Artificial Intelligence for Fast and Accurate 3-Dimensional Tooth Segmentation on Cone-beam Computed Tomography

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Tooth segmentation in cone-beam computed tomographic (CBCT) imaging is challenging due to low contrast and artifacts.
• Manual segmentation is time-consuming and labor-intensive.
• Current methods struggle with fully automated tooth segmentation based solely on CBCT intensity.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI)-driven tool for automated tooth segmentation on CBCT images.
• To improve the efficiency and accuracy of tooth segmentation in dental imaging.
• To explore AI applications in oral healthcare treatment planning.

Main Methods:

• Developed an AI-driven algorithm using a feature pyramid network for automatic tooth detection and segmentation.
• Utilized 433 CBCT scans of single- and double-rooted teeth for algorithm development and validation.
• Evaluated the AI tool using volume comparison, intersection over union, Dice score, surface deviation, and time efficiency.

Main Results:

• AI-driven segmentation demonstrated high accuracy, with mean intersection over union scores of 0.87-0.88.
• The AI tool achieved minimal average median surface deviation (7.85-9.96 μm) compared to manual segmentation.
• Automated segmentation was significantly faster, with fully automated AI (F-AI) being 12 times faster than manual methods.

Conclusions:

• The developed AI tool offers a fast and accurate approach for automated tooth segmentation on CBCT imaging.
• This AI-driven method shows potential for enhancing surgical and treatment planning in oral healthcare.
• The findings support the integration of AI in routine dental diagnostic workflows.