Automatic mandibular canal detection using a deep convolutional neural network.

Area of Science:

• Medical imaging and artificial intelligence.
• Deep learning applications in dentistry.

Background:

• Deep learning is increasingly used in healthcare for automated prediction and detection.
• Automated mandibular canal detection is crucial in dentistry for preventing nerve damage during surgery.

Purpose of the Study:

• To evaluate deep learning models for automated mandibular canal segmentation using Cone Beam Computed Tomography (CBCT).
• To explore the potential of 2D SegNet, 2D U-Net, and 3D U-Net for dental segmentation automation.

Main Methods:

• Experiments were conducted using preliminary automation software with 2D SegNet, 2D U-Net, and 3D U-Net models.
• Model performance was assessed based on global accuracy for mandibular canal detection.

Main Results:

• The 3D U-Net model achieved the highest global accuracy at 0.99.
• The 2D SegNet model demonstrated a global accuracy of 0.96.
• The 2D U-Net with adjacent images showed a global accuracy of 0.82.

Conclusions:

• Deep learning, particularly the 3D U-Net, shows high potential for accurate automated mandibular canal detection in CBCT scans.
• This technology can significantly improve dental treatment planning and reduce patient discomfort.
• Further research can explore deep learning for other dental imaging applications.