Study On The Temperature Field Of The Conformal Therapy By Radiofrequency Ablation In Spinal Tumor

Spinal tumor is a serious disease to harm human health. Because of itself complex environment and the limitation of surgical resection, this disease has been cured by radiofrequency ablation in clinic. However, the basis reference data of spinal tumors is lack, and the worst defect is that choosing treatment parameters of radiofrequency ablation always depends on the treatment experience, which could cause the temperature field not to cover the whole tumor tissue. Therefore, the conformal therapy of spinal tumor by radiofrequency ablation is one of the important clinical problems to be solved.This study applied the finite element method to simulate spinal tumor by radiofrequency ablation, and analyzed the temperature field and ablation area at different situations including different treatment parameters(voltage and time) and different electrode structures. Meanwhile, in order to improve the accuracy of clinical treatment and achieve to predict the treatment parameters of different spinal tumors, this study has put forwards three kinds of optimization algorithms, including Back Propagation Neural Network(BPNN), Generalized Regression Neural Network(GRNN) and Support Vector Machine(SVM). Besides, selecting the best algorithm is to establish the prediction and optimization system, which could provide reference data for surgical decision.The simulation results by single needle electrode: The treatment voltages ranged from 8V to 22 V, and the heating-up times ranged from 2min to 20 min. The ablation region was gradually expanding from ellipsoidal shape to spherical shape. When the heating-up time exceeded 15 min, the ablation region was increasing smaller and smaller with the heating-up time went on. It could be recommended that the time should be limited within 15 min when using single needle electrode in clinic. The volume range of ablation zone was 70-11928mm3, and the transverse diameter range of ablation zone was 5.4-29.6mm. The small diameter(<30mm) spinal tumor could be treated by the single needle electrode in clinic.The simulation results by multi needles electrode: The treatment voltages ranged from 11 V to 22 V, and the heating-up times ranged from 2min to 20 min. The ablation region was gradually expanding to form mushroom shape because of the electrode structure particularity. The volume range of ablation zone was 81-126411mm3, and the transverse diameter range of ablation zone was 31.4-63.1mm, meanwhile the minimum radius is 3.9-21.1mm. The big diameter(?30mm) spinal tumor could be treated by the multi needles electrode in clinic.Because the change of ablation zone was quite complex, and they had no linear relationship. So we applied three different optimization algorithms to extract their numerical characteristics to achieve the aim of predicting treatment parameters in this study. In SVM model based on genetic algorithm, the average relative error was minimum, compared with that in SVM model based on cross validation method. Besides, the correlation coefficient was almost close to 1. In BP neural network model, the prediction results was accurate relatively, but the worst defect was that its predicted value was randomness. Moreover, the average relative error of GRNN model was the highest in three model. So SVM model based on genetic algorithm was the best and optimal one to apply in this study. And we have established an automatic optimization to predict treatment parameters of spinal tumor software based on this optimal algorithm. In this software, we can write the input value(the transverse diameter, longitudinal diameter, depth and volume about unknown spinal tumors), and predict the treatment parameters accurately and automatically. Moreover, the evaluated part is to assess the accuracy of prediction value, especially the coverage rate of ablation zone. This software can provide reference data for surgical decision to ensure the accuracy of the operation and the practical operation.