Kinematic classification of mandibular movements in patients with temporomandibular disorders based on PCA

Abstract

This retrospective study aimed to kinematically classify mandibular movements collected during Temporomandibular Disorders (TMD) treatment, employing Fourier transformation (FT), Principal Component Analysis (PCA), and K-means clustering (k-means), and to investigate their correlation with symptoms of pain-related TMD. The study included five TMD participants diagnosed with myalgia (age: 39-86 years, with an SD of 18.96) and three healthy participants (age: 32-42 years, with an SD of 5.13) with no stomatognathic problems. TMD participants underwent tailored treatment for their symptoms, and their maximum unassisted mouth opening (MMO) was recorded randomly with a motion capture system (ARCUS digma II, Kavo, Biberach, Germany) at multiple time points. MMO for healthy participants served as a control. The dataset comprising 28 trials, was transferred to the AnyBody Modeling System (AnyBody Technology, Aalborg, Denmark) to extract joint angle time series, which were then transformed into Fourier series. Subsequently, PCA and k-means clustering were conducted. Two clusters were identified: Cluster 1, predominantly composed of symptomatic trials, and Cluster 2, mainly consisting of asymptomatic trials. Distinct transition pathways between the clusters were observed among participants, corresponding to the alleviation of pain-related symptoms during TMD treatment. These findings suggest that this approach has potential as an effective tool for diagnosing and assessing TMD by identifying symptomatic kinematic patterns and tracking temporal changes in mandibular movement. Despite the small dataset, these results suggest promise for a novel functional assessment method for TMD based on kinematic features.

Keywords: Fourier transformation (FT); K-means clustering (k-means); Kinematic analysis; Mandibular kinematics; Principal component analysis (PCA); Temporomandibular joint disorder (TMD).