[Development of a surface electromyography index system for orofacial muscles and validation of a discriminant model in unilateral molar occlusal interference].

OBJECTIVE

To construct a standardized unilateral molar occlusal interference model, to establish a comprehensive surface electromyography (sEMG)-based index system for orofacial muscle function, and to develop an accurate discriminant model, thereby providing an objective electrophysiological basis for occlusal interference diagnosis.

METHODS

Twenty-six healthy adult volunteers were recruited and provided written informed consent. Utilizing advanced digital dental technology, including intraoral scanning, computer-aided design (CAD), and additive manufacturing, a standardized occlusal inter-ference patch with a precise thickness was fabricated. This patch was adhesively bonded to the occlusal surface of the mandibular first molar to create a reversible unilateral occlusal interference model. A self-developed, multi-channel wireless sEMG system was employed to collect high-fidelity electromyographic signals from key bilateral masticatory muscles: the anterior temporal muscles, masseters, and the anterior bellies of the digastric muscles. Data were recorded during 10 standardized mandibular functional activities both before (baseline) and after the induction of interference. From the raw sEMG signals, a multi-dimensional index system comprising 56 distinct indicators across time, frequency, and complexity domains was constructed. Sophisticated statistical analyses, including paired-sample t-tests (or Wilcoxon signed-rank tests), principal component analysis (PCA) for dimensionality reduction, and stepwise Logistic regression analysis, were applied to screen for the most significant feature variables and to build the optimal discriminant model.

RESULTS

Forty valid interference models were successfully established. Statistical analysis revealed 253 sEMG indicators showed significant differences following interference induction (P < 0.05), with lateral-movement-related parameters demonstrating particular sensitivity. PCA extracted 19 principal components (PCs) explaining 85.5% of cumulative variance, where PC1 (muscle fatigue level) and PC2 (functional movement amplitude) represented the primary explanatory components. The optimal Logistic regression model incorporated 3 principal components. Cross-validation showed the model achieved a mean accuracy of 0.840, with mean sensitivity and specificity of 0.851 and 0.828, respectively, and a mean area under the curve (AUC) of 0.923.

CONCLUSION

The Logistic regression discriminant model for unilateral molar occlusal interference constructed in this study can effectively identify the occlusal interference state under the experimental conditions, demonstrating promising diagnostic potential.