Automated detection of carotid artery encasement on CT scans using an AI-driven approach

Carotid artery encasement can complicate the surgical resection of head and neck squamous cell carcinomas, increasing the risk of surgery. Incomplete tumor excision due to the proximity of the carotid artery may affect long-term outcomes by promoting tumor recurrence. This study leverages artificial Intelligence to detect carotid artery encasements in computed tomography images. CT images and tumor masks were sourced from the National Cancer Institute’s CPTAC-HNSCC dataset. The study included 100 patients, yielding 1,367 images with tumor presence. For HNSCC segmentation, a pre-trained ResNet-50 backbone of a U-Net model was fine-tuned on the CT images. Carotid segmentation was performed using the SAM-Med2D foundation model. Both models returned segmentations for the tumor and carotid, respectively. Encasement was assessed by calculating the minimum distance and encasement angle between the masks provided by the models. In validation, the tumor segmentation model achieved a mean F1 score of 86% and an Intersection over Union score of 75%, with 73% IoU for the tumor class and 99% for the non-tumor class. Carotid segmentation was most effective when using the bounding-box method, reaching 79% accuracy. Both models were successfully applied to the same image for the final assessment. This study developed an automated method for evaluating CAE in CT images, demonstrating promising potential for future clinical applications.