Prediction model based on preoperative CT findings for carotid artery invasion in patients with head and neck masses.

Objectives: To investigate the performance of a model in predicting carotid artery (CA) invasion in patients with head and neck masses using computed tomography (CT).

Methods: This retrospective study included patients with head and neck masses who underwent CT and surgery between January 2013 and July 2021. Patient characteristics and ten CT features were assessed by two radiologists. The patients were randomly allocated to a training cohort (n=106) and a validation cohort (n=109). Independent risk factors for CA invasion were assessed by univariate and multivariate logistic regression analyses. The predictive model was established as a nomogram using the training cohort. In addition, the calibration, discrimination, reclassification, and clinical application of the model were assessed in the validation cohort.

Results: A total of 215 patients were evaluated, including 54 patients with CA invasion. Vascular wall deformation (odds ratio [OR], 7.17; p=0.02) and the extent of encasement to the CA (OR, 1.02; p<0.001) were independent predictors of CA invasion in the multivariable analysis in the training cohort. The performance of the model was similar between the training and validation cohort, with an area under the receiver operating characteristic curve of 0.93 (95% confidence intervals [CI], 0.88-0.98) and 0.88 (95% CI, 0.80-0.96) (p=0.07), respectively. The calibration curve showed a good agreement between the predicted and actual probabilities.

Conclusion: A predictive model for carotid artery invasion can be defined based on features that come from patient characteristics and CT data to help in improve surgical planning and invasion evaluation.