A CT-Based Radiomics Nomogram for Preoperative Prediction of Lymph Node Metastasis in Periampullary Carcinomas

Purpose: To establish and validate a radiomics nomogram for preoperatively predicting lymph node (LN) metastasis in periampullary carcinomas. Materials and Methods: A total of 122 patients with periampullary carcinoma were assigned into a training set (n = 85) and a validation set (n = 37). The preoperative CT radiomics of all patients were retrospectively assessed and the radiomic features were extracted from portal venous-phase images. The one-way analysis of variance test and the least absolute shrinkage and selection operator regression were used for feature selection. A radiomics signature was constructed with logistic regression algorithm, and the radiomics score was calculated. Multivariate logistic regression model integrating independent risk factors was adopted to develop a radiomics nomogram. The performance of the radiomics nomogram was assessed by its calibration, discrimination, and clinical utility with independent validation. Results: The radiomics signature, constructed by seven selected features, was closely related to LN metastasis in the training set (p < 0.001) and validation set (p = 0.017). The radiomics nomogram that incorporated radiomics signature and CT-reported LN status demonstrated favorable calibration and discrimination in the training set [area under the curve (AUC), 0.853] and validation set (AUC, 0.853). The decision curve indicated the clinical utility of our nomogram. Conclusion: Our CT-based radiomics nomogram, incorporating radiomics signature and CT-reported LN status, could be an individualized and non-invasive tool for preoperative prediction of LN metastasis in periampullary carcinomas, which might assist clinical decision making.

Automated summary

A radiomics nomogram was established and validated for preoperative prediction of lymph node metastasis in periampullary carcinomas. The radiomics signature, constructed by seven selected features, showed a close relationship to LN metastasis in both the training and validation sets. The nomogram, incorporating radiomics signature and CT-reported LN status, demonstrated favorable calibration, discrimination, and clinical utility for predicting LN metastasis.