基于粒子群算法与反向传播神经网络的呼吸运动预测研究

在放射治疗过程中,呼吸运动会造成某些器官组织如肺、肝的靶区发生变化,从而降低放疗的效果,并且加大对正常组织器官的伤害。因此,在放疗过程中对靶区进行呼吸运动的实时估计是一项非常必要的工作。由于具备较好的非线性拟合能力,优化反向传播神经网络(BP-NN)已经被广泛应用于呼吸的预测,然而BP-NN容易陷入局部最优值。提出一种应用粒子群算法(PSO)优化BP-NN的方法减少陷入局部最优值的机率,提高呼吸运动预测的精度。首先,应用PSO算法寻找神经网络的最佳初始权值与阈值;然后,应用最优的初始权值与阈值建立神经网络(PSO-NN);最后,利用建立的PSO-NN网络进行呼吸预测。结果表明,11组肺癌病人呼吸运动预测实验对比结果表明,此算法(PSO-NN)相比单纯应用BP-NN算法的平均绝对误差由0.24减少到0.18(25%),互相关系数由0.82提高到0.86。所提出的算法可以有效地减少BP-NN陷入局部最优值的机率,提高预测的精度。     

Quantifying the effect of intrafraction motion during breast IMRT planning and dose delivery

Respiratory motion during intensity modulated radiation therapy (IMRT) causes two types of problems. First, the clinical target volume (CTV) to planning target volume (PTV) margin needed to account for respiratory motion means that the lung and heart dose is higher than would occur in the absence of such motion. Second, because respiratory motion is not synchronized with multileaf collimator (MLC) motion, the delivered dose is not the same as the planned dose. The aims of this work were to evaluate these problems to determine (a) the effects of respiratory motion and setup error during breast IMRT treatment planning, (b) the effects of the interplay between respiratory motion and multileaf collimator (MLC) motion during breast IMRT delivery, and (c) the potential benefits of breast IMRT using breath-hold, respiratory gated, and 4D techniques. Seven early stage breast cancer patient data sets were planned for IMRT delivered with a dynamic MLC (DMLC). For each patient case, eight IMRT plans with varying respiratory motion magnitudes and setup errors (and hence CTV to PTV margins) were created. The effects of respiratory motion and setup error on the treatment plan were determined by comparing the eight dose distributions. For each fraction of these plans, the effect of the interplay between respiratory motion and MLC motion during IMRT delivery was simulated by superimposing the respiratory trace on the planned DMLC leaf motion, facilitating comparisons between the planned and expected dose distributions. When considering respiratory motion in the CTV-PTV expansion during breast IMRT planning, our results show that PTV dose heterogeneity increases with respiratory motion. Lung and heart doses also increase with respiratory motion. Due to the interplay between respiratory motion and MLC motion during IMRT delivery, the planned and expected dose distributions differ. This difference increases with respiratory motion. The expected dose varies from fraction to fraction. However, for the seven patients studied and respiratory trace used, for no breathing, shallow breathing, and normal breathing, there were no statistically significant differences between the planned and expected dose distributions. Thus, for breast IMRT, intrafraction motion degrades treatment plans predominantly by the necessary addition of a larger CTV to PTV margin than would be required in the absence of such motion. This motion can be limited by breath-hold, respiratory gated, or 4D techniques.     

Automatic detection of erthemato-squamous diseases using adaptive neuro- fuzzy inference systems

A new approach based on adaptive neuro-fuzzy inference system (ANFIS) was presented for detection of erythemato-squamous diseases. The domain contained records of patients with known diagnosis. Given a training set of such records, the ANFIS classifiers learned how to differentiate a new case in the domain. The six ANFIS classifiers were used to detect the six erythemato-squamous diseases when 34 features defining six disease indications were used as inputs. To improve diagnostic accuracy, the seventh ANFIS classifier (combining ANFIS) was trained using the outputs of the six ANFIS classifiers as input data. The proposed ANFIS model combined the neural network adaptive capabilities and the fuzzy logic qualitative approach. Some conclusions concerning the impacts of features on the detection of erythemato-squamous diseases were obtained through analysis of the ANFIS. The performances of the ANFIS model were evaluated in terms of training performances and classification accuracies and the results confirmed that the proposed ANFIS model has some potential in detecting the erythemato-squamous diseases. The ANFIS model achieved accuracy rates which were higher than that of the stand-alone neural network model.     

Immobilization effect of air-injected blanket (AIB) for abdomen fixation

A new device for reducing the amplitude of breathing motion by pressing a patient's abdomen using an air-injected blanket (AIB) for external beam radiation treatments has been designed and tested. The blanket has two layers sealed in all four sides similar to an empty pillow made of urethane. The blanket is spread over the patient's abdomen with both ends of the blanket fixed to the sides of the treatment couch or a baseboard. The inner side, or patient side, of the blanket is thinner and expands more than the outer side. When inflated, the blanket balloons and effectively puts an even pressure on the patient's abdomen. Fluoroscopic observation was performed to verify the usefulness of AIB for patients with lung, breast cancer, or abdominal cancers. Internal organ movement due to breathing was monitored and measured with and without AIB. With the help of AIB, the average range of diaphragm motion was reduced from 2.6 to 0.7 cm in the anterior-to-posterior direction and from 2.7 to 1.3 cm in the superior-to-inferior direction. The motion range in the right-to-left direction was negligible, for it was less than 0.5 cm. These initial testing demonstrated that AIB is useful for reducing patients' breathing motion in the thoracic and abdominal regions comfortably and consistently.     

The effects of intra-fraction organ motion on the delivery of intensity-modulated field with a multileaf collimator

Intensity-modulated radiation therapy can be conveniently delivered with a multileaf collimator. With this method, the entire field is not delivered at once, but rather it is composed of many subfields defined by the leaf positions as a function of beam on time. At any given instant, only these subfields are delivered. During treatment, if the organ moves, part of the volume may move in or out of these subfields. Due to this interplay between organ motion and leaf motion the delivered dose may be different from what was planned. In this work, we present a method that calculates the effects of organ motion on delivered dose. The direction of organ motion may be parallel or perpendicular to the leaf motion, and the effect can be calculated for a single fraction or for multiple fractions. Three breast patients and four lung patients were included in this study,with the amplitude of the organ motion varying from +/- 3.5 mm to +/- 10 mm, and the period varying from 4 to 8 seconds. Calculations were made for these patients with and without organ motion, and results were examined in terms of isodose distribution and dose volume histograms. Each calculation was repeated ten times in order to estimate the statistical uncertainties. For selected patients, calculations were also made with conventional treatment technique. The effects of organ motion on conventional techniques were compared relative to that on IMRT techniques. For breast treatment, the effect of organ motion primarily broadened the penumbra at the posterior field edge. The dose in the rest of the treatment volume was not significantly affected. For lung treatment, the effect also broadened the penumbra and degraded the coverage of the planning target volume (PTV). However, the coverage of the clinical target volume (CTV) was not much affected, provided the PTV margin was adequate. The same effects were observed for both IMRT and conventional treatment techniques. For the IMRT technique, the standard deviations of ten samples of a 30-fraction calculation were very small for all patients, implying that over a typical treatment course of 30 fractions, the delivered dose was very close to the expected value. Hence, under typical clinical conditions, the effect of organ motion on delivered dose can be calculated without considering the interplay between the organ motion and the leaf motion. It can be calculated as the weighted average of the dose distribution without organ motion with the distribution of organ motion. Since the effects of organ motion on dose were comparable for both IMRT and conventional techniques, the PTV margin should remain the same for both techniques.     

Real-time prediction of respiratory motion based on local regression methods

Recent developments in modulation techniques enable conformal delivery of radiation doses to small, localized target volumes. One of the challenges in using these techniques is real-time tracking and predicting target motion, which is necessary to accommodate system latencies. For image-guided-radiotherapy systems, it is also desirable to minimize sampling rates to reduce imaging dose. This study focuses on predicting respiratory motion, which can significantly affect lung tumours. Predicting respiratory motion in real-time is challenging, due to the complexity of breathing patterns and the many sources of variability. We propose a prediction method based on local regression. There are three major ingredients of this approach: (1) forming an augmented state space to capture system dynamics, (2) local regression in the augmented space to train the predictor from previous observation data using semi-periodicity of respiratory motion, (3) local weighting adjustment to incorporate fading temporal correlations. To evaluate prediction accuracy, we computed the root mean square error between predicted tumor motion and its observed location for ten patients. For comparison, we also investigated commonly used predictive methods, namely linear prediction, neural networks and Kalman filtering to the same data. The proposed method reduced the prediction error for all imaging rates and latency lengths, particularly for long prediction lengths.     

A review on the clinical implementation of respiratory-gated radiation therapy

Respiratory-gated treatment techniques have been introduced into the radiation oncology practice to manage target or organ motions. This paper will review the implementation of this type of gated treatment technique where the respiratory cycle is determined using an external marker. The external marker device is placed on the abdominal region between the xyphoid process and the umbilicus of the patient. An infrared camera tracks the motion of the marker to generate a surrogate for the respiratory cycle. The relationship, if any, between the respiratory cycle and the movement of the target can be complex. The four-dimensional computed tomography (4DCT) scanner is used to identify this motion for those patients that meet three requirements for the successful implementation of respiratory-gated treatment technique for radiation therapy. These requirements are (a) the respiratory cycle must be periodic and maintained during treatment, (b) the movement of the target must be related to the respiratory cycle, and (c) the gating window can be set sufficiently large to minimise the overall treatment time or increase the duty cycle and yet small enough to be within the gate. If the respiratory-gated treatment technique is employed, the end-expiration image set is typically used for treatment planning purposes because this image set represents the phase of the respiratory cycle where the anatomical movement is often the least for the longest time. Contouring should account for tumour residual motion, setup uncertainty, and also allow for deviation from the expected respiratory cycle during treatment. Respiratory-gated intensity-modulated radiation therapy (IMRT) treatment plans must also be validated prior to treatment. Quality assurance should be performed to check for positional changes and the output in association with the motion-gated technique. To avoid potential treatment errors, radiation therapist (radiographer) should be regularly in-serviced and made aware of the need to invoke the gating feature when prescribed for selected patients.     

An optical flow based method for improved reconstruction of 4D CT data sets acquired during free breathing

Respiratory motion degrades anatomic position reproducibility and leads to issues affecting image acquisition, treatment planning, and radiation delivery. Four-dimensional (4D) computer tomography (CT) image acquisition can be used to measure the impact of organ motion and to explicitly account for respiratory motion during treatment planning and radiation delivery. Modern CT scanners can only scan a limited region of the body simultaneously and patients have to be scanned in segments consisting of multiple slices. A respiratory signal (spirometer signal or surface tracking) is used to reconstruct a 4D data set by sorting the CT scans according to the couch position and signal coherence with predefined respiratory phases. But artifacts can occur if there are no acquired data segments for exactly the same respiratory state for all couch positions. These artifacts are caused by device-dependent limitations of gantry rotation, image reconstruction times and by the variability of the patient's respiratory pattern. In this paper an optical flow based method for improved reconstruction of 4D CT data sets from multislice CT scans is presented. The optical flow between scans at neighboring respiratory states is estimated by a non-linear registration method. The calculated velocity field is then used to reconstruct a 4D CT data set by interpolating data at exactly the predefined respiratory phase. Our reconstruction method is compared with the usually used reconstruction based on amplitude sorting. The procedures described were applied to reconstruct 4D CT data sets for four cancer patients and a qualitative and quantitative evaluation of the optical flow based reconstruction method was performed. Evaluation results show a relevant reduction of reconstruction artifacts by our technique. The reconstructed 4D data sets were used to quantify organ displacements and to visualize the abdominothoracic organ motion.     

Application of adaptive neuro-fuzzy inference system for epileptic seizure detection using wavelet feature extraction

Intelligent computing tools such as artificial neural network (ANN) and fuzzy logic approaches are demonstrated to be competent when applied individually to a variety of problems. Recently, there has been a growing interest in combining both these approaches, and as a result, neuro-fuzzy computing techniques have been evolved. In this study, a new approach based on an adaptive neuro-fuzzy inference system (ANFIS) was presented for epileptic seizure detection. The proposed ANFIS model combined the neural network adaptive capabilities and the fuzzy logic qualitative approach. Decision making was performed in two stages: feature extraction using the wavelet transform (WT) and the ANFIS trained with the backpropagation gradient descent method in combination with the least squares method. Some conclusions concerning the impacts of features on the detection of epileptic seizures were obtained through analysis of the ANFIS. The results are highly promising, and a comparative analysis suggests that the proposed modeling approach outperforms ANN model in terms of training performances and classification accuracies. The results confirmed that the proposed ANFIS model has some potential in epileptic seizure detection. The ANFIS model achieved accuracy rates which were higher than that of the stand-alone neural network model.     

Similarity classifier with generalized mean applied to medical data

A new approach based on fuzzy similarity was presented for the detection of erythemato-squamous diseases, diabetes, liver disorders, breast cancer and thyroid. The domain contained records of patients with known diagnoses. The results were very promising with all data sets and some conclusions can be drawn that a fuzzy similarity model can be used for the diagnosis of patients taking into consideration the error rate. A fuzzy similarity classifier was used to detect the six erythemato-squamous diseases when 34 features defining six disease indications were used as inputs. The results confirmed that the proposed model has potential in detecting erythemato-squamous diseases. The fuzzy similarity model achieved accuracy rates (over 97%) which were higher than that of the stand-alone neural network model or the ANFIS model suggested in [E.D. Ubeyli, I. Güler, Comput. Biol. Med. 35(5) (2005) 421-433]. With PIMA Indian diabetes, the detection model has an error rate of about 25% which is much better than the overall rate of 33% for diabetes. The model was also tested with other data sets: thyroid and two breast cancer data sets where the average detection accuracy was over 96% for all cases, which is quite good. Also, the liver disorder data set gave promising results.     

The application of the sinusoidal model to lung cancer patient respiratory motion

Accurate modeling of the respiratory cycle is important to account for the effect of organ motion on dose calculation for lung cancer patients. The aim of this study is to evaluate the accuracy of a respiratory model for lung cancer patients. Lujan et al. [Med. Phys. 26(5), 715-720 (1999)] proposed a model, which became widely used, to describe organ motion due to respiration. This model assumes that the parameters do not vary between and within breathing cycles. In this study, first, the correlation of respiratory motion traces with the model f(t) as a function of the parameter n (n = 1, 2, 3) was undertaken for each breathing cycle from 331 four-minute respiratory traces acquired from 24 lung cancer patients using three breathing types: free breathing, audio instruction, and audio-visual biofeedback. Because cos2 and cos4 had similar correlation coefficients, and cos2 and cos1 have a trigonometric relationship, for simplicity, the cos1 value was consequently used for further analysis in which the variations in mean position (z0), amplitude of motion (b) and period (tau) with and without biofeedback or instructions were investigated. For all breathing types, the parameter values, mean position (z0), amplitude of motion (b), and period (tau) exhibited significant cycle-to-cycle variations. Audio-visual biofeedback showed the least variations for all three parameters (z0, b, and tau). It was found that mean position (z0) could be approximated with a normal distribution, and the amplitude of motion (b) and period (tau) could be approximated with log normal distributions. The overall probability density function (pdf) of f(t) for each of the three breathing types was fitted with three models: normal, bimodal, and the pdf of a simple harmonic oscillator. It was found that the normal and the bimodal models represented the overall respiratory motion pdfs with correlation values from 0.95 to 0.99, whereas the range of the simple harmonic oscillator pdf correlation values was 0.71 to 0.81. This study demonstrates that the pdfs of mean position (z0), amplitude of motion (b), and period (tau) can be used for sampling to obtain more realistic respiratory traces. The overall standard deviations of respiratory motion were 0.48, 0.57, and 0.55 cm for free breathing, audio instruction, and audio-visual biofeedback, respectively.     

A monoscopic method for real-time tumour tracking using combined occasional x-ray imaging and continuous respiratory monitoring

Three major linear accelerator vendors offer gantry-mounted single (monoscopic) x-ray imagers. The use of monoscopic imaging to estimate three-dimensional (3D) target positions has not been fully explored. The purpose of this work is to develop and investigate a robust monoscopic method for real-time tumour tracking, combining occasional x-ray imaging and continuous external respiratory monitoring, and compare this with an established stereoscopic method. Monoscopic estimation of 3D target positions is a two-step procedure. Step (1) is similar to the stereoscopic approach using combined occasional x-ray imaging and real-time external respiratory monitoring, i.e. to establish the correlation between the target coordinates T(x, y, z) and the external respiratory signal (R) (sECM: stereoscopic external correlation model). However, in monoscopic estimation, the correlation between the two coordinates (xp, yp) projected on the imager plane and the external respiratory signal (mECM: monoscopic external correlation model) is established. With only a single projection, the component of the 3D target position, which is along the x-ray imaging direction, is unresolved. Therefore, step (2) is used to estimate the unresolved component (z( parallel)) by building a correlation model between the unresolved component and the two other components projected on the imager (ICM: internal correlation model) with a prior 3D target trajectory that may be obtained by 4DCT, MV/kV imaging or 4DCBCT. At the time of prediction, (xp, yp) are estimated from (R) using the correlation model in step (1), and then z( parallel) is estimated from the estimated (xp, yp) using the correlation model in step (2). The performance of the proposed method was evaluated under various model update intervals and compared with the stereoscopic estimation method using 160 tumour trajectory and external respiratory motion data recorded at 25 Hz from 46 thoracic and abdominal cancer patients who underwent hypofractionated stereotactic radiotherapy by a CyberKnife system. The precision of the input data used in this study to represent tumour motion was assessed using x-ray imaging to be 1.5 +/- 0.8 mm. Monoscopic imaging every 30/60 s with updating ICM every 120/180 s can estimate target positions with a 1 mm root-mean-square error (RMSE) for 63/53% or a 2 mm RMSE for 93/91%, respectively. In contrast, stereoscopic x-ray imaging every 30/60 s can estimate target motion within a 1 mm RMSE for 72/58% or a 2 mm RMSE for 95/92%, respectively. The overall 3D error of the monoscopic estimation is approximately 10% higher than comparable stereoscopic imaging methods when the period between imaging is 1 s or more, and 40% higher for continuous imaging. The promising result may be explained by the fact that superior/inferior motion-the major axis of tumour motion-is fully resolved even in the monoscopic view for coplanar treatments, and tumour motion in each dimension is relatively well correlated.     

A comparison between amplitude sorting and phase-angle sorting using external respiratory measurement for 4D CT

Respiratory motion can cause significant dose delivery errors in conformal radiation therapy for thoracic and upper abdominal tumors. Four-dimensional computed tomography (4D CT) has been proposed to provide the image data necessary to model tumor motion and consequently reduce these errors. The purpose of this work was to compare 4D CT reconstruction methods using amplitude sorting and phase angle sorting. A 16-slice CT scanner was operated in ciné mode to acquire 25 scans consecutively at each couch position through the thorax. The patient underwent synchronized external respiratory measurements. The scans were sorted into 12 phases based, respectively, on the amplitude and direction (inhalation or exhalation) or on the phase angle (0-360 degrees) of the external respiratory signal. With the assumption that lung motion is largely proportional to the measured respiratory amplitude, the variation in amplitude corresponds to the variation in motion for each phase. A smaller variation in amplitude would associate with an improved reconstructed image. Air content, defined as the amount of air within the lungs, bronchi, and trachea in a 16-slice CT segment and used by our group as a surrogate for internal motion, was correlated to the respiratory amplitude and phase angle throughout the lungs. For the 35 patients who underwent quiet breathing, images (similar to those used for treatment planning) and animations (used to display respiratory motion) generated using amplitude sorting displayed fewer reconstruction artifacts than those generated using phase angle sorting. The variations in respiratory amplitude were significantly smaller (P < 0.001) with amplitude sorting than those with phase angle sorting. The subdivision of the breathing cycle into more (finer) phases improved the consistency in respiratory amplitude for amplitude sorting, but not for phase angle sorting. For 33 of the 35 patients, the air content showed significantly improved (P < 0.001) correlation with the respiratory amplitude than with the phase angle, suggesting a stronger relationship between internal motion and amplitude. Overall, amplitude sorting performed better than phase angle sorting for 33 of the 35 patients and equally well for two patients who were immobilized with a stereotactic body frame and an abdominal compression plate.     

Review on 4D models for organ motion compensation

Minimal invasive tumor therapies are getting ever more sophisticated with novel treatment approaches and new devices allowing for improved targeting precision. Applying these effectively requires precise localization of the structures of interest. Vital processes, such as respiration and heartbeat, induce organ motion, which cannot be neglected during therapy. This review focuses on 4D organ models to compensate for respiratory motion during therapy. An overview is given on the effects of motion on the therapeutical outcome, methods required to capture and quantify respiratory motion, range of reported tumor motion, types of surrogates used when tumors are not directly observable, and methods for temporal prediction of surrogate motion. Organ motion models, which predict the location of structures of interest from surrogates measured during therapy, are discussed in detail.     

Dosimetric impact of geometric errors due to respiratory motion prediction on dynamic multileaf collimator-based four-dimensional radiation delivery

The synchronization of dynamic multileaf collimator (DMLC) response with respiratory motion is critical to ensure the accuracy of DMLC-based four dimensional (4D) radiation delivery. In practice, however, a finite time delay (response time) between the acquisition of tumor position and multileaf collimator response necessitates predictive models of respiratory tumor motion to synchronize radiation delivery. Predicting a complex process such as respiratory motion introduces geometric errors, which have been reported in several publications. However, the dosimetric effect of such errors on 4D radiation delivery has not yet been investigated. Thus, our aim in this work was to quantify the dosimetric effects of geometric error due to prediction under several different conditions. Conformal and intensity modulated radiation therapy (IMRT) plans for a lung patient were generated for anterior-posterior/posterior-anterior (AP/PA) beam arrangements at 6 and 18 MV energies to provide planned dose distributions. Respiratory motion data was obtained from 60 diaphragm-motion fluoroscopy recordings from five patients. A linear adaptive filter was employed to predict the tumor position. The geometric error of prediction was defined as the absolute difference between predicted and actual positions at each diaphragm position. Distributions of geometric error of prediction were obtained for all of the respiratory motion data. Planned dose distributions were then convolved with distributions for the geometric error of prediction to obtain convolved dose distributions. The dosimetric effect of such geometric errors was determined as a function of several variables: response time (0-0.6 s), beam energy (6/18 MV), treatment delivery (3D/4D), treatment type (conformal/IMRT), beam direction (AP/PA), and breathing training type (free breathing/audio instruction/visual feedback). Dose difference and distance-to-agreement analysis was employed to quantify results. Based on our data, the dosimetric impact of prediction (a) increased with response time, (b) was larger for 3D radiation therapy as compared with 4D radiation therapy, (c) was relatively insensitive to change in beam energy and beam direction, (d) was greater for IMRT distributions as compared with conformal distributions, (e) was smaller than the dosimetric impact of latency, and (f) was greatest for respiration motion with audio instructions, followed by visual feedback and free breathing. Geometric errors of prediction that occur during 4D radiation delivery introduce dosimetric errors that are dependent on several factors, such as response time, treatment-delivery type, and beam energy. Even for relatively small response times of 0.6 s into the future, dosimetric errors due to prediction could approach delivery errors when respiratory motion is not accounted for at all. To reduce the dosimetric impact, better predictive models and/or shorter response times are required.     

Adaptive neuro-fuzzy inference systems for analysis of internal carotid arterial Doppler signals

In this study, a new approach based on adaptive neuro-fuzzy inference system (ANFIS) was presented for detection of internal carotid artery stenosis and occlusion. The internal carotid arterial Doppler signals were recorded from 130 subjects that 45 of them suffered from internal carotid artery stenosis, 44 of them suffered from internal carotid artery occlusion and the rest of them were healthy subjects. The three ANFIS classifiers were used to detect internal carotid artery conditions (normal, stenosis and occlusion) when two features, resistivity and pulsatility indices, defining changes of internal carotid arterial Doppler waveforms were used as inputs. To improve diagnostic accuracy, the fourth ANFIS classifier (combining ANFIS) was trained using the outputs of the three ANFIS classifiers as input data. The proposed ANFIS model combined the neural network adaptive capabilities and the fuzzy logic qualitative approach. Some conclusions concerning the impacts of features on the detection of internal carotid artery stenosis and occlusion were obtained through analysis of the ANFIS. The performance of the ANFIS model was evaluated in terms of classification accuracies and the results confirmed that the proposed ANFIS classifiers have some potential in detecting the internal carotid artery stenosis and occlusion. The ANFIS model achieved accuracy rates which were higher than that of the stand-alone neural network model.     

The impact of respiratory motion and treatment technique on stereotactic body radiation therapy for liver cancer

Stereotactic body radiation therapy (SBRT), which delivers a much higher fractional dose than conventional treatment in only a few fractions, is an effective treatment for liver metastases. For patients who are treated under free-breathing conditions, however, respiration-induced tumor motion in the liver is a concern. Limited clinical information is available related to the impact of tumor motion and treatment technique on the dosimetric consequences. This study evaluated the dosimetric deviations between planned and delivered SBRT dose in the presence of tumor motion for three delivery techniques: three-dimensional conformal static beams (3DCRT), dynamic conformal arc (DARC), and intensity-modulated radiation therapy (IMRT). Five cases treated with SBRT for liver metastases were included in the study, with tumor motions ranging from 0.5 to 1.75 cm. For each case, three different treatment plans were developed using 3DCRT, DARC, and IMRT. The gantry/multileaf collimator (MLC) motion in the DARC plans and the MLC motion in the IMRT plans were synchronized to the patient's respiratory motion. Retrospectively sorted four-dimensional computed tomography image sets were used to determine patient-organ motion and to calculate the dose delivered during each respiratory phase. Deformable registration, using thin-plate-spline models, was performed to encode the tumor motion and deformation and to register the dose-per-phase to the reference phase images. The different dose distributions resulting from the different delivery techniques and motion ranges were compared to assess the effect of organ motion on dose delivery. Voxel dose variations occurred mostly in the high gradient regions, typically between the target volume and normal tissues, with a maximum variation up to 20%. The greatest CTV variation of all the plans was seen in the IMRT technique with the largest motion range (D99: -8.9%, D95: -8.3%, and D90: -6.3%). The greatest variation for all 3DCRT plans was less than 2% for D95. Dose variations for DARC fell between the 3DCRT and IMRT techniques. The dose volume histogram variations for normal organs were negligible. Therefore, the IMRT technique may be a preferable treatment choice in cases where the target volume and critical organs are in close proximity, or when normal organ protection is a high priority, provided that motion effect for the target volume can be managed.     

Internal-external correlation investigations of respiratory induced motion of lung tumors

In gated radiation therapy procedures, the lung tumor position is used directly (by implanted radiopaque markers) or indirectly (by external surrogate methods) to decrease the volume of irradiated healthy tissue. Due to a risk of pneumothorax, many clinics do not implant fiducials, and the gated treatment is primarily based on a respiratory induced external signal. The external surrogate method relies upon the assumption that the internal tumor motion is well correlated with the external respiratory induced motion, and that this correlation is constant in time. Using a set of data that contains synchronous internal and external motion traces, we have developed a dynamic data analysis technique to study the internal-external correlation, and to quantitatively estimate its underlying time behavior. The work presented here quantifies the time dependent behavior of the correlation between external respiratory signals and lung implanted fiducial motion. The corresponding amplitude mismatch is also reported for the lung patients studied. The information obtained can be used to improve the accuracy of tumor tracking. For the ten patients in this study, the SI internal-external motion is well correlated, with small time shifts and corresponding amplitude mismatches. Although the AP internal-external motion reveals larger time shifts than along the SI direction, the corresponding amplitude mismatches are below 5 mm.