What If the External Crown Surface of Teeth Could Predict the Pulp Chamber? A DeepSDF-Based Approach

Area of Science:

• Dental anatomy and morphology
• Artificial intelligence in medicine
• 3D reconstruction and modeling

Background:

• Accurate spatial localization of the pulp chamber is crucial for safe and effective dental treatments.
• Current methods often require invasive procedures or advanced imaging.
• Predicting internal anatomy from external surfaces offers a non-invasive alternative.

Purpose of the Study:

• To evaluate the performance of DeepSDF models in predicting the 3D pulp chamber space and cervical line of molars.
• To assess the feasibility of using only external crown surfaces for pulp chamber reconstruction.
• To validate the models on both micro-CT derived data and clinical intraoral scans.

Main Methods:

• A DeepSDF network was trained on 265 micro-CT derived 3D tooth models.
• The model learned a joint latent representation of external and internal tooth anatomy.
• Evaluation involved reconstruction from external surfaces and comparison with ground truth (micro-CT and CBCT) on independent datasets and clinical intraoral scans.

Main Results:

• DeepSDF models accurately reconstructed pulp chambers on micro-CT data (mean surface distance 0.18 mm).
• On clinical intraoral scans, pulp chamber reconstruction showed a median distance error of 0.49 mm.
• Deviations were noted at pulp horns, and predicted volumes were smaller than ground truth.

Conclusions:

• DeepSDF models establish consistent relationships between external tooth morphology and pulp anatomy.
• While local variability exists in clinical applications, the overall pulp chamber position is reliably predicted.
• These reconstructions offer valuable non-invasive visualization for digital endodontics and augmented reality-assisted procedures.