Creation of virtual dental patients and the use of artificial intelligence in tooth replacement

Abstract

This thesis is composed of 9 chapters exploring 2 research directions. The first one was creating a Virtual Patient Model (VPM), including the determination of the maxilla position within a three-dimensional (3D) face, the hinge axis of the temporomandibular joint, and the ideal camera setting for capturing tooth color. The second direction reviewed the use of Artificial Intelligence (AI) in prosthodontics, developed a protocol for training and validating an AI system, and conducted two preliminary studies investigating the feasibility of the AI system. Chapter 1 is an introduction and provides a general overview of this thesis. Chapter 2 is a systematic review performed to determine Computer-Aided Design and Computer-Aided Manufacturing (CAD-CAM) techniques for the fabrication of digital complete dentures and their outcomes. From the 39 papers finally reviewed, the occlusal error was commonly seen with no consensus on methods of jaw registration found. Chapter 3 determined the accuracy of transferring a digital scan of maxillary teeth into a 3D face recorded by 3D stereophotogrammetry by using a virtual facebow. This is the first step in the creation of VPM. Forty-seven patients were recruited and the accuracy of the teeth-and-face reference relationship in VPM was found to be relatively small (<2 mm) when compared to that of the cone-beam computed tomography. n Chapter 4, a study was conducted to predict the hinge axis of casts mounted on an articulator by scanning two different jaw relationships, teeth closed and teeth slightly apart, and from this calculating the rotational center of the hinge axis. The estimated hinge axis discrepancy was found <5 mm when compared with the hinge axis of the articulator. In Chapter 5, a study was performed to investigate the accuracy of color captured by different settings of a professional Single-lens Reflex (SLR) camera and a smartphone camera. Compared with the spectrophotometer, the SLR camera has better color accuracy than a smartphone camera. Chapter 6 is a narrative review of the usage of AI in prosthodontics and Chapter 7 aims to develop a protocol for training and validating a novel AI system for designing dental prostheses. In Chapter 8, two preliminary studies investigated the use of AI in designing biomimetic occlusal surfaces of maxillary first molar dental prostheses by learning the features of adjacent remaining teeth. In the first study, 10 AI-designed dental prostheses were mixed and matched to their original natural teeth. The chance of the correct match was 1/10 for each pair. The result is promising with over 60% correct matching. The second study was a comparison of tooth morphology between dental prostheses designed by CAD software (Control) and by AI. The differences from the original teeth, in terms of median Hausdorff distance, AI-designed dental prostheses was 0.366 mm, while that of CAD-designed dental prostheses was 0.304 mm.