Currently, the use of dental devices and prosthetics requires individuals to adjust their oral behaviour so that activities such as speech and eating feel as natural as possible. However, such adjustments often end up producing unnatural speech patterns and/or unusual oral movements leading to complications, like an irregular bite.

Recent advancements in dental technologies have led to the realisation of new approaches allowing the effective development of customised dental devices and adhesive formulations. Such personalised devices require minimal adjustments, whilst achieving natural oral movements. The primary aim of this research is to develop technologies and approaches that will allow us to realise this goal. To achieve this, detailed models of the mouth and relevant facial/oral structures need to be developed, helping us understand the changes in speech and food processing that occur in a denture wearing individual. Although technologies currently exist that can 3D model the face by recording its outside, they are not able to also form and illustrate the movements of inside the mouth during speech and other oral movements.

This leads to two main research questions. Firstly, is it possible to 3D-model the movements of the tongue during speech and food-processing, by recording only the head and neck with a 360-degree camera? Secondly, how can this technology be applied in visualising the effect of dental prosthetics on one’s oral motion?

We have found real time MRI recordings of the mouth to be most suitable for our aim, capable of displaying the movements of each articulator, both internal and external, but in a 2-dimensional plane. The research aims to train machine learning models to analyse the two-dimensional real-time MRI recordings to form a 3D representation; In essence this can be likened to the task of creating audio data from a video of the mouth moving during speech (predicting speech).

The research explores the development of novel computer vision algorithm for constructing a representation of the mouth’s internal elements by mapping the 2D/3D view or action (i.e. time domain) of the external elements.