extravasation in lung cancer patients. The decision tree model included five features: diabetes, injection site, age, chemotherapy, and contrast agent temperature. The area under the ROC curve (AUC) of the decision tree model was 0.884, while that of the logistic regression model was 0.862. The Delong test result was Z = -0.608, P = 0.543. Conclusion: Age, diabetes, chemotherapy, injection rate, injection site (forearm small veins), and contrast agent temperature are influencing factors for contrast agent extravasation in lung cancer patients during CT examination. The decision tree prediction model constructed in this study has good performance and is beneficial for early and accurate identification of contrast agent extravasation.