School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Ave, 639798, Singapore.
School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Ave, 639798, Singapore.
Comput Biol Med. 2019 Oct;113:103419. doi: 10.1016/j.compbiomed.2019.103419. Epub 2019 Aug 28.
In the present study, an infrared (IR) thermal camera was used to map the temperature of the target skin surface, and the resulting thermal image was evaluated for the presence of carotid artery stenosis (CAS). In the presence of stenosis in the carotid artery, abnormal temperature maps are expected to occur on the external skin surface, which could be captured and quantified using IR thermography. A Duplex Ultrasound (DUS) examination was used to establish the ground truth. In each patient, the background-subtracted thermal image, referred to as full thermal image, was used to extract novel parametric cold thermal feature images. From these images, statistical features, viz., correlation, energy, homogeneity, contrast, entropy, mean, standard deviation (SD), skewness, and kurtosis, were calculated and the two groups of patients (control and diseased: a total of 80 carotid artery samples) were classified. Both cut-off value- and support vector machine (SVM)-based binary classification models were tested. While the cut-off value classification model resulted in a moderate performance (70% accurate), SVM was found to have classified the patients with high accuracy (92% or higher). This preliminary study suggests the potential of IR thermography as a possible screening tool for CAS patients.
在本研究中，使用红外（IR）热像仪来绘制目标皮肤表面的温度图，并用所得的热图像评估颈动脉狭窄（CAS）的存在。在颈动脉狭窄的情况下，预计在外部皮肤表面会出现异常的温度图，可以使用红外热成像来捕获和量化。使用双工超声（DUS）检查来建立基准。在每个患者中，使用背景减除后的热图像，即全热图像，来提取新的参数冷热特征图像。从这些图像中，计算统计特征，如相关性、能量、均匀性、对比度、熵、均值、标准差（SD）、偏度和峰度，并对两组患者（对照组和患病组：总共 80 个颈动脉样本）进行分类。测试了基于阈值和支持向量机（SVM）的二进制分类模型。虽然阈值分类模型的性能中等（准确率为 70%），但 SVM 发现可以对患者进行高精度分类（准确率在 92%以上）。这项初步研究表明，IR 热成像作为一种可能的 CAS 患者筛查工具具有潜力。
