Homogeneity and Conformity of Neural Network-Based Lung-IMRT Planning

Abstract

The IMRT planning technique applies the concept of irradiation, which is controlled automatically by a computer. An IMRT plan is aligned with a trial-and-error approach and still involves non-intuitive, iterative steps based on the planner’s subjective decision. The Neural Network method is used in radiotherapy planning in determining IMRT plans in lung cancer cases. This method is used to predict dose distribution based on previous planning data. The purpose of using this neural network method is to predict the dose distribution in the PTV volume with validation in the previous plan, also predicting the dose distribution for doses that cover 95% of the target volume. So, this can make it easier for a planner to make decisions objectively. The obtained results show that the quality of planning produced based on neural network modelling has a homogeneity index (HI) of 0,09 ± 0,02, and the conformity index (CI) of 1,2 ± 0,27 with an average dose 1,02 ± 0,01 was the mean received at the target organ. The maximum dose to the at-risk right lung organ is 0,82 ± 0,22 Gy, the left lung is 0,75 ± 0,29 Gy, the heart is 0,77 ± 0,14 Gy, and the spinal cord is 0,50 ± 0,14 Gy.