Development and validation of a contrast-enhanced CT-based radiomics nomogram for preoperative diagnosis in neuroendocrine carcinoma of digestive system.

Objectives: To develop and validate a contrast-enhanced CT-based radiomics nomogram for the diagnosis of neuroendocrine carcinoma of the digestive system.

Methods: The clinical data and contrast-enhanced CT images of 60 patients with pathologically confirmed neuroendocrine carcinoma of the digestive system and 60 patients with non-neuroendocrine carcinoma of the digestive system were retrospectively collected from August 2015 to December 2021 at Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, and randomly divided into a training cohort (n=84) and a validation cohort (n=36). Clinical characteristics were analyzed by logistic regression and a clinical diagnosis model was developed. Radiomics signature were established by extracting radiomic features from contrast-enhanced CT images. Based on the radiomic signature and clinical characteristics, radiomic nomogram was developed. ROC curves and Delong's test were used to evaluate the diagnostic efficacy of the three models, calibration curves and application decision curves were used to analyze the accuracy and clinical application value of nomogram.

Results: Logistic regression results showed that TNM stage (stage IV) (OR 6.8, 95% CI 1.320-43.164, =0. 028) was an independent factor affecting the diagnosis for NECs of the digestive system, and a clinical model was constructed based on TNM stage (stage IV). The AUCs of the clinical model, radiomics signature, and radiomics nomogram for the diagnosis of NECs of the digestive system in the training, validation cohorts and pooled patients were 0.643, 0.893, 0.913; 0.722, 0.867, 0.932 and 0.667, 0.887, 0.917 respectively. The AUCs of radiomics signature and radiomics nomogram were higher than clinical model, with statistically significant difference (Z=4.46, 6.85, both < 0.001); the AUC difference between radiomics signature and radiomics nomogram was not statistically significant (Z=1.63, = 0.104). The results of the calibration curve showed favorable agreement between the predicted values of the nomogram and the pathological results, and the decision curve analysis indicated that the nomogram had favorable application in clinical practice.

Conclusions: The nomogram constructed based on contrast-enhanced CT radiomics and clinical characteristics was able to effectively diagnose neuroendocrine carcinoma of the digestive system.