Research On Several Optimization Problems In IMRT Planning

Intensity modulated radiation therapy (IMRT) is an important radiation therapy method for tumor treatment, in which high-energy particles can be accurately cast into the target volumn while ensuring that the organs at risk (OAR) receive doses of lower than tolerated dose, thereby enhancing the cure rate of cancer patients and improving their quality of life after recovery. In order to furtherly improve the efficiency of IMRT, this paper investigated several core techniques including polygon clipping al-gorithm, segmentation algorithm and a distributed algorithm of Monte Carlo (MC) dose calculation.Contour outline of target volume or OAR usually involves graphic operations, such as merging margin lines, remove inside and outside lines etc., which are essen-tially the boolean calculations of union, intersection and difference between polygons. This paper presents a new polygon clipping algorithm that adopts circular linked list data structure, in which each node only has one single pointer, to represent the poly-gons and clipping results. The algorithm only needs to traverse the polygons’list one time, and its time complexity is 0 ((n1+n2+m)* log d). Compared to Vatti, Greiner, Liu Yongkui, Weiler algorithms, the algorithm in this paper not only can be applied to general polygons, such as perforated polygons, but also has a lower space and time complexity. The experimental results indicate that when the polygons has more vertices and fewer intersection points, the polygon clipping algorithm is more efficient, about 280 times faster than the Weiler algorithm.For IMRT static mode (Step and Shoot), based on Xia and Verhey algorithm, this paper proposes an improved segmentation (also be called intensity matrix splitting) algorithm that can furtherly reduce the total treatment time (T). The basic idea of the proposed algorithm is to repeatedly identify several optimized decomposition coeffi-cient vector, which has less number of elements and the sum of the elements is as small as possible, and then make the intensity matrix’s leftmost column elements gradually become zero, until the entire intensity matrix becomes zero matrix. Instead of using the combination of the total number of monitor unit (TNMU) and the number of segments (NS), the segmentation algorithm employs the total treatment time (T) as performance evaluation index because the T needs to consider not only TNMU and NS, but also the moving speed of multi leaf collimator (MLC) and the verification and recording Time (V&R). Taking these four parameters into account can better reflect the real efficiency of the algorithm and therefore has more practical significance. Through test with the random matrix data, the suggested algorithm, compared to Gal-vin, Bortfeld, Xia and Verhey, Siochi, Luan and Engel algorithms, always gives the shortest the total treatment time, which shows preliminarily that the improved seg-mentation algorithm is more efficient. Further, through a large number of clinical test data, the improved algorithm can still come to the same conclusion. Thus, the im-proved segmentation algorithm is more practical and feasible.Due to the high precision for dose calculation, Monte Carlo (MC) algorithm en-joys the "gold standard" in the radiation therapy field and has been widely studied in academic and industrial areas. MC algorithm needs to simulate a large number of par-ticles (typically> 10^6) to obtain almost identical results with the real experiments, so consumes very long computation time and is rarely applied to the actual treatment planning system (TPS). This article introduces the idea of distributed computing into MC dose calculation. Experiments with clinical data show that this strategy can greatly reduce the dose calculation time of TPS system and accelerate the speed of planning design. Moreover, the higher the performance of distributed computers, the faster the dose calculation. So, the proposed MC distributed algorithm provides a practical solution for the application of MC algorithm in TPS.Although three proposed optimization algorithm in this paper could not cover all the issues in IMRT, they have been able to make a lot of improvements on the effi-ciency of IMRT system. Physicians can more quickly and accurately outline the con-tours of organization structure including target volume and OAR by the optimized polygon clipping algorithm. Improved segmentation algorithm can not only save the time for radiation therapy and increase utilization of X radiation, but also improve tumor’s control rate and greatly protect the organs at risk. MC distributed dose calcu-lation algorithm being introduced into real TPS system can furtherly enhance the effi-ciency of IMRT. Therefore, several optimization problems in this paper are of great significance to the improvement and development of IMRT.