Accuracy and reliability of a fully automated landmark identification system on Cone Beam Computed Tomography

• Main Authors: Ghowsi, Ali, Hatcher, David, Suh, Heeyeon, Park, Joorok, Oh, Heesoo
• Format: Others
• Published: Scholarly Commons 2021
• Subjects: Automated Landmark; Accuracy; Reliability; CBCT; Dentistry; Endodontics and Endodontology; Orthodontics and Orthodontology
• Online Access: https://scholarlycommons.pacific.edu/dugoni_etd/17; https://scholarlycommons.pacific.edu/cgi/viewcontent.cgi?article=1019&context=dugoni_etd

ABSTRACT Introduction: Medical imaging continues to play an increasing role in health care and is an integral part of medicine and dentistry. Recent technological advancements have led to the development of fully automated landmark identification (ALI) systems capable of tracing Cone-Beam Computed Tomography (CBCT). The purpose of this study was to evaluate the accuracy and reliability of an ALI system as a tool for automatic landmark location compared to human judges. Methods: One hundred subjects’ CBCT volumes from multiple imaging centers were traced by two human judges who were calibrated and had an ICC close to 1. Fifty-three landmarks were identified in the x, y, and z coordinate planes using Checkpoint Software (Stratovan Corporation, Davis, CA). The ground truth was created by calculating the mean values of the x, y, and z coordinates for each landmark across both judges’ landmark identification. To evaluate the accuracy of ALI, the mean absolute error at each coordinate and mean error distance (mm) between the human landmark identification (ground truth) and the ALI were determined, and a successful detection rate (SDR) was calculated. Results: Overall, the ALI system was as successful at landmarking as the human judges with the exception of a few landmarks. The mean error distance for all 53 landmarks was 4.04 mm ± 6.5. Forty-nine out of 53 landmarks were located within a mean error of 4mm when the average for the coordinates of human judges was considered as a ground truth. Conclusion: Across all three coordinate planes, 96% of the landmarks had a mean absolute error of less than 4mm when compared to the ground truth. The ALI was more precise than humans when identifying landmarks on the same image at different times. This study demonstrates the promise of ALI in aiding orthodontists with landmark identification on CBCTs in the future.