Study On Human Body Modeling For Thermal Analysis Based On Infrared Image

In recent years,infrared imaging technology has been widely used as an important nondestructive testing technology,and it has also made important progress in biomedicine.It is found that the infrared thermal imaging technology has special advantages in early detection and diagnosis of tumors compared with X-ray imaging or ultrasound imaging.This is because the former detects changes in physiology of the body's tissues,and the latter detects morphological changes.As a portion of the human body,when one feels ill,the metabolism changes of tissue cells always precede the morphological changes.Due to this reason,supplemented by infrared thermal imaging technology,the diagnosis accuracy of X-ray tumor detection can be greatly improved.Therefore the diagnosis method based on infrared optics is becoming an attractive subject in the biomedical fields,and the human body modeling for thermal analysis is one of the essential parts of the subject.In this thesis,the establishment of the finite element models of layered medium heat sources and the prediction of the spatial information of abnormal heat source are studied.The main works are as follows.1.The finite element model of human leg is established.According to the clinical empirical parameters,the geometric structure of four-layers of human legs is established.The heat generation rate of blood perfusion as a function of polar radius coordinates in each layer is given.The numerical simulation results are then obtained with the model by loading the heat generation of basal tissue metabolism,blood perfusion and tumor tissue metabolism.By varying the radius and depth of the tumor,the influence of abnormal heat source with different size and different position on the temperature field,inside the body and at skin surface,is obtained.It is shown that in the range of tumor radius from 0.50 cm to 0.70 cm?initial stage?,a smaller tumor yields higher maximum temperature and narrower FWHM of the temperature distribution on the skin surface.2.The finite element model of human breast is established.According to the clinical empirical parameters,the geometric structure of four-lays of human breasts is established.The heat generation rate of blood perfusion as a function of polar radius coordinates and zenith angles in each layer is given.The numerical simulation results are then obtained with the model by loading the heat generation of basal tissue metabolism,blood perfusion and tumor tissue metabolism.By varying the radius and depth of the tumor,the influence of abnormal heat source with different size and different position on the temperature field,inside the body and at skin surface,is obtained.It is shown that as the tumor radius increases from 0.50cm,the feature temperature at skin surface corresponding to the tumor center position?y axis?decreases at first,and then increases.The translation point appears at the tumor radius of 0.65cm.3.The neural network prediction model is established.A preliminary BP neural network is established which is with single hidden layer and 73 input neurons.The transfer function of the hidden layer is hyperbolic tangent function,and that of the output is linear function.The training function is Levenberg-Marquardt BP algorithm.Based on the established finite element model of human breast the temperature distribution on the skin surface for certain size and certain position of tumors is obtained.One group of the input data of the net is formed by combining the temperature data with the information of depth and radius of the abnormal heat source.125 groups of input data,which are acquired by equidistant cycle transform of depths and radiuses of heat source,form the training set of the network.Other 25 groups form the testing set of the network.It is shown with the testing set that the mse error of the prediction of the network is in the order of magnitude of 10^-8m.This research shows that constructing suitable layered finite element model of human body to analyze the influence of the abnormal heat source with different size and different depth on the temperature distribution at skin surface is an important portion for the prediction of the heat source information.Taking the advantages of automatic judgment and continuous learning,and with accumulation of large amounts of clinical data,artificial neural network will play an important role in infrared biomedical diagnosis of artificial intelligence auxiliary.