Department of Mechanical Engineering, K. N. Toosi University of Technology, Tehran, Iran.
Int J Numer Method Biomed Eng. 2019 Jun;35(6):e3192. doi: 10.1002/cnm.3192. Epub 2019 Mar 7.
Thermography is a developing and noninvasive medical imaging technique that can be used for diagnosis of body disorders based on temperature deviation from normal body temperature. This research investigates the feasibility of thermography method in conjunction with artificial neural networks (ANNs) for detection of thyroid tumors. For this purpose, first, a 3-D model of the healthy human neck is constructed based on patient-specific computed tomography (CT) images. This model is used for analyzing bio-heat transfer in the human neck. The healthy thyroid gland is considered as a heat source and generates heat according to its temporal temperature. Finite element results verify the thermography potential for detection of thyroid gland location and estimation of its butterfly shape on the neck thermogram. The numerical analysis is carried out on 35 models with varying thermo-physical parameters of the healthy thyroid gland, including heat generation and blood perfusion. The acquired thermograms are used to develop an ANN for correlating the thermo-physical parameters of the gland and temperature profile on the neck surface. In the next stage, dynamic thermal images are captured from 10 healthy and three cancerous human cases. The experimental thermal images are analyzed by the developed ANN and the corresponding thermo-physical parameters are obtained. Results show that the estimated heat generation values for the healthy cases are about 3000 while it increases to more than 12 000 for the cases with tumors. This significant variation confirms the potential of dynamic thermography in diagnosis of thyroid tumors.
热成像技术是一种发展中的非侵入性医学成像技术，可用于根据体温与正常体温的偏差来诊断身体疾病。本研究探讨了热成像方法与人工神经网络（ANNs）结合用于检测甲状腺肿瘤的可行性。为此，首先，根据患者特定的计算机断层扫描（CT）图像构建了健康人颈部的 3D 模型。该模型用于分析人体颈部的生物传热。健康的甲状腺被视为热源，根据其时间温度产生热量。有限元结果验证了热成像在检测甲状腺位置和估计其在颈部热图上的蝶形方面的潜力。对 35 个具有不同健康甲状腺的热物理参数的模型进行了数值分析，包括产热和血流灌注。获得的热图像用于开发用于关联腺体的热物理参数和颈部表面温度分布的 ANN。在下一阶段，从 10 个健康人和 3 个癌症患者中捕获动态热图像。通过开发的 ANN 对实验热图像进行分析，并获得相应的热物理参数。结果表明，对于健康案例，估计的产热值约为 3000，而对于肿瘤患者，产热值增加到 12000 以上。这种显著的变化证实了动态热成像在甲状腺肿瘤诊断中的潜力。
