Study And Application Of The Highly Stable And Fast Optimization Algorithm For Intensity Modulated Proton Therapy

Proton radiation therapy has great advantages in radiotherapy of cancer owing to the Bragg peak and is being rapidly developed at home and abroad.As the advanced form,Intensity modulated proton therapy(IMPT)is based on the active scanning technique involving the modulation of the intensity weight of each scanning spot.IMPT can achieve excellent dose conformation in the tumor target,as well as better protection of organs at risk(OARs).However,due to special physical characteristics and delivery technique,the optimization for IMPT planning faces three major challenges:1)It involves many scanning spots whose optimal weight intensity is hard to find.And these spots lead to huge amount of dose martrix data requiring large memory and time.2)It can suffer from long scanning path sometimes needing further optimization effort caused by unnecessary pass-through in the zero-weight area.3)It often requires special steps to improve the plan stability and optimization robustness.The objective of this thesis research project is to develop and apply highly stable and fast optimization algorithms that solve the above-mentioned challenges and finally improve the practicality of IMPT.Three research tasks were carried out:(1)To study the intensity optimization method of IMPT;(2)To study the scanning path optimization method;(3)To study the robust optimization method.In order to complete the above three tasks,three corresponding methods were used:(1)A sparse matrix preprocessing method was developed using hybrid conjugate gradient algorithm to solve the intensity optimization problem.(2)An adaptive genetic algorithm was developed to solve the scanning path optimization problem.(3)A synchronous optimization based robust method was developed to reduce the influence of uncertainties.The following results were obtained:(1)The AAPM TG-119 and clinical cases for intensity optimization show that,the dose statistics of the optimized IMPT plan for each case can meet with the dose prescription.And,after sparse martrix handling,the storge needed for intensity optimization is reduced by 76.5%,80.3%,79.3%and 83.7%,with 2.46,1.91,2.45 and 1.75 times optimization speed-up,respectively.(2)The four tested cases for scanning path optimization show that,adaptive genetic algorithm based scanning path optimizaed is able to reduced the total planning path length by 27.2%,18.7%,25.4%and 33.0%,respectively.(3)The AAPM TG-119 and clinical cases for robust optimization showed that,the robust IMPT plans had narrower amplitude of dosimetric disturbance under uncertainty influence.Synchronous optimization method based robust optimization is found to improve the stability of IMPT plan.These results have lead to the following conclusions:the three tasks have been successfully completed and the set of optimization modules are found to be clinically useful for IMPT optimization.