Objective:

To explore the value of contrast-enhanced computed tomography (CECT)-based radiomics in predicting lymph node metastasis following neoadjuvant chemotherapy (NAC) in patients with advanced gastric carcer.

Methods:

Data of 226 advanced g

astric cancer patients who underwent NAC and radical resection were collected retrospectively between August 2012 and October 2020 from two cancer hospitals. Clinicopathological and radiological data of the enrolled patients were collected retrospectively. The enrolled patients were divided into the absence of lymph node metastasis and the presence of lymph node metastasis groups based on postoperative pathology results. Radiomics features were initially extracted from CECT images. Subsequently

radiomics predictive models were constructed using machine learning software respectively

which were then externally validated by the external independent validation set. Clinicopathological parameters of the enrolled patients were collectedand analyzed retrospectively. Univariate and multivariate logistic regression analyses were performed to screen for independent predictors. Finally

an integrated model combining clinicopathological predictive predictors and radiomics features was developed.

Results:

The area under curves (AUCs) of the radiomics predictive model based on arterial phase images achieved 0.972 and 0.701 for the training and external testing cohorts

respectively. The AUCs of the radiomics predictive model based on portal venous phase images achieved 0.917 and 0.736 for the training and testing cohorts

respectively. However

the AUC values of the radiomics predictive model based on the combination of arterial and venous phase images achieved 0.862 and 0.684 for the training and testing cohorts

respectively. Multivariate logistic regression analyses showed that there were no independent predictors of lymph node metastasis after NAC in advanced gastric cancer patients.

Conclusion:

Radiomics model of pretreatment CECT images can perform consistently in predicting lymph node metastasis after NAC in advanced gastric cancer patients. It could be served as a non-invasive biomarker to provide clinicians critical information for designing individualized therapies.