Transformative potential of deep learning in oral surgery

Abstract:

Cone-beam computed tomography (CBCT) is a cornerstone of modern digital dentistry, enabling 3D diagnosis and treatment planning for implantology, orthodontics, and oral surgery. While AI-driven automatic tooth segmentation holds promise for transforming these workflows, the clinical impact of CBCT-based tooth segmentation is critically limited by inadequate accuracy in rendering dental crowns, due to inherent image artifacts and low resolution. Although fusing CBCT with high-resolution optical scans (from intraoral or model scanners) mitigates this issue, this multi-modal approach introduces procedural complexity and cost. The purpose of this interdisciplinary project is to develop a novel deep learning framework that leverages optical scan data as a high-fidelity “teacher” to enhance CBCT segmentation accuracy. Our ultimate vision is to create a generative model that produces precise, model-free CBCT segmentations, thereby streamlining digital dental workflows.

Skills and experience required for the project:

This is an interdisciplinary project at the intersection of AI and dentistry, welcoming students from Engineering, Dentistry, and related fields.

Ideal candidates are expected to possess expertise in any of the following areas:

Technical AI & Programming:
Proficiency in Python and deep learning frameworks like PyTorch or TensorFlow.
Experience with 3D CNN architectures (e.g., 3D U-Net), Transformers (e.g., Swin Transformer), and Generative AI (e.g., GANs).

Medical Image Processing:
Experience with medical imaging libraries (e.g., ITK, SimpleITK) and handling DICOM files.
Understanding of medical image segmentation metrics (e.g., Dice coefficient, Hausdorff distance).

Dental Domain Knowledge:
Understanding of CBCT imaging principles, including common artifacts and anatomical challenges.
Knowledge of tooth morphology, the FDI tooth numbering system, and key anatomical structures.

Interested students with a relevant background are encouraged to make contact for further discussion.