Automated landmark identification on cone-beam computed tomography: Accuracy and reliability

Area of Science:

• Dentistry
• Medical Imaging
• Orthodontics

Background:

• Accurate landmark identification on cone-beam computed tomography (CBCT) is crucial for orthodontic diagnosis and treatment planning.
• Manual landmark identification by human judges can be time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To evaluate the accuracy and reliability of a fully automated landmark identification (ALI) system.
• To compare the performance of the ALI system against human judges for landmark localization on CBCT images.

Main Methods:

• 100 CBCT images were analyzed.
• Two human judges identified 53 landmarks in 3D coordinates.
• The ground truth was established by averaging human landmark coordinates.
• Accuracy was assessed using mean absolute error and mean error distance, with a success rate calculation.

Main Results:

• The ALI system achieved an average mean absolute error of 1.57 mm across all coordinates.
• 94% of landmarks had a mean absolute error under 3 mm.
• The system demonstrated a 75% success rate in detecting landmarks within a 4 mm error distance.

Conclusions:

• The ALI system exhibited clinically acceptable mean error distances, comparable to human judges.
• The ALI system demonstrated higher precision than humans for repeated landmark identification on the same image.
• The study highlights the potential of ALI as a valuable tool for orthodontists in CBCT landmark identification.