Dual-Stage Deeply Supervised Attention-Based Convolutional Neural Networks for Mandibular Canal Segmentation in CBCT...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oral and Maxillofacial Surgery

Background:

• Accurate segmentation of the mandibular canal in 3D CT scans is crucial for safe dental implant placement.
• Manual segmentation is time-consuming and prone to errors, potentially risking damage to the mandibular nerve.
• Existing automated methods may lack the precision required for clinical applications.

Purpose of the Study:

• To develop and validate a novel dual-stage deep learning scheme for automatic mandibular canal segmentation.
• To improve the accuracy and robustness of mandibular canal segmentation in Cone Beam Computed Tomography (CBCT) scans.
• To provide a clinically acceptable automated solution for dental implantology.

Main Methods:

• A dual-stage deep learning approach was proposed, starting with histogram-based dynamic windowing for CBCT scan enhancement.
• A 3D deeply supervised attention UNet architecture was used for localizing relevant Volumes of Interest (VOIs) containing the mandibular canals.
• A Multi-Scale input Residual UNet (MSiR-UNet) was employed for accurate final segmentation within the identified VOIs.

Main Results:

• The proposed method demonstrated improved performance in mandibular canal segmentation compared to existing techniques.
• The technique achieved results within a clinically acceptable range on both private and public CBCT datasets.
• The segmentation method proved robust across different CBCT scan types and fields of view.

Conclusions:

• The novel dual-stage deep learning scheme offers an accurate and robust solution for automatic mandibular canal segmentation.
• This automated approach can enhance the safety and efficiency of dental implant procedures.
• The method shows significant potential for clinical adoption in dental implantology.