SinusC-Net for automatic classification of surgical plans for maxillary sinus augmentation using a 3D distance-guided...

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Maxillary sinus floor augmentation is crucial for dental implant placement in the posterior edentulous region.
• Accurate pre-operative planning is essential for successful sinus augmentation procedures.
• Current methods for classifying surgical approaches can be time-consuming and subjective.

Purpose of the Study:

• To develop and evaluate an automated deep learning model for classifying surgical approaches in maxillary sinus floor augmentation.
• To utilize a 3D distance-guided network on CBCT images for enhanced precision.
• To improve the efficiency and accuracy of pre-operative planning for dental implantology.

Main Methods:

• A modified ABC classification method with five surgical approaches was applied.
• A two-stage deep learning model, SinusC-Net, was developed, comprising landmark detection and surgical approach classification.
• Volumetric regression and a 3D distance-guided network were employed for detection and classification, respectively, on CBCT images.

Main Results:

• The model achieved a mean MRE of 0.87 mm for landmark detection with 95.47% SDR within 2 mm.
• SinusC-Net demonstrated high classification performance with mean accuracy of 0.97, sensitivity of 0.92, specificity of 0.98, and AUC of 0.95.
• The 3D distance-guided network proved effective for accurate 3D anatomical landmark detection and surgical approach classification.

Conclusions:

• The developed deep learning model (SinusC-Net) accurately detects 3D anatomical landmarks and classifies surgical approaches for maxillary sinus floor augmentation.
• Automated classification using 3D distance-guidance offers a precise and efficient tool for pre-operative planning in implant dentistry.
• This AI-driven approach has the potential to enhance surgical outcomes and streamline the planning process for dental implant placement.