Elsevier

American Journal of Orthodontics and Dentofacial Orthopedics
美国正畸与牙面畸形矫治杂志

Volume 163, Issue 2, February 2023, Pages 260-271.e5
卷 163,第 2 期,2023 年 2 月,页 260-271.e5
American Journal of Orthodontics and Dentofacial Orthopedics

Digital orthodontics 数字正畸
Machine-learning–based detection of degenerative temporomandibular joint diseases using lateral cephalograms
基于机器学习的侧位头颅片退行性颞下颌关节疾病检测

https://doi.org/10.1016/j.ajodo.2022.10.015Get rights and content 获取权利和内容

Highlights 亮点

  • Cephalograms can contribute to the screening of degenerative temporomandibular joint disease.
    颅侧影可以帮助筛查退行性颞下颌关节疾病。

  • The history of limited mouth opening and crepitus is correlated with DJD.
    有限张口史和关节弹响与骨关节病相关。

  • Nomogram combined Ceph scores and clinical factors performed well in DJD screening.
    名义图结合颅骨评分和临床因素在退行性关节病筛查中表现良好。

Introduction 介绍

Degenerative temporomandibular joint diseases (DJDs) are common diseases in dental practice, characterized by a series of degenerative processes in the temporomandibular joint. Early clinical detection of DJD by dental practitioners can be beneficial to prevent or alleviate the further progression of the disease. This study aimed to develop a cephalogram-based multidimensional nomogram to screen DJD.
退行性颞下颌关节疾病(DJDs)是牙科常见疾病,特征是颞下颌关节的一系列退行性过程。牙科医生对 DJD 的早期临床检测可以有助于预防或缓解疾病的进一步发展。本研究旨在开发一种基于头影测量的多维标记图,用于筛查 DJD。

Methods 方法

A total of 502 patients (170 normal and 332 with DJD) were randomly assigned to a training set (n = 351) or a validation set (n = 151). Thirty-six cephalometric parameters were extracted from the cephalograms to be used as input for a predictive machine-learning algorithm. Multivariable logistic regression was used to construct a combined model for visualization in the form of a nomogram. Receiver operating characteristic curve, calibration testing, and decision curve analyses were conducted to evaluate the performance of the combined model.
共 502 名患者(170 名正常患者和 332 名患有 DJD 的患者)被随机分配到训练集(n = 351)或验证集(n = 151)。从头颅影像中提取了 36 个头颅测量参数,以作为预测机器学习算法的输入。使用多变量逻辑回归构建一个结合模型,以便以诺莫图的形式进行可视化。进行了受试者工作特征曲线、校准测试和决策曲线分析,以评估结合模型的性能。

Results 结果

A Ceph score consisting of 22 cephalometric parameters were significantly associated with DJD (P <0.01). A combined model that consisted of Ceph scores and clinical features (including age, gender, limited mouth opening, crepitus, etc.) performed well in the receiver operating characteristic curve (area under the curve, 0.893), calibration test, and decision curve analyses, indicating its potential clinical value.
一个由 22 个头影测量参数组成的 Ceph 评分与 DJD 显著相关(P <0.01)。一个结合了 Ceph 评分和临床特征(包括年龄、性别、开口受限、咔嗒声等)的综合模型在受试者工作特征曲线(曲线下面积,0.893)、校准测试和决策曲线分析中表现良好,显示其潜在的临床价值。

Conclusions 结论

This study constructed and verified a multidimensional nomogram consisting of Ceph scores and clinical features, which may contribute to the clinical screening of DJD in dental practice. Future studies are needed to test the reliability of the model with similar parameters.
本研究构建并验证了包含 Ceph 评分和临床特征的多维列线图,这可能有助于牙科实践中 DJD 的临床筛查。未来需要进行研究以测试该模型在类似参数下的可靠性。

Degenerative temporomandibular joint diseases (DJDs) are common pathologic conditions affecting the temporomandibular joint (TMJ) with a prevalence of 11%-20% among temporomandibular disorder (TMD) patients.1, 2, 3 DJD is characterized by a series of degenerative processes in the TMJ, including joint cartilage loss, osteoproliferative body formation, and subchondral bone remodeling and hardening. DJD is often accompanied by pain, dysfunction, joint noises, and compromised oral health–related quality of life.4,5 DJD with joint pain is classified as temporomandibular joint osteoarthritis (TMJOA), whereas DJD without joint pain is classified as osteoarthrosis.2 According to a previous study, 35% of the population with minimal condyle and/or eminence flattening show no symptoms.6 They may experience various morphologic and functional deformities when the breakdown of the TMJ begins.7 Thus, screening for DJD in the dental clinic during early, more treatable stages is challenging and important.
退行性颞下颌关节疾病 (DJDs) 是影响颞下颌关节 (TMJ) 的常见病理状况,在颞下颌关节紊乱 (TMD) 患者中患病率为 11%-20%。1, 2, 3 DJD 的特点是 TMJ 中的一系列退行性过程,包括关节软骨丢失、骨增生体形成以及软骨下骨重塑和硬化。DJD 通常伴有疼痛、功能障碍、关节噪音以及受损的口腔健康相关生活质量。45 有关节疼痛的 DJD 被归类为颞下颌关节骨关节炎 (TMJOA),而没有关节疼痛的 DJD 被归类为骨关节病。2 根据之前的一项研究,35% 的髁突和/或隆起轻微扁平的患者没有症状。6 当 TMJ 开始分解时,他们可能会经历各种形态和功能上的畸形。7 因此,在早期、更易治疗的阶段在牙科诊所筛查 DJD 具有挑战性且重要。

The diagnosis of DJD is typically made on the basis of a radiographic examination of the condyle and articular eminence.4 TMJ with erosive resorption, attrition, sclerosis, cyst-like changes, and osteophyte formation is classified as DJD.4 Orthopantomography (OPG), transcranial oblique lateral radiography, cone-beam computed tomography (CBCT), and magnetic resonance imaging are often used for the detection of DJD.8, 9, 10 Among them, CBCT is recognized for its advantages in an accurate, detailed presentation of the TMJ and can be used to demonstrate the degree and location of TMJ damage.11 However, CBCT is not commonly performed in clinical screening for DJD because of its high radiation dosage and cost. Generally, OPG is the most typical method for screening of the maxillofacial region.12 However, this method of DJD diagnosis is low in accuracy and sensitivity because of image distortion and overlap.12
DJD 的诊断通常基于对髁突和关节隆起进行的放射学检查。4 具有侵蚀性吸收、磨损、硬化、囊肿样改变和骨赘形成的 TMJ 被归类为 DJD。4 全景 X 线片 (OPG)、经颅斜位侧位 X 线片、锥形束计算机断层扫描 (CBCT) 和磁共振成像常用于检测 DJD。8, 9, 10 其中,CBCT 因其在准确、详细地呈现 TMJ 方面的优势而得到认可,可用于展示 TMJ 损伤的程度和位置。11 然而,由于 CBCT 的高辐射剂量和成本,它在 DJD 的临床筛查中并不常用。通常,OPG 是颌面部区域筛查最典型的技术。12 然而,由于图像失真和重叠,这种 DJD 诊断方法的准确性和敏感性较低。12

Lateral cephalograms are another plain radiograph commonly obtained in the dental clinic. They were originally used in diagnosis and evaluation in orthodontic practice and are widely used for evaluating craniomaxillofacial morphology.13, 14, 15 Lateral cephalograms are not used for imaging the TMJ because they are a 2-dimensional modality. Several investigations have reported that patients with DJD show special cephalometric features compared with normal patients.13,16 Patients with severe DJD have a short ramus height and backward-rotated mandible observable on cephalograms.17 Therefore, the use of lateral cephalograms can assist in DJD diagnosis. In addition, the standard landmarks used in cephalometric tracing make up for the problem of morphologic deviations in OPG application. Analyzing and summarizing cephalometric parameters strongly related to DJD would be helpful for clinical diagnosis.
侧位头颅 X 线片是牙科诊所常用的另一种普通 X 线片。它们最初用于正畸实践中的诊断和评估,并被广泛用于评估颅面形态。13, 14, 15 侧位头颅 X 线片不用于颞下颌关节成像,因为它们是二维模态。一些研究表明,与正常患者相比,DJD 患者表现出特殊的头影测量特征。1316 重度 DJD 患者的头颅 X 线片显示下颌支高度短,下颌后旋。17 因此,使用侧位头颅 X 线片可以帮助诊断 DJD。此外,头影测量描记中使用的标准标志弥补了 OPG 应用中形态偏差的问题。分析和总结与 DJD 密切相关的头影测量参数将有助于临床诊断。

Considering the high prevalence of degenerative alteration in TMD patients and the lack of simple and sensitive clinical diagnostic features,18 an accurate tool for primary diagnosis of DJD would be helpful for clinical practice. As an emerging field attracting numerous researchers’ interest, radiomic feature-based artificial intelligence (AI) models are widely used in diagnostics, decision-making, and outcome prediction in clinics.19, 20, 21, 22 Several investigators have constructed AI-based models for the detection of TMJOA using CBCT and OPG.19, 20, 21,23, 24, 25 However, the limited access to CBCT and variability of panoramic features among different devices restricts the application of these models. Moreover, their trained deep-learning models cannot explain and analyze the results.26 Following the “as low as reasonably achievable” principle with respect to radiation exposure, it is unethical to regularly perform CBCT on every patient for the screening of DJD. As most patients seeking orthodontic treatment are already subject to cephalograms, a cephalometric parameter-based DJD detection procedure may assist in the primary diagnosis of DJD without requiring additional radiation exposure. Therefore, we sought to build and verify a cephalometric parameter-based predictive machine-learning algorithm to assist in the clinical screening of DJD.
考虑到 TMD 患者中退行性变化的高发病率和简单而敏感的临床诊断特征的缺乏,一个用于 DJD 初步诊断的准确工具将对临床实践有帮助。作为一项吸引众多研究者兴趣的新兴领域,基于放射组学特征的人工智能(AI)模型在临床诊断、决策和结果预测中得到广泛应用。一些研究者已经构建了基于 AI 的模型,用于使用 CBCT 和 OPG 检测 TMJOA。然而,CBCT 的有限获取和不同设备之间全景特征的差异限制了这些模型的应用。此外,他们训练的深度学习模型无法解释和分析结果。遵循“合理可达到的最低限度”的辐射暴露原则,定期对每位患者进行 CBCT 筛查 DJD 是没有伦理的。 由于大多数寻求正畸治疗的患者已接受头部影像测量,因此基于头部测量参数的 DJD 检测程序可以在不需要额外辐射暴露的情况下,辅助初步诊断 DJD。因此,我们旨在构建并验证一种基于头部测量参数的预测机器学习算法,以帮助临床筛查 DJD。

Material and methods 材料和方法

This retrospective study was approved by the Ethics Committee of West China Hospital of Stomatology Sichuan University (WCHSIRB-2020-418). All patients and their legal guardians were informed of the possibility that their records might be used for research purposes, and oral informed consent was obtained. Patients from whom both CBCT of the TMJs and a lateral cephalogram were obtained during the period from 2018 to 2021 were randomly evaluated. Only the first scan accompanied by CBCT was included for training the AI model for patients with ≥1 lateral cephalogram. All the patients’ personal information was deidentified.
本回顾性研究获得了四川大学华西口腔医院伦理委员会的批准(WCHSIRB-2020-418)。所有患者及其法定监护人均被告知其记录可能用于研究目的,并获得了口头知情同意。在 2018 年至 2021 年期间,随机评估了同时获得颞下颌关节 CBCT 和侧面头影测量的患者。仅将第一次伴随 CBCT 的扫描纳入训练 AI 模型的患者,要求至少有 1 个侧面头影测量。所有患者的个人信息均已去标识化。

The inclusion criteria included (1) patients aged ≥18 years, (2) patients with TMJ-related clinical information, and (3) CBCT conducted within 1 week before/after lateral cephalogram. The exclusion criteria included (1) patients with cephalogram or CBCT from which features could not be successfully extracted; (2) patients with tumor or maxillofacial deformity that could cause joint deformity; (3) patients with a history of orthodontic treatment, plastic surgery, or other craniofacial surgery; and (4) patients with a history of TMJ treatment. Ultimately, 502 patients were included in this study. Among them, 351 patients (59 male and 292 female), with an average age of 32.29 ± 10.90 years (age range, 18-69 years), were randomly assigned to training set. The other 151 patients (29 male and 122 female), with an average age of 30.14 ± 9.51 years (age range, 18-64 years), constituted the validation set. The TMD clinical examination, including evaluation of limited mouth opening, deviation, clicking, and crepitus, was performed by the 2 TMJ specialists (with 8 and 10 years of experience, respectively) as well according to DC/TMD criteria. Any disagreement was resolved through consultation with a third specialist with 20 years of experience diagnosing and treating TMJ disorders. In addition, self-reported symptoms were extracted from the medical records. A flow diagram of this study is presented in Figure 1.
纳入标准包括:(1)年龄≥18 岁的患者;(2)具有颞下颌关节相关临床信息的患者;(3)在侧位头颅片拍摄前/后 1 周内进行的 CBCT。排除标准包括:(1)头颅片或 CBCT 无法成功提取特征的患者;(2)肿瘤或颌面畸形导致关节畸形的患者;(3)有正畸治疗、整形手术或其他颅面手术史的患者;(4)有颞下颌关节治疗史的患者。最终,502 名患者被纳入本研究。其中,351 名患者(男性 59 名,女性 292 名),平均年龄 32.29±10.90 岁(年龄范围 18-69 岁),被随机分配到训练集。另外 151 名患者(男性 29 名,女性 122 名),平均年龄 30.14±9.51 岁(年龄范围 18-64 岁),构成验证集。TMD 临床检查,包括评估张口受限、偏斜、弹响和摩擦音,由 2 名颞下颌关节专科医生(分别具有 8 年和 10 年经验)根据 DC/TMD 标准进行。 任何分歧通过与一位拥有 20 年颞下颌关节疾病诊断和治疗经验的第三位专家进行咨询来解决。此外,自我报告的症状从病历中提取。该研究的流程图如图 1 所示。

  1. Download: Download high-res image (1MB)
    下载:下载高清图片(1MB)
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Fig 1. Flowchart depicting the design of the study.
图 1. 研究设计流程图。

CBCT images were obtained from all patients using the 3D Accuitomo scanner (J Morita Corp, Kyoto, Japan). Scan settings were as follows: 90 kVp, 5 mA, exposure time of 17.5 seconds, voxel size of 0.25 mm, slice thickness of 0.25 mm, a field of view of 140 mm × 100 mm, and 360° rotation. The criteria for CBCT diagnosis of DJD were developed with reference to previous literature: (1) normal-sized condyle with osteosclerosis and/or joint flattening; (2) deformed condyle with a subcutaneous cyst, cortical resorption, or extensive osteosclerosis.27 Patients with unilateral or bilateral TMJ osteoarthritis were designated as DJD (Supplementary Fig 1). Two TMJ specialists (with 4 and 5 years of experience, respectively) were assigned to evaluate TMJ independently using the CBCT without access to the patients’ cephalograms. Any disagreement was resolved through consultation with a third specialist with 20 years of experience diagnosing and treating TMJ disorders.
所有患者均使用 3D Accuitomo 扫描仪(J Morita 公司,日本京都)获取 CBCT 图像。扫描设置如下:90 kVp,5 mA,曝光时间 17.5 秒,体素大小 0.25 mm,切片厚度 0.25 mm,视野 140 mm × 100 mm,360°旋转。CBCT 诊断 DJD 的标准参考了以往文献:(1)髁突大小正常,伴有骨硬化和/或关节平坦;(2)髁突变形,伴有皮下囊肿、皮质吸收或广泛骨硬化。27 例单侧或双侧颞下颌关节骨关节炎患者被指定为 DJD(补充图 1)。两位颞下颌关节专家(分别具有 4 年和 5 年经验)被分配独立评估颞下颌关节,不接触患者的头颅侧位片。任何意见分歧都通过与第三位具有 20 年颞下颌关节疾病诊断和治疗经验的专家协商解决。

Cephalometric parameter extraction was performed on lateral cephalograms, which were acquired using the same radiographic machine. Cephalometric tracing was conducted with the software Uceph (version 4.4.2; Yacent, China) by 2 orthodontists (X.F. and X.X.) with 5 years of experience each (Supplementary Fig 2). The orthodontists were blinded to the clinical information and diagnosis of all patients. A total of 36 cephalometric parameters were digitized and divided into 7 categories: cranial base relationship (3 parameters), size and position of the maxilla (2 parameters), size and position of the mandible (6 parameters), the relationship between maxilla and mandible (2 parameters), vertical dimension (11 parameters), dentoalveolar features (10 parameters) and facial profile (2 parameters). The reference plane was the Frankfort horizontal plane. Detailed values of these cephalometric parameters in the normal and DJD groups are listed in Supplementary Table I. To assess the reproducibility of the cephalometric parameters, 2 orthodontists traced 50 randomly selected cephalograms and repeated these tracings after 30 days. The interclass and intraclass correlation coefficients (ICCs) and Bland-Altman analysis were computed to evaluate the intraobserver and interobserver reproducibility of cephalometric parameters.
侧位头颅 X 线片采用同一台放射设备获取,并对其进行头影测量参数提取。头影测量描记由两位具有 5 年经验的正畸医生(XF 和 XX)使用 Uceph 软件(版本 4.4.2;雅森,中国)进行(补充图 2)。正畸医生对所有患者的临床信息和诊断均不知情。共数字化了 36 个头影测量参数,并将其分为 7 类:颅底关系(3 个参数)、上颌骨大小和位置(2 个参数)、下颌骨大小和位置(6 个参数)、上颌骨和下颌骨之间的关系(2 个参数)、垂直维度(11 个参数)、牙槽特征(10 个参数)和面部轮廓(2 个参数)。参考平面为法兰克福水平面。正常组和 DJD 组这些头影测量参数的详细数值列于补充表 I 中。为了评估头影测量参数的重复性,两位正畸医生对 50 张随机选择的头颅 X 线片进行了描记,并在 30 天后重复描记。 组间和组内相关系数 (ICC) 和 Bland-Altman 分析被用来评估头影测量参数的观察者内和观察者间重复性。

The least absolute shrinkage and selection operator (LASSO) logistic regression algorithm was applied to choose the most effective and reproducible DJD-related cephalometric parameters out of the 36 listed cephalometric parameters.28 Cephalometric parameters with nonzero coefficients in the LASSO regression were selected to generate a risk score (referred to as a Ceph score) in the training set. The Ceph score was then used to predict the TMJ status in both the training and validation sets using a Mann-Whitney U test. The predictive accuracy of the Ceph score was evaluated by the receiver operating characteristic (ROC) test and area under the curve (AUC) in both sets.
最小绝对收缩和选择算子(LASSO)逻辑回归算法被应用于从 36 个列出的头影测量参数中选择最有效且可重复的与 DJD 相关的头影测量参数。28 在 LASSO 回归中具有非零系数的头影测量参数被选择以生成风险评分(称为 Ceph 评分),该评分在训练集中使用。然后,Ceph 评分用于通过 Mann-Whitney U 检验预测训练集和验证集中的 TMJ 状态。通过接收器操作特征(ROC)测试和曲线下面积(AUC)评估 Ceph 评分在两个集合中的预测准确性。

Multivariable logistic regression analysis was conducted on the training set to create a clinical model with the following factors: age, gender, and the experience of limited mouth opening, deviation, clicking, crepitus, joint pain, and orofacial pain.
对训练集进行了多变量逻辑回归分析,以创建一个临床模型,考虑以下因素:年龄、性别,以及有限开口、偏移、点击声、摩擦声、关节疼痛和面部口腔疼痛的经历。

Ceph scores were combined with the clinical model above to develop a diagnostic model for DJD. A combined model with all features was constructed on the basis of a multivariate logistic regression model. A nomogram was generated on the basis of the combined model in the training set using the R package rms as a quantitative tool to predict the individual probability of DJD.
Ceph 评分与上述临床模型相结合,开发了 DJD 的诊断模型。基于多元 Logistic 回归模型构建了包含所有特征的组合模型。使用 R 包 rms 在训练集上基于组合模型生成列线图,作为定量工具预测 DJD 的个体概率。

We compared the discriminatory performance of the 3 established models (clinical model, Ceph score, and combined model) using ROC curves and AUC values in both the training and validation sets. Then, calibration curves and the Hosmer-Lemeshow test were also employed to evaluate the calibration of the combined model. To further evaluate the efficiency of the combined model for different demographic characteristics, a stratified analysis was conducted. The performance of the combined model was evaluated by the AUC values in different subgroups, including gender (male or female), sagittal dimension (skeletal Class I, skeletal Class II, or skeletal Class III), and vertical dimension (low angle, average angle, or high angle). The calibration curves were plotted using R package rms.
我们使用 ROC 曲线和 AUC 值在训练集和验证集中比较了 3 种已建立模型(临床模型、Ceph 评分和组合模型)的判别性能。然后,还使用校准曲线和 Hosmer-Lemeshow 检验来评估组合模型的校准。为了进一步评估组合模型对不同人口特征的效率,进行了分层分析。组合模型的性能通过不同亚组的 AUC 值进行评估,包括性别(男性或女性)、矢状尺寸(骨骼 I 类、骨骼 II 类或骨骼 III 类)和垂直尺寸(低角度、平均角度或高角度)。校准曲线使用 R 包 rms 绘制。

To evaluate the clinical application potential of our combined model, we conducted a decision curve analysis (DCA) to further assess the net benefit acquired from the deployment of the clinical model and the combined model. The performance of these 2 models was analyzed at different threshold probabilities, and the model with the larger region under the curve was chosen for better clinical outcomes.29
为了评估我们组合模型的临床应用潜力,我们进行了决策曲线分析(DCA),以进一步评估部署临床模型和组合模型获得的净收益。分析了这两个模型在不同阈值概率下的性能,并选择曲线下面积更大的模型以获得更好的临床结果。29

Statistical analysis 重试    错误原因

Normally distributed variables were compared using Student’s t test. Continuous variables that were not normally distributed were compared using the Mann-Whitney U test. The discrete variables were analyzed using the chi-square test. All the statistical analyses used in this study were performed with SPSS (version 20; IBM, Armonk, NY), R (4.1.0; R Foundation for Statistical Computing, Vienna, Austria), and EmpowerStats 2.2.0.11 (X&Y solutions, Inc, Boston, Mass). P <0.05 was considered statistically significant.
正态分布变量通过学生 t 检验进行比较。非正态分布的连续变量通过曼-惠特尼 U 检验进行比较。离散变量使用卡方检验进行分析。本研究中使用的所有统计分析均在 SPSS(版本 20;IBM,阿蒙克,纽约)、R(4.1.0;维也纳,奥地利,R 统计计算基金会)和 EmpowerStats 2.2.0.11(X&Y 解决方案公司,波士顿,马萨诸塞州)中进行。P <0.05 被认为具有统计学意义。

Results 结果

The study workflow is presented in Figure 1. A total of 502 patients were included in this study and were randomly assigned to the training and testing groups at a ratio of 7:3. Clinical characteristics and cephalometric measurement values of included patients are presented in Table I and Supplementary Table II. The intraobserver ICCs and the interobserver ICCs were both >0.75. Bland-Altman plots of ANB, FH-PP, and S-N are shown in Supplementary Figure 3. Bias in the Bland-Altman analysis ranged from 0.08 to 0.61. All P values in the Bland-Altman analysis were >0.05, representing acceptable intraobserver and interobserver reproducibility of cephalometric parameters. No significant difference was found in any clinical feature between the training and the validation sets, which indicated excellent equivalency between the 2 sets. Significant differences were found in gender and experience of limited mouth opening and crepitus between the normal and DJD groups in both the training and test sets (Table I).
研究工作流程如图 1 所示。本研究共纳入 502 名患者,按 7:3 的比例随机分配到训练组和测试组。纳入患者的临床特征和头影测量值见表 I 和补充表 II。观察者间信度(ICC)和观察者内信度(ICC)均>0.75。ANB、FH-PP 和 S-N 的 Bland-Altman 图示于补充图 3 中。Bland-Altman 分析中的偏倚范围为 0.08 到 0.61。Bland-Altman 分析中的所有 P 值均>0.05,表示头影测量参数具备可接受的观察者内和观察者间重复性。在训练集和验证集中,未发现任何临床特征的显著差异,表明两组之间具有良好的等效性。在训练集和测试集中,正常组与 DJD 组在性别及有限开口经验与关节摩擦声方面存在显著差异(表 I)。

Table I. Characteristics of patients in the training and validation set
表 I. 训练集和验证集患者特征

Characteristics 特征Training set (n = 351) 训练集 (n = 351)Test set (n = 151) 测试集 (n = 151)
DJD (n = 228) DJD (n=228)Normal (n = 123) 正常 (n = 123)P values P 值DJD (n = 104) DJD (n=104)Normal (n = 47) 正常 (n = 47)P values P 值
Age, mean ± standard deviation
年龄,均值 ± 标准差
31.81 ± 10.4333.17 ± 11.750.18930.33 ± 9.5829.74 ± 9.560.803
Gender, n (%) 性别,n (%)
 Male 男性28 (12.3)31 (25.2)0.00225 (24.0)4 (8.5)0.018
 Female 女性200 (87.7)92 (74.8)79 (76.0)43 (91.5)
Limited mouth opening, n (%)
张口受限,n (%)
 Yes 88 (38.6)29 (23.6)0.00334 (32.7)8 (17.0)0.034
 No 140 (61.4)94 (76.4)70 (67.3)39 (83.0)
Deviation, n (%) 重试    错误原因
 Yes 104 (45.6)60 (48.8)0.32450 (48.1)29 (61.7)0.084
 No 重试    错误原因 124 (54.4)63 (51.2)54 (51.9)18 (38.3)
Clicking, n (%) 点击,n (%)
 Yes 171 (75.0)82 (66.7)0.06384 (80.8)34 (72.3)0.171
 No 重试    错误原因 57 (25.0)41 (33.3)20 (19.2)13 (27.7)
Crepitus, n (%) 气泡声, n (%)
 Yes 86 (37.7)16 (13.0)<0.000133 (31.7)8 (17.0)0.043
 No 142 (62.3)107 (87.0)71 (68.3)39 (83.0)
Pain, n (%) 疼痛, n (%)
 Yes 188 (82.5)101 (82.1)0.52387 (83.7)37 (78.7)0.303
 No 重试    错误原因 40 (17.5)22 (17.9)17 (16.3)10 (21.3)
Ceph score, median (interquartile range) 重试    错误原因 68.62 (67.86-69.32)66.76 (65.98-67.60)<0.000168.14 (67.17-69.15)66.74 (66.11-67.25)<0.0001

A total of 36 cephalometric parameters were used for the LASSO analysis. LASSO regression was employed to choose the optimized cephalometric parameters for the cephalometrics model. In this manner, 22 cephalometric parameters with nonzero coefficients were selected to construct the cephalometrics signature (Fig 2, A-C), which consisted of 7 vertical dimension parameters, 5 dentoalveolar parameters, 4 mandibular skeletal parameters, 3 cranial base parameters, 1 facial profile parameter, 1 maxillary skeletal parameter, and 1 maxillary/mandibular parameter. Using all these cephalometric parameters, a formula for calculating a risk score (Ceph score) was established on the basis of a multi-logistic regression, as presented in the Supplementary Material.
共使用了 36 个头影测量参数进行 LASSO 分析。采用 LASSO 回归选择优化的头影测量参数,以构建头影测量模型。通过这种方式,选择了 22 个具有非零系数的头影测量参数用于构建头影测量特征(图 2,A-C),其中包括 7 个垂直维度参数、5 个牙槽参数、4 个下颌骨骼参数、3 个颅底参数、1 个面部轮廓参数、1 个上颌骨骼参数和 1 个上颌/下颌参数。利用所有这些头影测量参数,基于多重逻辑回归建立了计算风险评分(Ceph 评分)公式,如补充材料所示。

  1. Download: Download high-res image (826KB)
    下载:下载高分辨率图像 (826KB)
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Fig 2. Cephalometrics feature selection using the LASSO binary logistic regression model: A, Tuning parameter (λ) selection in the LASSO model used 10-fold cross-validation via minimum criteria. Dotted vertical lines were drawn at the optimal values using the minimum criteria and the λ standard error of the minimum criteria (1-standard error [SE] criteria). The optimal λ value of 0.0027 with log (λ) = −5.8992 was selected (1-SE criteria) according to 10-fold cross-validation; B, LASSO coefficient profiles of the 36 cephalometrics features. The dotted line was plotted at the value selected using 10-fold cross-validation in A, in which optimal λ resulted in 22 nonzero coefficients; C, 22 cephalometric features were selected for the signature building.
图 2. 使用 LASSO 二元逻辑回归模型的头影测量特征选择:A,LASSO 模型中调节参数 (λ) 的选择使用通过最小标准的 10 倍交叉验证。根据最小标准和最小标准的 λ 标准误差(1 标准误差 [SE] 标准),在最佳值处绘制了虚线垂直线。根据 10 倍交叉验证,选择的最佳 λ 值为 0.0027,log (λ) = -5.8992(1-SE 标准);B,36 个头影测量特征的 LASSO 系数轮廓。虚线是根据 A 中 10 倍交叉验证选择的值绘制的,最佳 λ 导致 22 个非零系数;C,选择了 22 个头影测量特征用于特征构建。

The Ceph score of each patient was calculated. A significant difference was detected between patients with and without DJD in the training set (P <0.001, Fig 3, A), which was verified in the validation set (P <0.001, Fig 3, C). The AUC value of Ceph scores was 0.861 (95% confidence interval (CI), 0. 822-0.900, Fig 3, B) in the training set and 0.812 (95% CI, 0.738-0.886; Fig 3, D) in the validation set, which demonstrated that Ceph scores have a good discriminatory ability.
每个患者的 Ceph 评分均已计算。在训练集中,患有和未患 DJD 的患者之间存在显著差异(P <0.001,图 3,A),并在验证集中得到验证(P <0.001,图 3,C)。Ceph 评分的 AUC 值为 0.861(95%置信区间(CI),0.822-0.900,图 3,B)在训练集中和 0.812(95%置信区间,0.738-0.886;图 3,D)在验证集中,这表明 Ceph 评分具有良好的鉴别能力。

  1. Download: Download high-res image (373KB)
    下载:下载高分辨率图像(373KB)
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Fig 3. The ROC curves of the Ceph score in the (A) training set and the (B) validation set. The box-dot plots of the Ceph score in the training set (A) and the validation set (C). Plots (B) and (D) show the ROC curves of the Ceph score in the training and validation sets, respectively.
图 3. Ceph 评分在(A)训练集和(B)验证集中的 ROC 曲线。训练集(A)和验证集(C)中的 Ceph 评分箱点图。图(B)和(D)分别显示了 Ceph 评分在训练集和验证集中的 ROC 曲线。

A combined model was constructed using multivariable regression analysis on the basis of traditional clinical features and Ceph scores. A detailed description of the features is shown in Table II. Ceph scores, the experience of limited mouth opening, and crepitus were all significantly correlated with DJD. A nomogram based on these features was established (Fig. 4, A). The ROC test showed that the combined model scored 0.893 (95% CI, 0.847-0.919) on the training set, 0.828 (95% CI, 0.757-0.899) on the validation set, and thus performed better than the clinical model, which scored 0.701 (95% CI, 0.645-0.756) on the training set and 0.603 (95% CI, 0.505-0.701) on the validation set (Figs 4, B and C). Significant differences were shown in AUC values between the combined model and the clinical model (P <0.001), which confirmed the combined model’s superior predictive performance. In addition, the combined model revealed a good capacity for identifying DJD in different subgroups (Table III), indicating its utility for recognizing the disease in different stratification contexts.
通过基于传统临床特征和颅面评分(Ceph scores)的多变量回归分析构建了一个联合模型。特征的详细描述见表 II。颅面评分、张口受限的经验和关节摩擦音与退行性关节病(DJD)均显著相关。基于这些特征建立了一个列线图(图 4,A)。ROC 检验显示,联合模型在训练集上的得分为 0.893(95% CI, 0.847-0.919),在验证集上的得分为 0.828(95% CI, 0.757-0.899),因此表现优于临床模型,后者在训练集上的得分为 0.701(95% CI, 0.645-0.756),在验证集上的得分为 0.603(95% CI, 0.505-0.701)(图 4,B 和 C)。联合模型与临床模型之间的 AUC 值显示出显著差异(P <0.001),这证实了联合模型的优越预测性能。此外,联合模型在不同亚组中识别 DJD 的能力良好(表 III),表明其在不同分层背景下识别疾病的实用性。

Table II. Risk factor for DJD
表 II. DJD 的风险因素

Variable 变量Combined model (95% CI) 联合模型 (95% CI)Clinical model (95% CI) 临床模型 (95% CI)
Odds ratio 赔率比P values P 值DJDP values P 值
Age 重试    错误原因 0.9781(0.9520-1.0049)0.10850.9847 (0.9636-1.0064)0.1656
Gender 重试    错误原因 0.8911(0.4149-1.9139)0.76760.4451 (0.2418-0.8193)0.0093
Limited mouth opening 口腔张开受限2.1320(1.0890-4.1739)0.02721.8815 (1.1117-3.1842)0.0185
Deviation 偏差0.9991(0.5459-1.8287)0.99770.7347 (0.4558-1.1844)0.2058
Clicking 点击1.4780 (0.7591-2.8777)0.25041.4814 (0.8750-2.5080)0.1435
Crepitus 捻发音4.2697 (2.0247-9.0040)<0.00014.0898 (2.2162-7.5471)<0.0001
Pain 疼痛1.2900 (0.5956-2.7940)0.51840.7766 (0.4122-1.4630)0.4339
Ceph score 头颅评分3.5658 (2.6574-4.7848)<0.0001NANA

NA, not available. 不可用。

  1. Download: Download high-res image (433KB)
    下载:下载高分辨率图像 (433KB)
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Fig 4. Combined model for the prediction of DJD: A, The nomogram was developed in the training set, with the Ceph score and clinical features. Plots (B) and (C) show the ROC curves for the combined model, Ceph score, and clinical model in the training and validation sets, respectively.
图 4. DJD 预测的综合模型:A,列线图是在训练集中开发的,包含 Ceph 评分和临床特征。图(B)和(C)分别显示了训练集和验证集的综合模型、Ceph 评分和临床模型的 ROC 曲线。

Table III. The AUC values of the combined model for stratified analysis in different subgroups
表 III. 不同亚组分层分析的联合模型 AUC 值

Subgroup 亚组Patients 患者AUC values (95% CI) 重试    错误原因
Gender 性别
 Male (n = 88) 男性 (n = 88)DJD (n = 53)0.812 (0.717-0.907)
Normal (n = 35) 正常 (n = 35)
 Female (n = 414) 重试    错误原因 DJD (n = 279) DJD (n = 279)0.877 (0.841-0.912)
Normal (n = 135) 正常 (n = 135)
Sagittal dimension 矢状维度
 Skeletal Class I (n = 256)
骨骼 I 型(n = 256)
DJD (n = 148) 重试    错误原因 0.816 (0.763-0.869)
Normal (n = 108) 正常 (n = 108)
 Skeletal Class II (n =196)
骨骼 II 类 (n =196)
DJD (n = 160) 退行性关节病 (n = 160)0.885 (0.820-0.949)
Normal (n = 36) 正常 (n = 36)
 Skeletal Class III (n = 50)
骨骼 III 类 (n = 50)
DJD (n = 24) DJD (n = 24)0.899 (0.806-0.992)
Normal (n = 26) 正常(n = 26)
Vertical dimension 垂直维度
 Low angle (n = 167) 重试    错误原因 DJD (n = 98) DJD (n=98)0.847 (0.786-0.909)
Normal (n = 69) 正常 (n = 69)
 Average angle (n = 264) 重试    错误原因 DJD (n = 174) DJD (n = 174)0.855 (0.807-0.902)
Normal (n = 90) 正常 (n = 90)
 High angle (n = 71)
高角度 (n = 71)
DJD (n = 60) 退行性关节病 (n = 60)0.949 (0.882-1.000)
Normal (n = 11) 正常 (n = 11)

The calibration curves of the combined model between predicted and actual DJD showed good consistency in both the training and validation sets (Figs 5, A and B). According to the Hosmer-Lemeshow test, nonsignificant P values were found in both the training set (P = 0.863) and validation set (P = 0.881), demonstrating that the combined model achieved acceptable goodness of fit.
综合模型的校准曲线在预测和实际 DJD 之间在训练集和验证集中都显示出良好的一致性(图 5,A 和 B)。根据 Hosmer-Lemeshow 检验,训练集(P = 0.863)和验证集(P = 0.881)均发现了不显著的 P 值,表明综合模型达到了可接受的拟合优度。

  1. Download: Download high-res image (197KB)
    下载:下载高分辨率图像 (197KB)
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Fig 5. The calibration curves in the training set (A) and validation set (B) represent the predictive performance of the combined model. Gray represents a perfect prediction, and pink represents the predictive performance of the combined model.
图 5. 训练集 (A) 和验证集 (B) 中的校准曲线代表组合模型的预测性能。灰色代表完美预测,粉红色代表组合模型的预测性能。

The DCA for the combined model and the traditional clinical model is presented in Figure 6. The decision curve demonstrated that the combined model would be more beneficial than the “treat all patients” strategy or the “treat none” strategy when the threshold probability was >0.312.
组合模型和传统临床模型的 DCA 见图 6。决策曲线表明,当阈值概率>0.312 时,组合模型比“治疗所有患者”策略或“不治疗”策略更有益。

  1. Download: Download high-res image (162KB) 重试    错误原因
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Fig 6. DCA for the combined model compared with the clinical model. The y-axis represents the standardized net benefit. The x-axis represents the threshold probability. Gray represents the possibility that all patients had DJD. Black represents the possibility that no patients had DJD. The decision curves in the validation set showed that using the combined model to predict DJD adds more benefit than treating all or no patients when the threshold probability is >0.312.
图 6. 综合模型与临床模型的决策曲线分析(DCA)。y 轴代表标准化净收益。x 轴代表阈值概率。灰色表示所有患者都有退行性关节病(DJD)的可能性。黑色表示没有患者有退行性关节病(DJD)的可能性。在验证集中,决策曲线显示,当阈值概率 >0.312 时,使用综合模型预测退行性关节病(DJD)相比于对所有患者或无患者进行治疗具有更多益处。

Discussion 讨论

In this study, we built and validated a combined model consisting of a cephalometric parameter-based Ceph score and traditional clinical features, which can assist in the clinical screening of DJD. This combined model and its nomogram displayed superb resolution ability on the training and validation sets in the ROC test. The AUC values of the Ceph score (0.861 in the training set, 0.812 in the validation set) and the combined model (0.893 in the training set, 0.828 in the validation set) were significantly higher than those of the clinical model (0.701 in the training set, 0.603 in the validation set). Furthermore, the DCA showed that the combined model could be more effective for clinical decision-making than the clinical model. The standard and well-developed landmarks used in cephalography make it possible for this cephalometric parameter-based combined model to be widely used in diagnosing DJD.
在这项研究中,我们构建并验证了一个结合模型,该模型由基于头影测量参数的 Ceph 评分和传统临床特征组成,可以辅助 DJD 的临床筛查。这个结合模型及其列线图在 ROC 测试中表现出优越的分辨能力,在训练集和验证集上的 AUC 值分别为 Ceph 评分(训练集为 0.861,验证集为 0.812)和结合模型(训练集为 0.893,验证集为 0.828),明显高于临床模型(训练集为 0.701,验证集为 0.603)。此外,DCA 显示结合模型在临床决策中可能比临床模型更有效。头影测量中使用的标准和完善的标志点使得这个基于头影测量参数的结合模型能够在 DJD 的诊断中得到广泛应用。

We established that cephalometrics could distinguish patients with DJD from normal patients through a comprehensive analysis of cephalometric parameters, including cranial base relationship, size and position of the maxilla, size and position of the mandible, the relationship between maxilla and mandible, vertical dimension, dentoalveolar features, and facial profile. Considering potential doubts that could arise regarding the reproducibility and robustness of cephalometric parameters, we took strong precautions to ensure the objectivity and reproducibility of our cephalometric model. The diagnosis of DJD and cephalometric feature extraction used in this study have been commonly applied and verified in previous studies. Two orthodontists traced the cephalograms based on the established standards, and both ICCs and the Bland-Altman analysis were used to evaluate subjectivity and operator error. Random errors for the digital tracing were observed in the Bland-Altman analysis, which might decrease the reliability of the model. Future researchers are encouraged to test the reliability of the model with similar cephalometric parameters.
我们建立了头影测量可以通过对头影测量参数的综合分析,将退行性关节病患者与正常患者区分开来,包括颅底关系、上颌骨的大小和位置、下颌骨的大小和位置、上颌骨与下颌骨之间的关系、垂直维度、牙槽特征和面部轮廓。考虑到可能对头影测量参数的可重复性和稳健性产生的疑虑,我们采取了强有力的预防措施,以确保我们的头影测量模型的客观性和可重复性。本研究中用于诊断退行性关节病和头影测量特征提取的方法在之前的研究中得到了普遍应用和验证。两位正畸医师根据既定标准描绘头影图,同时采用了一致性相关系数(ICC)和 Bland-Altman 分析来评估主观性和操作错误。在 Bland-Altman 分析中观察到数字描绘的随机误差,这可能降低模型的可靠性。鼓励未来研究者使用类似的头影测量参数测试模型的可靠性。

All 36 cephalometric parameters from 7 categories describing the relationships between cranial and maxillofacial soft and hard tissues and occlusion were subjected to analysis to rule out the influences of scale differences. To prevent overfitting, parameters less correlated with DJD were excluded from the final calculation. Using all the above methods, a comparatively evidence-based cephalometric model was built for DJD screening in dental practice.
描述颅面软硬组织和咬合关系的 7 个类别中的所有 36 个头影测量参数均经过分析,以排除尺度差异的影响。为了防止过度拟合,与 DJD 相关性较低的参数被排除在最终计算之外。使用上述所有方法,构建了一个比较基于证据的头影测量模型,用于牙科实践中的 DJD 筛查。

In our cephalometric model, 22 cephalometric parameters with strong correlations with osseous abnormalities were extracted, most of which were vertical dimension parameters. Specifically, counterclockwise rotation of the occlusal plane and short posterior facial height were closely related to DJD. Similarly, several investigators have reported that a tendency toward mandibular counterclockwise rotation, mandibular retrognathism, and a shorter mandible are associated with TMJOA.13,16,30,31 Moreover, in the sagittal dimension, larger ANB angles were positively related to TMJOA, which is consistent with the higher prevalence of TMJOA observed in patients with Class II patients malocclusion compared with those with Class I and Class III malocclusion.32 On one hand, a hyperdivergent skeletal pattern and mandibular retrognathism could result from aggressive condylar resorption. It has been reported that the digastric muscle and the mylohyoid muscle retract the mandible downward and backward in patients with short mandibular ramus height.17,33 On the other hand, these 2 clinical features have been regarded as risk factors for TMD.34 In patients with DJD, compressive deflection loaded on the condyle is produced with masticatory muscular force during the interdigitation of teeth, which may result in compressive condyle resorption.17 Regarding mandibular skeletal parameters, we found that increased articular angle (S-AR-GO) is positively associated with DJD. A larger articular angle, reflecting a backward position of the mandible, might cause an unfavorable muscle attachment position and abnormal muscle activity, leading to tooth extrusion and subsequent counterclockwise mandibular rotation.35,36 Moreover, abnormal overbite has also been shown to be related to DJD. Although controversy remains regarding the relationship between overbite and TMD, it is undisputed that normal occlusal guidance is important to a healthy muscle-occlusal system and TMJ.37, 38, 39
在我们的头影测量模型中,提取了 22 个与骨骼异常强相关的头影测量参数,其中大部分为垂直维度参数。具体而言,咬合平面的逆时针旋转和短后面高度与 DJD 密切相关。同样,一些研究人员报告说, mandibular 逆时针旋转、下颌后缩和下颌短小的倾向与 TMJOA 相关。此外,在矢状面上,较大的 ANB 角度与 TMJOA 呈正相关,这与 II 类患者的 TMJOA 患病率高于 I 类和 III 类患者的观察结果一致。一方面,超发散的骨骼模式和下颌后缩可能是由于侵袭性髁头吸收所致。据报道,二腹肌和舌骨下肌在下颌支高较短的患者中向下和向后收缩下颌。另一方面,这两种临床特征被认为是 TMD 的风险因素。34 在患有 DJD 的患者中,牙齿嵌合期间咀嚼肌力量产生的髁突压缩偏转会导致髁突压缩性吸收。17 关于下颌骨骼参数,我们发现关节角增大 (S-AR-GO) 与 DJD 正相关。较大的关节角反映了下颌骨向后移位,可能导致肌肉附着位置不利和肌肉活动异常,从而导致牙齿突出和随后下颌骨逆时针旋转。3536 此外,异常的深覆合也与 DJD 相关。尽管关于深覆合与颞下颌关节紊乱之间关系的争议仍然存在,但正常咬合引导对健康的肌肉-咬合系统和颞下颌关节至关重要,这一点是毫无争议的。37, 38, 39

Among all clinical factors included in the combined model, the history of limited mouth opening (odds ratio, 2.132; 95% CI, 1.089-4.174; P = 0.027) and crepitus (odds ratio, 4.270; 95% CI, 2.025-9.004; P <0.0011) were significantly correlated with DJD (Table II). Limited mouth opening is a typical symptom of disc displacement without reduction. The prevalence of early-stage DJD increases from 24% to 60% 1 month after limited mouth opening occurs.40 Previous studies implied that the prevalence of crepitus is significantly higher in patients with TMJOA,41 and our results are consistent with this finding. Usually, clicking indicates disc displacement with reduction,42 whereas crepitus is more often related to erosion and the formation of subchondral cysts and osteophytes in the condyles,43 indicating a late stage of degeneration. Pain is a complicated situation that can derive from muscles, the TMJ or other orofacial areas.44 The relationship between pain and DJD is inconclusive because around one-third of patients with DJD might not suffer from pain,45,46 and conversely, patients with orofacial pain might not have osseous changes in the TMJ.43,47 Therefore, it is reasonable that pain symptoms alone exhibited limited ability in our model to predict DJD because of the ambiguity of pain-related symptoms. However, with the contribution of the Ceph score and clinical factors, the combined model achieved better diagnostic ability.
在合并模型中包含的所有临床因素中,有限开口史(优势比,2.132;95% CI,1.089-4.174;P = 0.027)和关节弹响(优势比,4.270;95% CI,2.025-9.004;P <0.0011)与 DJD 显著相关(表 II)。有限开口是无复位椎间盘移位的典型症状。有限开口发生后 1 个月,早期 DJD 的患病率从 24%增加到 60%。40 之前的研究表明,颞下颌关节骨关节炎患者的关节弹响患病率明显更高,41 我们的结果与这一发现一致。通常,弹响表示有复位椎间盘移位,42 而关节弹响则更多地与髁突的软骨下囊肿和骨赘的侵蚀和形成有关,43 这表明退变的晚期。疼痛是一个复杂的情况,可能源于肌肉、颞下颌关节或其他口面部区域。44 疼痛与 DJD 之间的关系尚无定论,因为大约三分之一的 DJD 患者可能不会出现疼痛,4546 反之,口面部疼痛患者可能不会出现颞下颌关节的骨性改变。因此,由于疼痛相关症状的模糊性,仅疼痛症状在我们的模型中预测 DJD 的能力有限是合理的。然而,随着头影测量评分和临床因素的贡献,组合模型获得了更好的诊断能力。

Compared with previous research using radiographic examination to diagnose TMJOA,19, 20, 21,23 our combined model is advantageous for several reasons. First, the clinical application potential of this combined model under various circumstances was validated, as the stratified analysis performed for further evaluation showed. In addition to cephalometric parameters, conventional clinical features were adopted in the combined model. The combination of cephalometric and clinical features offers a convenient DJD screening method without any special imaging beyond what is routinely done. There is no need for specialized equipment such as CBCT, which is not commonly available in an average dental clinic. Within limiting radiation exposure, the screening of DJD is completed. Some newly graduated or general dentists may have limited knowledge about DJD and have difficulty diagnosing TMJ status in their orthodontic practice.48,49 An algorithm based solely on cephalometric measurements in our study can be built into software or a Web site, and using this algorithm to analyze the cephalograms of patients could help clinicians to assess the joint status of patients and identify patients with a higher likelihood of developing DJD. After identifying patients with high odds of DJD, the dentist could perform a more comprehensive clinical examination and imaging of the TMJ or refer the patient to a TMJ specialist. 重试    错误原因

Several limitations of this study should be noted. First, this research was conducted in a single hospital. Although the cephalometric parameters based on standard and well-developed landmarks performed well in numerous cephalometric analyses,13,14,16 further investigations based on multicentric databases are still required to validate the accuracy and robustness of this model. Second, the cephalometric tracing was conducted by 2 orthodontists, but this might be acceptable because the landmarks were identified strictly on the basis of cephalometric analysis criteria that are accepted internationally. Moreover, the intraobserver ICCs and the interobserver ICCs confirmed good intraobserver and interobserver reproducibility of cephalometric parameters. Finally, to achieve data integrity, only the most active features were included in the predictive model, and it is possible that some other relevant cephalometric features were excluded.
本研究存在一些局限性。首先,这项研究是在一家医院进行的。尽管基于标准和完善的标志的颅面测量参数在许多颅面测量分析中表现良好,131416 但仍需要基于多中心数据库的进一步研究来验证该模型的准确性和稳健性。其次,颅面测量描记是由两位正畸医生进行的,但这可能是可以接受的,因为标志的识别严格基于国际公认的颅面测量分析标准。此外,组内 ICC 和组间 ICC 证实了颅面测量参数良好的组内和组间重复性。最后,为了实现数据完整性,预测模型中只包含了最活跃的特征,因此可能排除了其他一些相关的颅面测量特征。

Conclusions 结论

A novel cephalogram-based combined model was constructed and verified to screen for DJD. The model consists of Ceph scores and clinical features (including age, gender, limited mouth opening, crepitus, etc.). This combined model could enhance the accuracy of DJD screening and increase the optimization of decision-making in dental practice. Future studies are needed to test the reliability of the model with similar cephalometric parameters.
一种新型颅影测量基础的组合模型被构建并验证用于筛查 DJD。该模型由颅影分数和临床特征(包括年龄、性别、张口受限、关节响等)组成。这个组合模型可以提高 DJD 筛查的准确性,并增加牙科实践中决策的优化。未来的研究需要测试该模型在类似颅面测量参数下的可靠性。

Author credit statement 作者署名声明

Xinyi Fang contributed to conceptualization, methodology, software, validation, formal analysis, investigation, original draft preparation, and visualization; Xin Xiong contributed to conceptualization, methodology, software, investigation, and original draft preparation; Jiu Lin contributed to validation, formal analysis, investigation, original draft preparation, and visualization; Yange Wu contributed to resources and data curation; Jie Xiang contributed to resources and data curation; and Jun Wang contributed to manuscript review and editing, supervision, project administration, and funding acquisition.
方新怡在概念化、方法论、软件、验证、正式分析、调查、原始草稿准备和可视化方面做出了贡献;熊鑫在概念化、方法论、软件、调查和原始草稿准备方面做出了贡献;林久在验证、正式分析、调查、原始草稿准备和可视化方面做出了贡献;吴扬根在资源和数据管理方面做出了贡献;向杰在资源和数据管理方面做出了贡献;王俊在手稿审阅和编辑、监督、项目管理和资金获取方面做出了贡献。

Supplementary Data 补充数据

Calculation formulas for Ceph score, clinical model, and combined model
Ceph 评分、临床模型和组合模型的计算公式

Ceph score = −0.00776 × N-S-Ar (SaddleAngle, °) + 0.20498 × S-Ar-Go (ArticularAngle, °) − 0.09766 × S-N (Anterior Cranial Base, mm) − 0.06531 × S-Ar (Posterior Cranial Base, mm) − 0.02252 × Co-A (Midface Length, mm) + 0.3145 × Ar-Go-Me (Gonial/JawAngle, °) − 0.009 × Dc-Xi-Pm (°) + 0.2901 × Y-Axis (SGn-FH, °) − 0.22797 × Y-Axis Length (mm) + 0.20486 × Go-Me (Mandibular body length, mm) + 0.26303 × ANB − 0.29693 × FMA (FH-MP) + 0.03459 × PP-FH (°) − 0.02125 × MP-OP (°) + 0.27777 × PP-OP (°) − 0.27614 × ANS-Xi-Pm (°) + 0.03083 × U1-SN (°) + 0.07146 × FMIA (L1-FH) + 0.24926 × U6.PP.MM + 0.26686 × L1-Me (mm) − 0.1302 × Overbite (mm) + 0.0174 × LL-EP (mm) 重试    错误原因

Clinical score = 0.81005 − 0.80942 × (1, Gender = male) − 0.01537 × Age + 0.63207 × (1, Limited mouth opening = yes) − 0.30825 × (1, Deviation = yes) + 0.39300 × (1, Clicking = yes) + 1.40849 × (1, Crepitus = yes) − 0.25286 × (1, Pain = yes)
临床评分 = 0.81005 − 0.80942 × (1, 性别 = 男) − 0.01537 × 年龄 + 0.63207 × (1, 限制嘴开口 = 是) − 0.30825 × (1, 偏斜 = 是) + 0.39300 × (1, 咔嗒声 = 是) + 1.40849 × (1, 磨擦声 = 是) − 0.25286 × (1, 疼痛 = 是)

Nomoscore = −85.75810 − 0.11527 × (Gender = male) − 0.02214 × Age + 0.75705 × (1, Limited mouth opening = yes) − 0.00087 × (1, Deviation = yes) + 0.39070 × (1, Clicking = yes) + 1.45155 × (1, Crepitus = yes) + 0.25464 × (1, Pain = yes) + 1.27139 × Ceph score.
Nomoscore = −85.75810 − 0.11527 × (性别 = 男) − 0.02214 × 年龄 + 0.75705 × (1, 张口受限 = 是) − 0.00087 × (1, 偏斜 = 是) + 0.39070 × (1, 响声 = 是) + 1.45155 × (1, 摩擦音 = 是) + 0.25464 × (1, 疼痛 = 是) + 1.27139 × 头影测量评分。

  1. Download: Download high-res image (473KB)
    下载:下载高分辨率图像 (473KB)
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Supplementary Fig 1. Images of health TMJ and TMJs with degenerative joint diseases: A, Flattening; B, Erosion; C, Osteophyte; D, Subcutaneous cyst; E, Extensive sclerosis; F, Erosion (coronal view); G, Erosion (transverse view); H, ∼l healthy joint; K, Healthy joints (transverse view).
补充图 1. 健康 TMJ 和患有退行性关节疾病的 TMJ 的图像:A,扁平化;B,侵蚀;C,骨赘;D,皮下囊肿;E,广泛性硬化;F,侵蚀(冠状面);G,侵蚀(横断面);H,∼l 健康关节;K,健康关节(横断面)。

  1. Download: Download high-res image (497KB)
    下载:下载高分辨率图像 (497KB)
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Supplementary Fig 2. Cephalometric tracing.
补充图 2. 头影测量描记。

  1. Download: Download high-res image (649KB)
    下载:下载高分辨率图像(649KB)
  2. Download: Download full-size image
    下载:下载完整尺寸图像

Supplementary Fig 3. Bland-Altman plots demonstrate the bias for cephalometric variables: A-C, ANB; D-F, FH-PP; G-I, S-N. Only 3 parameters were shown in this figure. A, D, and G, Showed intraobserver error for observer 1. B, E, and H, Showed intraobserver error for observer 2. C, F, and I, Showed interobserver error. Red indicates the 95% limits of test-retest agreement. Black indicates the mean of the differences, which are close to 0, indicating low test-retest bias.
补充图 3. Bland-Altman 图示说明了头影测量变量的偏差:A-C,ANB;D-F,FH-PP;G-I,S-N。此图中仅显示了 3 个参数。A、D 和 G 显示了观察者 1 的观察者内误差。B、E 和 H 显示了观察者 2 的观察者内误差。C、F 和 I 显示了观察者间误差。红色表示测试-重测一致性的 95%限值。黑色表示差异的平均值,接近 0,表明测试-重测偏差低。

Supplementary Table I. Detailed description of extracted cephalometrics features
补充表 I. 提取的头颅测量特征的详细描述

Variables 变量DJDNormal 正常P value p 值
Cranial base 重试    错误原因
 N-S-Ar (saddle angle) (°)
N-S-Ar(鞍角)(°)
125.2 ± 5.4126.6 ± 4.90.004
 S-N (anterior cranial base) (mm)
S-N(颅底前部)(mm)
62.2 ± 3.163.3 ± 3.3<0.001
 S-Ar (posterior cranial base) (mm)
S-Ar(后颅底)(mm)
32.4 ± 3.233.2 ± 3.30.013
Maxillary skeletal 上颌骨骼
 Co-A (midface length) (mm)
共同 A(中面长度)(毫米)
80.1 ± 4.681.1 ± 6.10.044
 SNA (°)82.2 ± 3.581.5 ± 3.50.033
Mandibular skeletal 下颌骨骼
 SNB (°)77.5 ± 3.878.1 ± 4.00.089
 S-Ar-Go (articular angle) (°)
S-Ar-Go (关节角) (°)
151.9 ± 6.7148.6 ± 5.9<0.001
 Ar-Go-Me (gonial/Jaw angle) (°)
阿-go-me(下颌角/下颌角度)(°)
117.7 ± 6.5116.8 ± 5.6 重试    错误原因 0.121
 Dc-Xi-Pm (°)36. 9 ± 6.0 36.9 ± 6.037.3 ± 4.90.467
 Ar-Go (ramus height) (mm)
Ar-Go(髁突高度)(mm)
45.8 ± 4.4 重试    错误原因 48.9 ± 4.5 重试    错误原因 <0.001
 Go'-Me (mandibular body length) (mm)
下颌体长度 (Go'-Me) (毫米)
68.5 ± 4.768. 8 ± 5.20.592
Maxillary/Mandibular 上颌/下颌
 ANB (°)4.7 ± 2.83.4 ± 2.2<0.001
 Wits (mm) 智力 (毫米)0.9 ± 3.80.2 ± 2.90.043
Vertical dimension 垂直维度
 FMA (FH-MP) FMA(FH-MP)25.0 ± 5.622.8 ± 5.2<0.001
 SN-MP (°) SN-MP(°)34.9 ± 6.132.4 ± 5.9<0.001
 PP-FH (°) PP-FH (°)−0.17 ± 2.8−0.2 ± 2.70.759
 OP-FH (°) 术后热 (°)8.7 ± 3.97.5 ± 3.70.001
 MP-OP (°) MP-OP(°)16.3 ± 4.215.4 ± 3.80.018
 PP-OP (°) PP-OP(°)8.5 ± 3.66. 8 ± 3.2<0.001
 ANS-Xi-Pm (°)47.2 ± 4.947.0 ± 4.40.637
 N-Me (anterior face height) (mm)
N-Me(前脸高度)(mm)
114.8 ± 6.2115.3 ± 7.30.400
 S-Go (mm) S-Go (毫米)75.8 ± 5.979.1 ± 6.3<0.001 重试    错误原因
 Y-axis (SGn-FH) (°) Y 轴 (SGn-FH) (°)61.5 ± 3.660.7 ± 3.2 重试    错误原因 0.012
 Y-axis length (mm) Y 轴长度 (毫米)115. 4 ± 6.2117.2 ± 7.50.004
Dentoalveolar 牙槽骨的
 U1-L1 (interincisal angle) (°)
U1-L1(切牙间角)(°)
125.6 ± 12.8126.2 ± 10.20.625
 U1-SN (°)102.2 ± 8.6103.5 ± 8.10.087
 FMIA (°) 股骨内侧髁 (°)57. 7 ± 9.2 57.7 ± 9.259.3 ± 7.00.040
 IMPA (L1-MP)97.3 ± 8.197.8 ± 6.50.487
 U6-PP (mm) U6-PP (毫米)22.5 ± 2.222.8 ± 2.30.173
 L6-MP (mm) L6-MP(毫米)31.9 ± 2.831.9 ± 2.80.971
 U1-ANS (mm) U1-ANS (毫米)28.7 ± 2.628.1 ± 2.60.011
 L1-Me (mm) L1-Me (毫米)40.0 ± 3.139.8 ± 3.30.508
 Overjet (mm) 超喷射(毫米)4.2 ± 1.83.7 ± 1.40.006
 Overbite (mm) 前突 (毫米)2.6 ± 2.02.5 ± 1.40.671
Facial profile 面部轮廓
 UL-EP (mm) UL-EP (毫米)0.7 ± 2.8−0.1 ± 2.2 -0.1 ± 2.20.001
 LL-EP (mm) 重试    错误原因 0.8 ± 2.80.8 ± 2.30.876

Note. Values are presented as mean ± standard deviation.
注:值以平均值±标准差表示。

Supplementary Table II. Diagnostic performance of models on the training and validation sets
补充表 II. 模型在训练集和验证集上的诊断性能

Models 模型Training set (n = 332) 训练集(n = 332)Test set (n = 141) 测试集 (n = 141)
Sensitivity 敏感性Specificity 特异性Accuracy 准确性AUC (95% CI)Sensitivity 敏感性Specificity 特异性Accuracy 准确性AUC (95% CI) 重试    错误原因
Clinical model 临床模型0.5130.8050.6150.701 (0.645-0.756)0.6060.6380.6160.603 (0.505-0.701)
Ceph score 头颅评分0.7500.8460.7840.861 (0.822-0.900)0.7020.8940.7620.812 (0.738-0.886)
Combined model 组合模型0.8730.7560.8320.883 (0.847-0.919)0.7020.8720.7550.828 (0.757-0.899)

Download: Download video (41MB)

Video.

References

Xinyi Fang and Xin Xiong are joint first authors and contributed equally to this work.
方新义和熊新为共同第一作者,对这项工作贡献相同。

All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest, and none were reported.
所有作者已完成并提交了 ICMJE 潜在利益冲突披露表格,并且未报告任何冲突。

This research was funded by the National Natural Science Foundation of China (nos. 81771114 and 81970967), the Sichuan Science and Technology Program (no. 2020YFS0173), and the Major Special Science and Technology Project of Sichuan Province (no. 2022ZDZX0031).
本研究得到了中国国家自然科学基金(项目号:81771114 和 81970967)、四川省科技计划(项目号:2020YFS0173)以及四川省重大专项科技项目(项目号:2022ZDZX0031)的资助。

View Abstract 查看摘要