Machine Learning-Based Localization and Quantification of Maximum Bite Force Using a Fiber Optic Mouthguard

Abstract

Bite force measurement is a critical parameter in dentistry and orthodontics, essential for the optimization of dental prostheses, diagnosis of bruxism and temporomandibular disorders, and monitoring of rehabilitation outcomes. While conventional bite force sensors offer limited spatial resolution and suffer from electromagnetic interference susceptibility, fiber optic sensors provide a compelling alternative due to their electrical passiveness, high sensitivity, and capability for dense distributed measurements. This thesis presents the novel application of machine learning to fiber optic bite force sensor data for automated localization and quantification of applied force on a dental mouthguard. The sensor platform used for this study is a silicone mouthguard (Sorta Clear 18), that contains an eight-branch scattering level multiplexing (SLMux) fiber optic network. This fiber optic network is then interrogated using a Luna Optical Backscattered Reflectometer (OBR) model 4600. In total there were 1056 readings taken from the mouthguard over the course of this study. These readings were obtained by uniformly distributing the application of twelve weights (100 – 1200g) across a matrix of eighty-eight points on the surface of the mouthguard. The raw output from the OBR was processed within a MATLAB pipeline and converted into 2D heatmap images. Those images were subsequently fed into machine learning algorithms.