Prediction And Optimization Of Light Field Distribution In Spherical Tumour Based On Machine Learning

In the era of great health,the medical field is booming.People are also trying to use physical means to treat diseases,such as lighting the affected parts.Photodynamic therapy is a method of producing active oxygen by light to kill cancer cells.Photodynamic therapy has a huge advantage over traditional methods in the treatment of tumors,but photodynamic therapy still has the problem of unstable treatment.The main reason is the lack of precision treatment standards.Light distribution in photodynamic therapy is an important research content.This paper studies the prediction and optimization of the internal light flux distribution of spheroid tumors based on Monte Carlo method and machine learning algorithm.A Monte Carlo method was used to establish a spherical tumor model under red light with a wavelength of 640 nm.The internal light flux distribution was studied.It was found that the number of incident photons had little effect on the light flux distribution inside the tumor.The distance from the light source to the tumor was higher.Small,the more uniform the distribution of light flux inside the tumor;the smaller the tumor radius,the more uniform the distribution of light flux inside the tumor;when the light source is placed symmetrically around the tumor,as the number of light sources increases,the distribution of light flux inside the tumor becomes more and more uniform.The light distribution of the tumors under different illumination conditions and different tumor sizes was established into a data set.The CART decision tree algorithm in machine learning was used to learn the light flux distribution data set under single light source conditions,and the effect on the internal light flux distribution of the tumor was found.The order of large to small is: tumor radius,distance from light source to tumor,number of photons,and position of light source.Then the GDBT decision tree algorithm in machine learning is constructed and the light flux distribution data set under multi-light source conditions is used to learn.The problem of light source placement under multi-light source illumination is solved,and the light source placement scheme is optimized by precise calculation.In this paper,the Monte Carlo tissue optical model is combined with the machine learning method,and the Monte Carlo method is used to construct the model.The influence of each physical quantity on the internal light flux distribution of the tumor is studied.The training data set is generated,and the importance of each physical quantity is calculated by using machine learning.Sorting,and suggesting the light source placement scheme under the ideal light distribution,has important guiding significance for the control of light distribution in precision dose photodynamic therapy.