Detecting missing teeth on PMCT using statistical shape modeling.

Area of Science:

• Forensic Dentistry
• Medical Imaging Analysis
• Computer Vision

Background:

• Accurate tooth identification in 3D medical images is crucial for victim identification and forensic investigations.
• Existing methods may struggle with incomplete or pathological dental data.

Purpose of the Study:

• To evaluate a novel tooth detection approach for mandibles using statistical shape models.
• To assess the performance of this method on diverse cases including missing teeth and pathologies.

Main Methods:

• Developed a statistical shape model from a complete mandible and teeth dataset.
• Fitted the model to target mandibles from CT images to reconstruct teeth and generate presence/absence maps.
• Evaluated accuracy on 76 diverse mandible cases.

Main Results:

• Achieved approximately 90% accuracy for front teeth (incisors, canines).
• Performance decreased for molars, particularly wisdom teeth, due to higher false-positive rates.
• The method relies solely on shape information, independent of imaging modality.

Conclusions:

• The approach provides reliable tooth count estimation (excluding wisdom teeth), identification, and reconstruction for forensic applications.
• It offers a non-modality-dependent solution applicable to various 3D scans and medical images.
• The method's target-agnostic nature allows potential adaptation for detecting missing parts in other organs.