Co-Mask R-CNN: collaborative learning-based method for tooth instance segmentation.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Artificial Intelligence

Background:

• Traditional tooth image analysis often misses crucial shape and position details.
• Existing methods may lack robustness and accuracy in segmentation.

Purpose of the Study:

• To develop an advanced computer-aided diagnosis method for precise tooth image analysis.
• To enhance tooth instance segmentation by integrating complementary image information.

Main Methods:

• Image enhancement to create edge-enhanced tooth images.
• A collaborative learning strategy using Mask Region-based Convolutional Neural Network (Co-Mask R-CNN).
• Simultaneous input of original and edge images with a two-stream encoder and attention-based feature fusion.

Main Results:

• Co-Mask R-CNN demonstrated superior segmentation accuracy and robustness compared to alternative networks.
• The method effectively leverages complementary information for enhanced tooth analysis.
• Experimental validation on a proprietary dataset confirmed the method's effectiveness.

Conclusions:

• Co-Mask R-CNN offers a promising approach for precise tooth segmentation in dental diagnostics.
• This method provides a reliable foundation for subsequent tooth disease diagnosis and treatment planning.
• The integration of complementary information significantly improves tooth image analysis outcomes.