The main objectives of this study are (i) to perform automated segmentation of facial regions from thermograms using k-means clustering algorithm and to classify the data into normal and orofacial pain (OFP) categories using various machine learning classifiers (ii) to implement the convolutional neural network (CNN) for classification of normal and OFP subjects which involves automated feature extraction and feature selection process. Fifty normal and 50 diseased cases suffering from orofacial pain were included in the study. Facial thermograms were segmented using k-means algorithm, then statistical features were extracted and classified into normal and OFP using various machine learning classifier. Further, the deep learning networks such as VGG-16 and DenseNet-121 were used for automated feature extraction and classification of facial thermograms. The facial temperature variations of 3.46%, 3.4%, and 3.27% were observed in the front, right and left side facial regions respectively between the normal and the OFP subjects. Machine learning classifiers such as support vector machine (SVM) and random forest (RF) classifier provided the highest accuracy of 99%. On the other hand, deep learning models such as modified VGG-16 achieved an average accuracy of 97% compared to modified DenseNet-121 which produced an average accuracy of 68% in classification of normal and OFP thermograms. Thus, computer aided diagnosis of facial thermography could be used as a viable screening device for a reliable identification of tooth pathology before the occurrence of structural changes and complications.

中文翻译：

---

使用机器学习和深度学习方法从热谱图自动检测口面部疼痛

本研究的主要目标是 (i) 使用 k 均值聚类算法从热像图中自动分割面部区域，并使用各种机器学习分类器将数据分类为正常和口面部疼痛 (OFP) 类别 (ii) 以实现卷积神经网络 (CNN) 用于对正常和 OFP 主题进行分类，涉及自动特征提取和特征选择过程。该研究包括患有口面部疼痛的 50 名正常病例和 50 名患病病例。使用k-means算法对面部热像图进行分割，然后使用各种机器学习分类器提取统计特征并将其分类为正常和OFP。更远，VGG-16 和 DenseNet-121 等深度学习网络用于自动特征提取和面部热像图分类。在正常和OFP受试者的正面、右侧和左侧面部区域分别观察到3.46%、3.4%和3.27%的面部温度变化。支持向量机 (SVM) 和随机森林 (RF) 分类器等机器学习分类器提供了 99% 的最高准确率。另一方面，与改进的 DenseNet-121 相比，改进的 VGG-16 等深度学习模型的平均准确度为 97%，后者在分类正常和 OFP 热像图时的平均准确度为 68%。因此，