Artificial Intelligence-Design of Maxillary Single-tooth Dental Prostheses

Professor Lam, Yu Hang Walter   (Principal Investigator (PI))

Dr Koohimoghadam Mohamad   (Co-Investigator)

Emeritus Professor Wang Wen Ping   (Co-Investigator)

Professor McGrath Colman   (Co-Investigator)

Dr Luk Wai Kuen   (Co-Investigator)

Professor Tsoi Kit Hon   (Co-Investigator)

Dr Hsung Tai Chiu   (Co-Investigator)

30

2022-01-01

2025-06-27

945740

Artificial Intelligence-Design of Maxillary Single-tooth Dental Prostheses

Artificial intelligence, Dental Prostheses

Dentistry

Biology and Medicine (M)

17126021

General Research Fund (GRF)

2021

Completed

1 1. The overarching aim of this project is to determine whether Artificial Intelligence (AI) is better at predicting the occlusal morphology and 3D position of missing tooth/teeth based on the features of remaining dentition than conventional non-AI approaches. Dental prostheses should be designed to follow the natural teeth (biomimetric) when replacement of missing teeth. Occlusal morphology of dental prostheses are the most critical parameters when replacement of missing teeth. Based on our previous studies, we have developed virtual patient models (VPMs) to simulate individual patient’s jaw movements for designing the occlusal morphology of dental prostheses. Despite computer-assisted design (CAD) software are available for assisting the design of dental prostheses, considerable clinical time are still required to capture patient’s data to simulate jaw movement and significant human input are needed to optimize the occlusal morphology of dental prostheses. Teeth of an individual patients are genetically controlled and exposed to mostly identical oral environment, therefore the occlusal morphology and 3D position of teeth in a dental arch are inter-related. It is hypothesized that AI can automated designing the single-tooth dental prostheses from the features of remaining dentition. This is one of the pioneer studies of application of AI in Prosthodontics to determine the occlusal morphology and 3D position of natural teeth. This study use natural teeth as gold standard instead of use teeth generated by dental technicians. Considerable clinical time is usually required to adjust the occlusal surface of dental prostheses to fit into patient’s mouth (occlusion). 2 Objective 1: To compare four deep learning methods/algorithms in interpreting and learning the features of 3D models Currently, it is uncertain which deep learning methods/algorithms is the best in interpreting and learning the features of 3D models. Following our experience in using the AI to detect the gingivitis on 2D photography, the method/algorithm used will affect the outcome of AI system. This project will compare four methods/algorithms in interpreting the 3D models in deep learning. We will first input 200 maxillary dentate training models into our AI system. Then the computer will learn the relationship between individual tooth and remaining dentition. We will collect 100 maxillary dentate teeth models specifically for the validation process. One natural tooth will serve as control and will be removed. We will determine the best deep learning algorithm by compare the geometric morphometric (occlusal morphology) and 3D position of teeth generated by different deep learning methods/algorithms to that of the original natural tooth. The null hypothesis is deep learning methods/algorithms in interpreting the 3D models do not affect the AI prediction of the occlusal morphology and 3D position of single-missing teeth. 3 Objective 2: To compare the AI system with maxillary tooth model alone to maxillary and mandibular (antagonist) models We will determine if the maxillary arch alone or if both the maxillary and mandibular (antagonist) arch are needed for the prediction of single-missing tooth. In theory, the mandibular models provide additional information for the prediction of vertical height and 3D position of missing teeth in the maxillary models. The null hypothesis is deep learning with or without the antagonist (mandibular) teeth models do not result in any differences in the occlusal morphology and 3D position of single-missing tooth predicted by AI. 4 Objective 3a: To compare the occlusal morphology and 3D position of the single-tooth dental prostheses designed by trained AI and by dental technicians Objective 3b: To compare the time required for designing dental prostheses for replacement of single missing tooth by AI and by dental professionals The final objective of this study is to compare 1) the occlusal morphology and 3D position and 2) time required for the AI-designed and human-designed (manual and computer-assisted) single-tooth dental prostheses. The occlusal morphology and 3D position of the dental prostheses will be compared to that of the natural teeth. We will also record and compare the time spent for these design processes. The null hypothesis is when comparing to natural teeth, the occlusal morphology and 3D position of dental prostheses for replacement of single missing tooth designed by AI do not different to that designed by dental professionals. Moreover, the time spend in designing of dental prostheses by AI do not different to that designed by human.