Automated orofacial virtual patient creation: A proof of concept

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Surgery

Background:

• Virtual patients (VPs) are crucial for treatment planning and communication.
• Current VP creation methods can be time-consuming and require manual adjustments.
• AI offers potential for automating and improving VP construction.

Purpose of the Study:

• To develop an AI-driven approach for constructing virtual patients using facial, intraoral, and CT scans.
• To quantitatively compare the accuracy, consistency, and efficiency of AI-driven VP registration against AI-refined and semi-automatic methods.
• To qualitatively assess user satisfaction with AI-generated VPs as a clinical communication tool.

Main Methods:

• Constructed VPs using an AI approach with facial, intraoral, and low-dose CT scans.
• Quantitatively analyzed VP accuracy, consistency, and time efficiency using Mimics software.
• Assessed user satisfaction via surveys with dentists and laypersons on realism and utility.

Main Results:

• AI-driven and AI-refined registration achieved 85% accuracy in the upper/middle face and 30% in the lower third.
• AI-driven registration was significantly faster (26.5s vs. 385s for semi-automatic), an 18-fold reduction.
• Users reported high satisfaction with VP realism and usefulness for communication and planning.

Conclusions:

• AI-driven registration provides a clinically accurate, fast, and consistent method for creating VPs from multimodal scans.
• Automated AI segmentation and registration of maxillofacial structures enhance clinical efficiency and accuracy for VPs.
• AI-powered VP creation facilitates widespread use in diagnosis, treatment planning, and communication.