Clinical-oriented 3D visualization and quantitative analysis of gingival thickness using convolutional neural...

Area of Science:

• Biomedical Engineering
• Dental Imaging
• Artificial Intelligence in Medicine

Background:

• Traditional gingival thickness (GT) assessment lacks spatial detail, hindering precise implant surgery planning.
• Current methods offer only point measurements or basic classifications, failing to capture 3D variations.
• A need exists for advanced tools to provide comprehensive spatial GT data.

Purpose of the Study:

• To develop a Convolutional Beam Computed Tomography (CBCT)-based 3D visualization system for gingival thickness assessment.
• To utilize deep learning for creating a novel spatial assessment tool for dental implant surgery.
• To improve the accuracy and predictability of implant procedures through enhanced GT evaluation.

Main Methods:

• Collected CBCT and intraoral scanning (IOS) data from 50 patients.
• Employed DeepLabV3+ architecture for semantic segmentation of gingival and bone tissues.
• Developed a 3D visualization algorithm transforming 2D slices into continuous 3D surfaces with gradient color mapping.

Main Results:

• Achieved 85.92% mIoU for semantic segmentation accuracy.
• Successfully constructed a 3D spatial distribution model of gingival thickness.
• Demonstrated millimeter-precision quantification and multi-angle GT assessment, overcoming 2D limitations.

Conclusions:

• The developed system advances GT assessment from qualitative to spatial quantitative analysis.
• This 3D visualization tool aids in identifying high-risk areas for personalized surgical planning.
• The system enhances predictability for aesthetic and complex dental implant cases.