| The effect of radiation therapy is always an important issue of clinical care. The purpose of the ideal radiation therapy is accurate to give tumor high doses at the same time decrease as much as possible on normal tissue around target area. With "precise positioning, precise plan, precise treatment" is the characteristic of accurate radiotherapy technology basic requirements in cancer treatment is on high precision, high doses, realize high curative effect and the low damage. In recent years in the domain of conformal radiotherapy, three dimensional conformal intensity-modulated radiotherapy technology and image guided radiotherapy technology for the small target and fixed tumor volume have good dose distribution and treatment effect, to a great extent, solve the problems of the static and rigid target in dose conformal radiation. These technology makes the tumors in the plan target volume (PTV) and shape more close to the clinical target volume (CTV), dose distribution on target area conformal better adjust. With the further development of the imaging technology, image guided radiotherapy become the key to solve the problem of uncertainty of the tumor in breathing condition between the inter-fractional and intra-fractional setup, and the latter is always the hot and the difficulty spot in field research. Because the IGRT technology is comparatively mature:earlier location technology is using electronic portal imaging device (EPID) which is integrated in accelerator position, in order to set a precise initial setup, EPID using bone structure and implanted reference mark to confirme target. Single plate X-ray imaging system is used to IGRT, not only can get better perspective image, but also can through the X-ray detection system rotating around the center of cone tomography, to get better effect of soft tissue resolution3D object imaging data.Along with the development of real-time imaging technology, dynamic adaptive radiation therapy as a real-time image guided radiotherapy form develop; and points in intra-fractional tumor tracking technology is becoming the research focus, also it’s an extremely challenging problems. Organs motion along with breathing motion is serious influence the current clinical radiation effect:in order to ensure that the beam fully cover tumor, usually, draw the outline of the target area edge enough, especially the treatment of tumor in chest and abdomen. The accurate radiotherapy in cancer treatment of the target, the target area with the inherent human body breathing movement changed position, in Yu[2] use the Picker company produces the picker-2000sCT machine to measure54cases peripheral lung cancer target displacement with breathing, and have a linear multivariate regression analysis with SPSS statistical software package, the results showed that Z axis displacement0.66±0.46cm (range0.10-2.00cm), the X axis displacement0.30±0.25cm (range0.00-0.90cm), the Y axis displacement0.37±0.34cm (range0.00～0.90cm); Erridge [3] use EPID system to observe25cases of small cell lung cancer lung tumor movement, found that the tumor movement at7.3mm±2.7mm in lateral direction,12.5mm±7.3mm in cranial-caudal direction, and9.4mm±5.2mm in ventral dorsal direction; Shimizu use CT scans patients with lung tumor under free breathe state, and calculated the displacement of the tumor with the relationship between the each side of triangle, the result show that lower lobes tumor average displacement9.1mm (3.4～24.0mm) in cranial-caudal direction,10.1mm (0-22.0mm) in lateral direction, the middle of the lung tumors average displacement6.2mm (2.4～11.3mm) in cranial-caudal direction, therefore, Shimizu points that it is necessary for the radiation plan put a safety boundary; Seppenwoolde implant a2mm diameter of gold mark in20patients around the tumor before treatment, take the movement of the gold mark in or around tumor on behalf of the movement of the tumor. CT scan under the quiet breathing condition, found that the lower lobes, not infringe upon the spine or chest wall such fixed structure of the tumors moving range are12mm±6mm in cranial-caudal direction, the upper lobe or infringe upon the fixed structure of tumors moving range are2mm±2mm in cranial-caudal direction.Based on the above analysis, the tumor target displacement with breathing motion will cause adverse impact in accurate radiotherapy location, so in order to ensure clinical tumor target area (CTV) to receive high doses of radiation, the traditional way is, when the tumor clinical target area (CTV) is sketched out, in consideration of a target of the breathing displacement, then continue to expand the border into the target area CTV plan target area (PTV), PTV will be as high dose reference area in the formulation of the radiation plan. Even though the tumors receiving high enough doses, but at the same time, more normal tissue around the target area will also be included into the high dose area, it will increase the probability of normal tissue complications occurred. So, expand the boundary of the target area it isn’t to meet requirements and expectations of accurate radiotherapy.In order to control the harmful effects of tumor displacement with breathing motion in accurate radiotherapy treatment, according to the related literature reports at home and abroad, the traditional breath control technologies and methods were adopted and promoted in clinical tumor radiation therapy, there are mainly including:1) deep inspiratory breath-hold method;2) thoracoabdominal passive pressure technology;3) active breathing control technology;4) respiratory gated technology;5) expand the radiation therapy target area boundary, etc. generation of These technicals have a positive impact for radiotherapy, but not a long time, people found that the technologies can not achieve the purpose of people expected very well. Deep inspiratory breath-hold method requires patients maintain a respiratory state during the course of treatment, that requirement is bad for those patients whose lung with functional deficiency; while passive pressure technology is simple, but the precision is poor; active breathing control technology’s weakness is that patients with residual volume are different before breath control every time, and repeat breathe in control cause fatigue and most of patients have respiratory system symptom and the tolerance is poor; respiratory gated technology requires radiation and patient breathing cycle synchronization, only in a small breathing period open ray radiation, this increases the treatment time; expand the radiation therapy target area boundary increased the radiation volume of health tissue around the tumor. The traditional methods have strict requirement of the patient’s respiratory condition, low efficiency in treatment, and health tissue around tumor always get more damage. So the tumor tracking technology more and more become the hot. But because of the correlation relationship between the breathing and tumor movement, we must analyze the change of tumor motion with breathing in radiation therapy every time. For dynamics of cancer radiation treatment research mainly include two parts:one is real-time tracking the position and shape of the tumor, another is dose transmission method that under the condition of tumor shape changed. The former considered the space physics positioning and beam tracking problem; The latter considered the treatment plan and tumor characteristics that adapt to the biological the radiation dose problem. For radiation treatment of tumors in thoracoabdominal with breathing movement, the first one is particularly important. But now because of restrictions of image sampling rate and system delays the recent available image can’t accurately reflect the location of the tumor in the image guided radiotherapy. Therefore, in order to effectively deal with the system time delay, control the process of dynamic radiation therapy with breathing signal, has to predict breathing motion signal or tumor position. Due to breathing motion signal inherent has very strong specificity, among different people and same person in different time the breathing pattern will have change, so precise and efficient respiratory motion signal prediction algorithm become one of the important direction in tumor radiation therapy studies,.This paper is dedicated to the study of prediction technology on thoracoabdominal tumor displacement with breathing movement. Because of the complexity and uncertainty of the breathing rhythms, we need a proper mathematical model to improve the accuracy of real-time prediction of breathing movement. This paper use Genetic Algorithm (Genetic Algorithm, the GA) to optimize BP (Back Propagation, BP) neural network parameters. This method is composed of three parts: (1) construct a BP neural network structure, determine the coefficients of the neural network including the weight and threshold of each node as the parameters set of genetic algorithm;(2) obtain data of breathing motion and select data to traine GA-BP neural network, it is very important to select proper train data set in neural network forecast, generally neural network training require of the selected samples must be more than a cycle and can reflect the basic characteristics of the overall data, optimize the parameters of the BP neural network with the selected training data set;(3) when get satisfy optimization parameters, then we can use the optimized BP neural network to predict breathing motion. Note that breathing motion has the very strong specificity, so the trained neural network can only be sure that specific individual, who provide the training sample set,has a good result. The data source is obtained by way of video capture marks on the chest wall with breathing movement displacement of video detection, tracking and record the marker coordinate data of movement; because of incorporation noise in the video target detection, tracking and recording, in order to obtain good optimization, training and prediction result, the coordinate data of markers were smoothed preprocessing by Matlab7.6. With the purpose of evaluating the prediction accuracy of breathing motion predict, this paper analyzes MRS (mean root square error) and MAPE(mean absolute percentage error) of the prediction of markers’trajectory to evaluate the precision of prediction algorithms. At the same time in order to show the validity of the adopted algorithm in this paper, the same data used in the existing prediction method including the BP neural network, and linear forecasting, MRS(most recent sample) method with the same evaluation standard, the experimental results reveal that the proposed method for different delay time effectively reduce forecasting errors. In addition, we also try the method of multi-marks on chest wall to predict breathing motion, the soruce of data collected by the same way, in the chest wall place a estimated marker and three referenced points, detecting and tracking these four points motion, using the data of three referenced points in genetic algorithm to optimize the BP neural network. The results of experiments reveal that multi-marks prediction accuracy is higher than single one mark prediction accuracy. |
PDF Full Text Request