The Method For Dose Calculation Of A-Si EPID Based On Neural Network

With the rapid progress in physics and biological theory,radiotherapy technology has undergone great development and become one of the three major tumor treatment methods.In order to improve the therapeutic results and reduce the side effects of radiation on the human body,it is necessary to precisely control the dose distribution of the irradiation field when the radiation therapy is performed on the patient,while minimizing or even preventing radiation from harming healthy tissues and organs near the target area.Because the field complexity and dose rate in intensity modulation radiation therapy is higher than that of other radiotherapy,there are likely to be errors in all aspects of treatment,and therefore,Dose-verification is needed for the plan of intensity modulation radiation therapy.Currently,dose-verification based on aSi EPID is one of the promising dosimetry techniques.The work of this paper is to study the dose response characteristics of a-Si EPID subfield,and discuss the method of dose calibration of a-Si EPID using computer technology,and provide technical support for the dose-verification and other follow-up work.Firstly,collected the gray image of a-Si EPID device with digital medical accelerator at different output dose,and collected the spatial distribution data of irradiation dose with Threedimensional water phantom system.Since the spatial resolution of the grayscale image from the a-Si EPID is much higher than the resolution of the dose distribution data,in order to obtain sufficient data,it is necessary to perform data structure conversion and data supplement on the low-resolution dose distribution data of the collected text format.In the data acquisition process,there are often many factors can affect the accuracy of data acquisition,while the neural network has strong ability to self-learning,to deal with inaccurate information capabilities,strong anti-interference and anti-noise ability.In this paper,neural networks are used to process the data of radiotherapy.Based on the analysis of the characteristics of various interpolation and fitting algorithms,and according to the characteristics of the dose distribution,the cubic spline interpolation method is used to interpolate the dose stationary region,the regression neural network based on genetic algorithm is used to fit the region of complicated dose distribution,and finally forming a 2048 × 2048 two-dimensional matrix,and utilizes the BP neural network model to study the dose response of amorphous silicon EPID.Then,fusing the 1-12 MU images by the magnitude of the radiation dose,in view of the dispersion of the response properties of the amorphous silicon EPID sub field and the spatial characteristics of the 1-12 MU fusion image,the space linear neural network model was designed and constructed,evaluated the model by using validated samples,and indicated the feasibility of using the model to achieve dose calibration.