Joint Craniomaxillofacial Bone Segmentation and Landmark Digitization by Context-Guided Fully Convolutional Networks.

Area of Science:

• Medical Imaging
• Computer-Aided Surgery
• Biomedical Engineering

Background:

• Accurate 3D models from CBCT are crucial for craniomaxillofacial (CMF) deformity treatment planning.
• Current methods often treat bone segmentation and landmark digitization as separate tasks, ignoring their correlation.
• Existing approaches fail to utilize spatial context information within CBCT images.

Purpose of the Study:

• To develop a novel method for joint bone segmentation and landmark digitization using CBCT images.
• To incorporate spatial context information, specifically displacement maps, into the modeling process.
• To improve the accuracy of 3D model generation for CMF applications.

Main Methods:

• A context-guided fully convolutional network (FCN) was proposed.
• An initial FCN was trained to learn displacement maps, capturing spatial context.
• A multi-task FCN was developed to perform joint bone segmentation and landmark digitization, guided by displacement maps.

Main Results:

• The proposed method was evaluated on 107 subjects across two centers.
• Experimental results demonstrated superior performance compared to state-of-the-art methods.
• The joint approach showed significant improvements in both bone segmentation and landmark digitization accuracy.

Conclusions:

• Jointly performing bone segmentation and landmark digitization with spatial context improves accuracy.
• The proposed context-guided FCN offers a more effective approach for 3D modeling in CMF deformities.
• This method enhances the foundation for precise treatment planning in CMF surgery.