Evaluation of Deep Learning Algorithm by Carotid Computed Tomography Angiography

This study was conducted to evaluate the reproducibility and plaque differences of the carotid artery by comparing Deep Learning-Brain Angiography Subtraction (DL-BAS) and the conventional Three Dimensional Digital Subtraction Angiography (3D-DSA) with Non-DSA images. Ninety-one patients who underwent head and neck CT angiography (CTA) were divided into the normal group, the experimental group with atherosclerotic plaque, and the experimental group with stent insertion. The circumference, area, and volume of the carotid lumen and the plaque volume were measured using the Distance Pair and Plaque Analysis tools of Aquarius iNtuition Edition software (Terarecon, California, USA). The images applied with DL-BAS and 3D-DSA techniques were compared with the Non-DSA images. As a result, there was no significant difference among the three groups in the normal group. However, in the group with atherosclerotic plaque and the group with stent insertion, the conventional 3D-DSA technique showed differences in carotid artery reproducibility and plaque differences. In conclusion, it is thought that the accuracy of diagnosis can be improved by using the Artificial Intelligence (AI)-based DL-BAS technique for follow-up examinations in patients with carotid atherosclerosis or stent placement.