Introducing a new auto edge detection technique capable of revealing cervical root resorption in CBCT scans with pronounced metallic artifacts

Abstract Cervical resorption is a serious threat to the longevity of the teeth. In this study, the Canny edge-detection algorithm was applied on CBCT images to compare the accuracy of original and Canny views for diagnosing cervical resorption in endodontically treated teeth. Intracanal metallic posts were inserted in 60 extracted teeth being randomly divided into three groups: control, 0.5 mm, and 1 mm cervical resorption. CBCT scans of the teeth were presented to three observers in both original and Canny formats with the accuracy being determined by receiver operating characteristic (ROC) analysis. The DeLong test was used for paired comparisons with the significance level set at 0.05. The highest accuracy belonged to Canny images in 1 mm resorption, followed by Canny images in 0.5 mm resorption, original images in 1 mm resorption, and original images in 0.5 mm resorption, respectively. The Canny images were significantly more accurate in the diagnosis of 0.5 mm (p