BACKGROUND Temporomandibular disorders (TMDs) are pathological conditions affecting the temporomandibular joint and/or masticatory muscles. The current diagnosis of TMDs is complex and multi-factorial, including questionnaires, medical testing and the use of diagnostic methods, such as computed tomography and magnetic resonance imaging. The evaluation, like the mandibular range of motion, needs the experience of the professional in the field and as such, there is a probability of human error when diagnosing TMD. The aim of this study is therefore to develop a method with infrared cameras, using the maximum range of motion of the jaw and four types of classifiers to help professionals to classify the pathologies of the temporomandibular joint (TMJ) and related muscles in a quantitative way, thus helping to diagnose and follow up on TMD. METHODS Forty individuals were evaluated and diagnosed using the diagnostic criteria for temporomandibular disorders (DC/TMD) scale, and divided into three groups: 20 healthy individuals (control group CG), 10 individuals with myopathies (MG), 10 individuals with arthropathies (AG). A quantitative assessment was carried out by motion capture. The TMJ movement was captured with camera tracking markers mounted on the face and jaw of each individual. Data was exported and analyzed using a custom-made software. The data was used to identify and place each participant into one of three classes using the K-nearest neighbor (KNN), Random Forest, Naïve Bayes and Support Vector Machine algorithms. RESULTS Significant precision and accuracy (over 90%) was reached by KNN when classifying the three groups. The other methods tested presented lower values of sensitivity and specificity. CONCLUSION The quantitative TMD classification method proposed herein has significant precision and accuracy over the DC/TMD standards. However, this should not be used as a standalone tool but as an auxiliary method for diagnostic TMDs.

中文翻译：

---

通过生物力学的面部特征识别颞下颌综合征的关节炎和肌病。

背景技术颞下颌疾病（TMD）是影响颞下颌关节和/或咀嚼肌的病理状况。当前对TMD的诊断是复杂和多因素的，包括调查表，医学检测以及使用诊断方法（例如计算机断层扫描和磁共振成像）。评估，如下颌运动范围一样，需要专业人员的经验，因此，诊断TMD时很可能会出现人为错误。因此，本研究的目的是开发一种使用红外摄像头的方法，该方法利用下颌的最大运动范围和四种类型的分类器来帮助专业人员以定量方式对颞下颌关节（TMJ）和相关肌肉的病理学进行分类，从而有助于诊断和跟踪TMD。方法采用颞下颌疾病诊断标准（DC / TMD）对40名患者进行评估和诊断，将其分为三组：20名健康人（对照组CG），10名肌病患者（MG），10名关节病患者（AG） ）。通过运动捕捉进行定量评估。TMJ运动是通过安装在每个人的面部和颌骨上的摄像机跟踪标记来捕获的。使用定制软件导出和分析数据。使用K最近邻（KNN），Random Forest，朴素贝叶斯（NaïveBayes）和支持向量机（Support Vector Machine）算法，将数据用于识别每个参与者并将其置于三个类别之一。结果在对三组进行分类时，KNN达到了显着的准确性和准确性（超过90％）。测试的其他方法显示出较低的灵敏度和特异性值。结论本文提出的定量TMD分类方法具有优于DC / TMD标准的精确度和准确性。但是，不应将其用作独立工具，而应作为诊断TMD的辅助方法。