Three-dimensional Finite Element Simulation Of Brain Magnetic Induction Tomography Forward Problem

Magnetic Induction tomography is a new research branch of electrical Impedancetomography technology. It uses the coil contactless with the body surface to generate thealternating magnetic field. The inside of electrical conductivity object around the coilproduce secondary induction magnetic field. When the conductivity distribution within theobject changed, the induced magnetic field has been changing. Induction magnetic fielddetect by the coil assembly. Applying reconstruction algorithm, then the internal object orits conductivity distribution changes in the image are reconstructed. Alternating magneticfield as excitation source, can easily penetrate the insulation good skull; Contactlessdetection method to avoid the situation of cannot paste electrode, such as trauma, and canbe accurate positioning in advance; The MIT is a low cost, functional imaging same as theother imaging bioelectric impedance technique. Cerebral hemorrhage or edema can cause electrical conductivity distribution changes of cerebral tissue, MIT technique is well suitedfor long time monitoring, which has important medical significance.The research of MIT mainly focused in the computer simulation, data acquisitionsystem, imaging algorithm and software system. Some of them have got very goodprogress and confirmed the feasibility to detect the human tissue. The research resultsshow that the data accuracy, stability, sensitivity and image spatial resolution of MITsystem is not high that limited MIT applying to the clinical application. The key issue ofMIT imaging quality is the results accuracy of the forward and inverse problem. Due tothe relationship of boundary data measured by detecting coil and the target conductivitydistribution is complicated nonlinear relation with special boundary condition, so thesolution of the inverse problem, namely the process of image reconstruction relies on theforward problem solutions providing more realistic and accurate boundary parameters, tomodify and optimize image reconstruction algorithm and improve the accuracy of thesolution. Forward problem solving usually adopts the finite element numerical simulationmethods, this method can better simulate the real situation and is convenient and fasttransformation experimental conditions to study the change rule, which is not easy toimplement in living organisms. The currently problem of MIT simulation research is thatuse mostly round sphere, cylinder and geometric model instead of a brain model, thedifference between the rules model and the real shape of the human brain is very big, itwill bring large error in the calculation and inverse problem, which cause the error ofimage reconstruction. Also some research medical image reconstruction method is adoptedto establish the brain model. The steps and cycle of this kind of model establishmentmethod is long. The brain model contains real brain anatomical details will greatlyincrease the finite element mesh and calculated pressure, affect the efficiency ofsubsequent calculations. There are a few studies using2d head model. While thecalculation process is simple, but the distribution of magnetic field is three-dimensional inreal human brain, the2d approximation is bound to induce large errors.Aiming at these problems in the MIT simulation research, this study constructed areal head model of the structure of the human brain and coil system model. Based on the brain and ball finite element simulation model, the changes of MIT solving results due tothe model geometric structure condition is studied, the specific work are as follows:1. Establishment of true human brain3d finite element modelAnalysis of brain structure, the skull is the support structure of brain; includeparenchyma inward, outward to the scalp. Therefore, model can be build based on skullbrain model. Skull is not rule geometry, but overall structure is egg-shaped. Through theappropriate segmentation, it can use different size of the ellipsoid, sphere and cubegeometry to approximate. Using geometric modeling tools provided by ComsolMultiphysics3.5a electromagnetic field numerical calculation software package, normalhuman skull contour model first constructed, which the distance from front to back isabout19cm, the distance from left to right is17cm and the distance from top to bottom is17cm. Then based on the skull contour model, scale out0.5cm as head cortex, reduce1cminside as skull layer formation, skull internal form parenchyma layer. Using packages, thebrain model is divided for tetrahedron three-dimensional finite element subdivision.Finally the human brain3d finite element model of the brain is constructed consisting ofthe real three layers. The structure simulation degree of human brain model is good,nearby the reconstruction of3d real human brain. Especially the construction of the brainmodel is convenient, fast and can provide a suitable simulation model for brain MIT study2. Establishment of the coil modelPrevious MIT coil system used more abstract line forming ring coil, but the real coilof wire has a certain diameter and the turn of coil has a certain thickness. Given theimportance of the coil as a sensor in the MIT system, better coil simulation research isvery necessary. A hollow cylinder coil model is established through Comsol Multiphysics3.5a finite element simulation software, the thickness of the cylinder wall simulate thewire diameter and the length of the cylinder to simulate the number of turns. Simulationcomputation results show that the structure and running state can be well used to simulatethe MIT solenoid coil and the exciting current and magnetic induction intensity in thespace domain and time domain parameter are good consistent with the actual trend.Considering the exciting current of MIT systems generally MHz above, the metal conductor skin effect in high frequency electromagnetic field is obvious. The model canbe further simplified as coil cylinder model. Compared with hollow cylinder model andcylindrical coil model, the finite element number, solving the degrees of freedom andsolving time reduced greatly. The calculation pressure is reduced and the efficiency of thesolution is improved. Considering the calculation accuracy and lower calculation pressureof simulation calculation of complex3d model, the cylinder model is more suitable for thesubsequent simulation research. Compared with the existing research adopts linearexcitation coil and the point, its structure is more close to reality and provides a kind ofaccurate, rapid, efficient research method as the MIT excitation source and data detection.3. The forward problem study of MIT based on real human modelA ball model is established that the volume is same with real brain model. the ballmodel also has three layers structure consisting of scalp, skull and brain parenchyma, thethickness of the scalp is0.5cm, the thickness of the skull is1cm and the skull layer radiusis16cm. The real contour model and balls are placed in brain16channel simulation coil.The same conductivity hemorrhage disturbance is set in the two models respectively.Applying10MHz and1A sinusoidal current in the excitation coil, the MIT forwardproblem is calculated, and the phases on detection coil are measured. Then changing thevolume of hemorrhage disturbance, position and shape, phases are calculated underdifferent situation. Calculation results show that the phase difference and phase data trendon15test coils is consistent with the existing simulation research reports with samehemorrhage in both model. The phase generated by the rules ball model is symmetry,basically it is a symmetrical distribution with8detecting coil as the center; The simulationresults of real contours brain model show that the phase difference of the results on thesymmetry position coil is not same, the phase difference of1to7detecting coil is notsame with the phase difference of its symmetrical position9to15detection coil. Thedifference is around10%, of which4and12, the difference of detecting coil is the largest.This difference is due to brain model geometry asymmetry, leading to the distancebetween coil and the model is not completely consistent. Analyzing the structure of brainmodel, the detecting coil is near the model from1to7. The research shows that: in the case of same conditions, the difference of MIT brain model geometry will affect forwardproblem results, the differences of rules ball model and brain model are bound to affect theaccuracy of the inverse problem results.4. The image reconstruction based on real human brain modelMIT simulation image are reconstructed by image reconstruction software based onthe modified Newton-Raphson reconstruction algorithm. The image reconstructionsoftware is based on the two-dimensional circular domain for image reconstruction. Aspherical hemorrhage model with radius of2cm and the electrical conductivity of1S/m,2S/m and3S/m at the model center are reconstructed. The results showed that thereconstruction image of both model can reflect the location of the object. Brain modelreconstruction result is higher contrast and show more obvious. Moving the hemorrhagedisturbance near to the excitation coil, the reconstruction results can reflect the location ofthe object, the difference of object location is not obvious. when hemorrhage disturbanceposition move to the edge of distance from center of6cm, image reconstruction resultsshow that the perturbation target location of real brain closer to the center than the ball,difference of distance is1cm. A rectangle hemorrhage disturbance, which length, widthand height is4cm,2cm,2cm is reconstructed, the results showed that disturbance targetsof brain model is more close to the outline of the setting of disturbance, and disturbancetargets of ball model is more different. Due to the solution of the inverse problem is basedon the subdivision of two-dimensional circular domain, these differences may be moreapparent, target position and contour may be more accurately using real human braingeometry for solving inverse problems.Under the same experimental conditions, that the simulation results of real brainmodel and rule ball model is different, and this difference is caused by the modelgeometry differences. This suggest us that: simulation result of forward problem is able toprovide a more realistic and accurate testing data and the reconstructed image will be moreaccurate using real human brain contour model.