Index of content:
Volume 35, Issue 6, June 2008
- Joint Imaging/Therapy Scientific Session: Room 351
- Advances in Cone Beam CT
TH‐D‐351‐07: Evaluation of Automatic Volume Match Function for Kilovoltage Cone‐Beam CT (CBCT) Guided Patient Setup35(2008); http://dx.doi.org/10.1118/1.2962929View Description Hide Description
Purpose: To evaluate the automatic volume match function provided by a commercial Cone‐beam CT(CBCT) guided patient setup system. Method and Materials:CBCTimages from 10 patients treated with stereotactic body radiotherapy for primary livercancers in the stereotactic frame were acquired by On‐board Imager (OBI, Varian Medical System) following initial treatment setup. Manual volume registration of CBCT to the planning CT was performed by physicians to adjust patient positioning. Retrospective automatic volume match was also performed for each dataset and compared to the manual registration. To further assess the automatic volume match, simulation of patient rotational offsets was generated. Automatic volume matching of the simulated data was used to investigate potential setup errors due to patient roll. Results: A total of 27 CBCT datasets were acquired and analyzed. The average differences between manual and automatic match are 2.4±1.7mm, 3.2±2.6mm and 1.8±1.6mm in the right‐left, superior‐inferior and anterior‐posterior directions, respectively. The simulation study demonstrated a significant limitation of the automatic 4 degrees of freedom (4DOF) correction mode with regard to the patient's rotational offset. This occurs because the 6 degrees of freedom (6DOF) registration algorithm is always internally applied, even in 4DOF shift calculation mode, although the two dimensions of pitch/roll offsets are not displayed. Our simulation revealed that this can cause patient setup error which becomes significant as the distance between the patient rotation axis and the treatment isocenter increases. Conclusion: Caution needs to be taken when using automatic volume match function for CBCT‐guided patient setup since the conventional treatment couch is only capable of translational and limited rotational shifts. Using the full degrees of freedom automatic volume match without caution can lead to significant error in patient treatment.
- Imaging for Therapy Assessment
TH‐D‐351‐01: High Temporal Resolution and Streak‐Free Four‐Dimensional Cone‐Beam Computed Tomography35(2008); http://dx.doi.org/10.1118/1.2962922View Description Hide Description
Purpose: Respiratory gated CBCT (e.g. 4D CBCT) with an on‐board imager has been introduced to track tumor motion for lungcancer patients. However, the image quality is comprised by aliasing artifacts and lower CNR, due to the limited number of projections. The temporal resolution of 4D CBCT is often chosen to be around 500∼1000ms to increase the sampling at each phase. In this paper, we present a method to simultaneously achieve high temporal resolution and streaking artifacts‐free images in 4D CBCT.Method and Materials: The enabling technique is an image reconstruction method called Prior Image Constrained Compressed Sensing (PICCS). By incorporating a prior image into the reconstruction, PICCS enables accurate image reconstruction from few projections. Home made physical motion phantoms were scanned on a Varian Trilogy system. Projection data were retrospectively gated based on the phase information obtained by synchronizing the ‘x‐ray on’ signal and the motion signal. Results: The gating window was chosen so that only one projection was selected in each respiratory cycle. Streak artifacts free 4D CBCTimages were achieved using PICCS from these 11–14 selected projections, while other reconstruction algorithms like FDK and conventional Compressed Sensing failed to generate meaningful images. Temporal information was well preserved in these images and the temporal resolution in these images was primarily limited by the detector read out speed which was about 100ms. Moving objects were segmented out from images at each phase and the motion trajectory was extracted. Agreement was observed between the extracted motion profile and the programmed motion profile. Conclusion: A new algorithm, PICCS, was proposed to mitigate undersampling streaking artifacts in 4D CBCT. High temporal resolution was achievable by using fewer projections. The trajectory of moving objects could be extracted from the high quality 4D CBCTimages.
TH‐D‐351‐02: A Novel Digital Tomosynthesis (DTS) Reconstruction Method Using Prior Information and a Deformation Model35(2008); http://dx.doi.org/10.1118/1.2962923View Description Hide Description
Purpose: Digital Tomosynthesis (DTS) is a quasi‐3D imaging technique which reconstructsimages from a limited angle of on‐board projections with significantly lower dose and shorter acquisition time than full cone‐beam CT(CBCT). However, DTS imagesreconstructed by the conventional filtered back projection method have low plane‐to‐plane resolution and can't provide full volumetric information for target localization. In this study, we developed a novel DTS reconstruction method using prior information and a deformation model to recover volumetric information. Method and Materials: A patient's previous CBCT or CT data were used as the prior information, and the new patient volume was considered as a deformation of the prior volume. The deformation fields were solved by minimizing bending energy and maintaining data fidelity. A nonlinear conjugate gradient method was used as the optimizer. The algorithm was tested using simulated projections of a Shepp‐Logan phantom, liver and head‐and‐neck patient data. The accuracy of the reconstruction was evaluated by comparing both pixel value and contour differences between DTS and CBCTimages.Results: In the liver patient study, the systematic and random errors for the live contour reconstructed using a 60‐degree scan angle were 0.5 and 1.6mm, respectively, showing the new organ volume was accurately reconstructed. The pixel signal‐to‐noise ratio (PSNR) for 60‐degree DTS reaches 23.5dB. In the head‐and‐neck patient study, the method using 60‐degree scan was able to reconstruct the 8.1 degree rotation of the bony structure with 0.0 degree error. The PSNR for 60‐degree DTS reaches 24.2dB. Conclusion: A novel reconstruction method was developed to reconstruct DTS images using prior information and a deformation model. Volumetric information was accurately obtained using a 60‐degree scan angle. Preliminary validation of the algorithm showed that it is both technically and clinically feasible for image‐guidance in radiation therapy.
Partially supported by Grants from NIH and Varian Medical Systems.
35(2008); http://dx.doi.org/10.1118/1.2962924View Description Hide Description
Purpose: In order to track tumor motion for patients with lungcancer four‐dimensional CBCT (4D CBCT) techniques have been introduced to deal with respiratory motion by gating projections into several phases. However, due to the limited gantry rotation speed, fewer than 100 projections are available for the image reconstruction at each phase. Thus, severe undersampling streaking artifacts plague 4D CBCTimages. In this presentation, we propose a simple scheme to significantly reduce the streaking artifacts. Method and Materials: A prior image is first reconstructed using all of the cone‐beam projection data without gating. This image volume is then reprojected to generate a synthesized projection data set. The difference projections generated from these two data sets are then gated and reconstructed to generate a difference image for each respiratory phase. The difference image is added back to the prior image to generate the final 4D CBCTimage volume for each phase. A home‐made motion phantom was built and scanned on a Varian Trilogy system. Projection data were retrospectively gated based on the phase information. Results: For a given phase, only ∼12 projections were selected (i.e. one projection for each respiratory cycle) to reconstruct the 4D CBCTimages. The fidelity of stationary objects was preserved and descent reconstructions of moving objects were obtained. Streak artifacts were significantly reduced in the reconstructed images. A figure of merit to characterize the streak artifacts strength was introduced and about 70% streak artifacts reduction was achievable compared with traditional FDK reconstruction.Conclusion: An algorithm has been proposed to reduce undersampling streaking artifacts in 4D CBCT. In physical phantom studies, we demonstrated that the streaking artifacts were effectively mitigated (70% reduction compared with FDK reconstruction). This correction scheme enables gating of the 4D CBCT data in a very narrow window (12∼95ms) which significantly improves the temporal resolution.
TH‐D‐351‐04: Cone Beam CT Beam Hardening and Scatter Preprocessing for Improved Image Quality in Image‐Guided Adaptive Radiation Therapy35(2008); http://dx.doi.org/10.1118/1.2962926View Description Hide Description
Purpose: To improve cone beam CT(CBCT)image quality in terms of CT number uniformity and accuracy to better support key image‐guidedadaptive radiation therapy tasks such as intensity‐driven deformable image registration and adaptive treatment planning.Method and Materials: Subtractive scatter corrections to Varian on‐board imager (OBI) CBCT projections utilized Monte Carlo simulated scatter profiles based upon the known phantom geometry. A first‐order water linearization correction was developed based upon the measured x‐ray beam central ray spectrum with mathematical spectral correction for the bowtie filter thickness associated with each detector pixel. Projections corrected for scatter and beam hardening and further corrected using Varian's scatter normalization phantom, were reconstructed with an in‐house FDK reconstruction engine. The preprocessing algorithm was applied to both half‐fan and full‐fan projection sets with and without the bowtie filter. Results:CBCTimages corrected with the described preprocessing method showed improved CT number uniformity and reduction of the notorious CBCT cupping artifact. For a 20 cm diameter water phantom, the cupping artifact was reduced from about 15% to within 2%. Conclusion: Model‐based scatter and beam‐hardening preprocessing procedures improve the on‐board CBCTimage quality, improving its utility for image‐guidedadaptive radiation therapy.Conflict of Interest: This work was supported by NIH P01 CA116602 and a grant from Varian Medical Systems.
35(2008); http://dx.doi.org/10.1118/1.2962927View Description Hide Description
Purpose: Cone‐beam CT(CBCT) is being increasingly used in radiation therapy. However, as compared to conventional CT, the degraded image quality of CBCT hampers its applications. Due to the large volume of x‐ray illumination in CBCT, scatter is considered as one of the fundamental limitations of CBCTimage quality. Many scatter correction algorithms have been developed, while drawbacks still exist. Here, we propose a new scatter correction method which is particularly useful in radiation therapy.Method and Materials: Since the same patient is scanned repetitively during one radiation treatment course, we measure the scatter distributions in one scan, and use the measured scatter distributions to estimate and correct scatter in the following scans. A partially blocked CBCT is used in the scatter measurement scan. The x‐ray beam blocker has a strip pattern, such that the whole‐field scatter distribution can be estimated from the detected signals in the shadow region and the patient rigid transformation can be determined from the reconstructed image using the illuminated detector projection data. From the derived patient transformation, the measured scatter is then modified and used for scatter correction in the following regular CBCT scans. Results: The proposed method has been evaluated using Monte Carlo simulations and physical experiments on an anthropomorphic chest phantom. The results show a significant suppression of scatter artifacts using the proposed method. Using the reconstruction in a narrow collimator geometry as a reference, the comparison also shows that the proposed method reduces reconstruction error from 13.2% to 0.8%. Conclusion: This work indicates that much improved CBCTimage quality is achievable using the proposed scatter correction method in radiation therapy. Our method is very attractive in applications where high CBCT reconstruction accuracy is critical, for example, dose calculation in adaptive radiation therapy.
TH‐D‐351‐06: Comparison Between 2D Monte Carlo Modeled and Experimental Cone‐Beam CT X‐Ray Projections35(2008); http://dx.doi.org/10.1118/1.2962928View Description Hide Description
Purpose: Fast and accurate modeling of cone‐beam CT(CBCT) x‐ray projection data can improve cone‐beam CT(CBCT)image quality either by conditioning projection data prior to image reconstruction or by supporting rigorous comparative simulation studies of competing image reconstruction and processing algorithms. In this study, we compare Monte Carlo‐ computed x‐ray projections with projections experimentally acquired from our Varian Trilogy CBCTimaging system for phantoms of known design. Method and Materials: Our recently developed Monte Carlo photon‐transport code, PTRAN, was used to compute primary and scatter projections for cylindrical phantoms of known diameter (CatPhan and NA model 76‐410) with and without bow‐tie filter and antiscatter grid for both full‐ and half‐fan geometries. The simulations were based upon measured 120 kVp spectra, beam profiles, and flat‐panel detector (4030CB) point‐spread functions. The beam‐stop array method was used to acquire scatter and SPR distributions from the OBI images. The biasing of scatter measurements due to the long detector PSF tails was corrected either by a lead mask or by deconvolution. Computed projections were compared to flat‐ and dark‐field corrected 4030CB images.Results: The simulated primary profiles agree with experiment within 3%, while the simulated scatter profiles agree within 8–10%. Both PSF measurements and mask measurements indicate that scatter radiation values can be biased by as much as 7% detector PSF tails. Conclusion: In agreement with the literature, the difference between simulated and measured projection data is of the order of 6–8%. Higher accuracy can be achieved mainly by improving the beam modeling and correcting the non linearities induced by the detector PSF.
This project was supported in part by grants from Varian Medical Systems and NCI (R01 CA 75371).
- Target Localization for IGRT
TU‐C‐351‐01: Design of Magnetic Resonance Imaging Protocol for Accelerated Partial Breast Irradiation in Prone Position35(2008); http://dx.doi.org/10.1118/1.2962455View Description Hide Description
Purpose: Accurate delineation of tumor bed is prerequisite for an effective treatment of breast cancer with accelerated partial breast irradiation (APBI) technique. Magnetic resonance imaging(MRI) has high contrast mechanisms that could enable precise contouring of the lumpectomy cavity. We investigated practical issues pertinent to MR‐guided breast radiation therapy using customized pulse sequences. Method and Materials: Unilateral breast MRimages were obtained from two healthy volunteers in prone position on a dedicated coil. Radiation therapists positioned the volunteers in reproducible setup using a waterproof marker. We acquired 6‐min 3D MRimages to analyze setup uncertainty and fast 2D MRimages of 0.5‐s temporal resolution to detect respiratory motion. An accurate 3D warping‐correction algorithm was evaluated and used to restore spatial fidelity from geometric distortions due to MR gradient non‐linearities. MRimages of surgical clips as fiducial were obtained using pulse sequences designed to provide high signal from lumpectomy cavity. Results: Average breathing motion in the breast was found to be less than 0.5 mm in prone position. Setup deviations of 1 cm were observed between series of intended prone positioning. Rigid image registration based on local anatomy structure resulted in a residual error which was up to 5 mm at the peripheral region of the breast. Gradient‐induced non‐linearity led to 1 cm distortion in the uncorrected phantom image at the region 15 cm away from the main magnet axis. The 3D correction algorithm reduced the deviation to < 1 mm. We observed 4 mm signal void artifacts of surgical clips. Conclusion: We obtained MRimages of suitable quality from volunteers using pulse sequences customized for lumpectomy site identification. MR‐guided APBI in prone position would require 5 mm uniform expansion of clinical target volume with the employment of an appropriate on‐board imaging technique.
35(2008); http://dx.doi.org/10.1118/1.2962460View Description Hide Description
Purpose:USimage guidance systems which are calibrated for soft tissue applications will introduce errors in depth‐from‐transducer representation when used in media with a different speed of sound propagation (e.g. fat). In this study we utilized a standard US phantom to demonstrate the existence of the speed artifact when using a commercial USimage guidance system to image through simulated layers of body fat, and compared the results with calculated/predicted values. Method and Materials A General Purpose Ultrasound Phantom (speed of sound‐SOS = 1540 m/s) Model ATL 040 (CIRS, Norfolk, VA) was imaged on a LightSpeed RT CT scanner (GE Health Care, Waukesha, WI) at 0.625‐mm slice thickness and 0.5mm pixel. Target‐simulating wires inside the phantom were contoured and later transferred to the BAT Sxi unit (Nomos Corp., Cranberry, PA). Various thicknesses (1–8cm) of fat simulating material (SOS = 1435 m/s), manufactured by CIRS, were placed on top of the phantom to study the depth‐related alignment error. In order to demonstrate that the speed artifact is not caused by adding additional layers on top of the phantom, we repeated these measurements in an identical setup using tissue simulating material (SOS=1540 m/s) for the top layers. Result and Discussion: For the fat simulating material used in this study, we observed the magnitude of the depth‐related alignment errors resulting from this speed artifact to be 0.73 mm per cm of fat imaged‐through. The measured alignment errors caused by the speed artifact agreed with the calculated values within measurement uncertainties for the five different thicknesses of fat simulating material studied here. Conclusion: We demonstrated the depth‐related alignment error due to speed artifact when using USimage guidance for radiation treatment alignment. When possible, care should be taken to avoid imaging through a thick layer of fat for larger patients in US alignments.
35(2008); http://dx.doi.org/10.1118/1.2962462View Description Hide Description
Purpose: Recent studies of prostate intrafraction motion have been inconclusive. One cine‐MRI study demonstrated that pre‐treatment rectal filling status is a significant predictor of intrafraction motion magnitude. However, studies using the Calypso system show that pre‐ and post‐imaging is not an adequate test of prostate intrafraction motion. Our purpose is to determine the effectiveness of pre‐ and post‐imaging in predicting prostate intrafraction motion. Method and Materials: Pre‐ and post‐CBCT images and intrafraction kV fluoroscopy were used to determine the prostate position via fiducial markers of 15 patients over 279 fractions. For each fraction, rectal filling status was documented and bladder volume estimated on both the pre‐ and post‐CBCTs. Also documented was whether a visible change occurred in rectum shape between CBCTs. The ability of rectal and bladder filling to predict the measured intrafraction motion was evaluated using receiver operator characteristic (ROC) analysis. A model assuming a linear motion pattern during the treatment fraction was applied and the RMS error calculated. Results: Initial rectal filling status was a significant predictor of large intrafraction motion while effects from bladder filling during treatment were inconclusive. The sensitivity of detecting intrafraction motion above a given threshold using post‐imaging was improved when change in rectal filling status was taken into account. For detecting motion >5 mm, the sensitivity increased from 0.88 to 0.93 with a decline in specificity from 0.96 to 0.76. The rms‐error decreased from 2.4 mm to 1.4 mm when the linear model was used as opposed to assuming no intrafraction motion (p<0.0001). Conclusion: Pre‐ and post‐imaging is a significant predictor of intrafraction motion especially when rectal filling status is taken into account. A linear model is more accurate than assuming no intrafraction motion. Effects of bladder filling are inconclusive and require further study. Conflict of Interest: Partially supported by NIH Grant CA118037.
TU‐C‐351‐04: Cone Beam CT Acquisition During Volumetric Arc Radiotherapy Delivery: Correction for Induced Artifacts35(2008); http://dx.doi.org/10.1118/1.2962465View Description Hide Description
Purpose: To evaluate the feasibility of kV cone beam CT(CBCT) acquisition simultaneously with volumetric modulated arc therapy (VMAT) delivery, and to test a method to correct for degradation of image quality due to VMAT delivery.Method and Materials: A commercial CBCT system was modified to enable simultaneous CBCT acquisition with VMAT delivery.CBCT scans of an image quality phantom were acquired during VMAT delivery while varying the VMAT parameters. Dose rate, energy, and field size of the VMAT delivery, and phantom size, were varied to evaluate the effect on image quality. The mean and standard deviation of the signal in a known location was quantified both in the raw 2D projection images and also in the reconstructed 3D CBCTimages. A nonlinear filter was tested to remove structural artifact and noise. An analytical scatter correction model was developed and used to remove scatter generated by the VMAT beam. Results: Structural artifact was reduced in the CBCT projections with a nonlinear filter. Scatter generated from the VMAT delivery varied with field size and dose rate, and minimally with phantom size. An analytical scatter model was constructed based on the VMAT fluence (field area times dose rate) for each CBCT projection, and applied to reduce scatter per projection. Applying the model improved uniformity from 7.9% to 3.0% and improved the contrast to noise ratio from 0.97 to 1.84. Conclusion: Megavoltage scatter, and its per projection variation, is the largest component contributing to degradation of CBCTimage quality during VMAT delivery. The degradation was reduced with a scatter model based on the VMAT delivery. A secondary component was structural artifact related to the repetition rate of the megavoltage beam and the readout mechanism of the kV detector. Conflict of Interest: Research sponsored by Elekta, Inc.
TU‐C‐351‐05: Daily Alignment Results for In‐Room CT‐Guided Stereotactic Body Radiation Therapy for Lung Cancer35(2008); http://dx.doi.org/10.1118/1.2962468View Description Hide Description
Purpose: To assess daily bone alignment results and changes in soft tissuetumor position during hypofractionated, in‐room computed tomography (CT)‐guided stereotactic body radiation therapy(SBRT) of lungcancer.Method and Materials: Daily alignment results during SBRT were analyzed for 117 tumors in 112 patients. Patients received 40–50 Gy of SBRT in 4–5 fractions to the target using an integrated CT‐LINAC system. The free‐breathing CT scans acquired during treatment set‐up were retrospectively re‐aligned to match with each of the bony references and the gross tumor volume (GTV) defined on the reference CT by rigid registration, and the daily deviations were calculated. Results: The mean (±SD) three‐dimensional (3D) shift from the initial skin marks to the final bone‐aligned positions was 9.4 ± 5.7 mm. The mean daily GTV deviation from the bone position was 0.1 ± 3.8 mm in the anterior‐posterior (AP) direction, −0.01 ± 4.2 mm in the superior‐inferior (SI) direction, and 0.2 ± 2.5 mm in the lateral direction. A statistically significant trend (linear fit with R2>0.8) in the change in GTV position relative to the bone was observed in 15 (13%), 11 (9%), and 21 (18%) cases along the AP, SI, and lateral directions, respectively. There were no significant associations between the trends in GTV movement and clinical factors. A margin of 10 mm around the ITV covered the inter‐fractional organ motion errors in 96.4% of tumors in the AP direction, 100% of tumors in the SI direction, and 100% of tumors in the lateral direction. Conclusion: 3D bone alignment using daily in‐room CT‐guided SBRT has good accuracy. However, a substantial number of tumors showed trends in position changes over 4 or 5 days. An isotropic margin distance of 10 mm around the ITV was necessary for adequate coverage of inter‐fractional organ motion errors of all cases.
TU‐C‐351‐06: Comparison of Multiple 3D‐3D Anatomy‐Based Rigid Image Registration Methods for Prostate Patient Setup Before External‐Beam Radiotherapy35(2008); http://dx.doi.org/10.1118/1.2962470View Description Hide Description
Purpose: To assess the variability in translational setup corrections computed via several different 3D to 3D rigid registration methods and conditions using only the image intensity information. Method and Materials: Fifteen pairs of male pelvic CTs were rigidly registered using multiple registration metrics, methods, and image content. Similarity was measured using mutual information and mean‐squared‐difference. The registrations were iterated using two different search algorithms. The effect of image resolution was observed by downsampling in each slice by factors of 2 and 4. The effects of image content were observed by using the entire image and then using only bony landmarks in the similarity measures. The uncertainty associated with the choice of source and target images was revealed by switching the roles of the two registered images. For each image pair the translational shifts for all the registration methods and conditions were compiled and the standard deviation around the mean shifts was computed. Results: Reversibility errors ranged from 0 to 1.8 mm for the various algorithms and conditions. The radial standard deviation of the various methods and conditions was 0.8 mm when the entire image was registered and 0.3 mm when only bony landmarks were used. Downsampling by a factor of 2 improved robustness with only a small loss in precision. Mutual information failed in nearly all cases when the images were thresholded to isolate bony features. Conclusion: Rigid image registrations have an intrinsic uncertainty that depends on the algorithm, the resolution of the images, and the features used to establish rigid congruence. Registration uncertainty should be accounted for in planning margins and patient setup procedures. In the absence of an absolute ground truth test of registration accuracy, the standard deviation in the shifts calculated by several different methods provides a useful estimate of that uncertainty.
Supported in part by NIH P01CA116602.
35(2008); http://dx.doi.org/10.1118/1.2962473View Description Hide Description
Purpose: Existing 4D CT scans show artifacts in a large percentage of cases. Additional dose is also given to ensure redundancy in data to minimize artifacts. To solve this a Prospective Displacement and Velocity based (PDV CT) cine method for acquisition of 4D CTimages was developed. It is compared against retrospective methods for efficiency, dose, and accuracy. Effects of different model parameters are also illustrated. Method and Materials: With the PDV CT method, image acquisition is only performed if both displacement and velocity of the respiratory signal are within predetermined tolerances. Respiratory signals from a Varian Real Time Position Management (RPM) system for 24 patients (103 sessions) under free breathing conditions were used. To surmount system latency and enhance efficiency, a linear adaptive prediction algorithm was used. The root‐mean‐square of differences between displacements and velocities of the respiratory signal corresponding to subsequent images, were calculated in order to evaluate the accuracy of each method. Results: A reduction in patient dose during image acquisition between 22 – 50% was achieved depending on the parameters chosen. The mean root‐mean‐square difference show PDV CT produces similar results than retrospective displacement sorting in general, although differences ∼20% smaller was achieved for some parameters. Velocity RMS differences improved between 30% – 45% when compared to retrospective phase sorting. The efficiency in acquisition time compared to retrospective phase sorting varied from ∼10% for displacement and velocity tolerances of 1mm and 4mm/s, respectively, to 80% – 93% for 4mm and 4mm/s. Conclusion: The reduced variation in the displacement and velocity of the respiratory signal indicates that the PDV CT method described here, could be a valuable tool for reducing artifacts in 4D CTimages, and more importantly, substantial dose reduction to the patient, although the price may be acquisition time. Conflict‐of‐Interest: Research supported by 1P01CA116602.
TU‐C‐351‐08: Application of Principal Component Analysis for Marker‐Less Lung Tumor Tracking with Beam's‐Eye‐View EPID Images35(2008); http://dx.doi.org/10.1118/1.2962477View Description Hide Description
Purpose: Recent studies have shown the impact of beam's‐eye‐view (BEV) imaging during radiotherapy to monitor the tumor location. This information can be used for real‐time interventions, treatment setup and verification, adaptive radiotherapy and delivered dose calculation. Although the tumor‐lung tissue density contrast can be poor, we show that it can be sufficient for tracking without implanted markers. The pre‐treatment 4DCT information used to plan the treatment is employed to ascertain the tumor motion during the treatment.Method and Materials: Exit radiation is passively acquired during lungtumorradiotherapy by operating an EPID in cine mode. The cine‐EPID in‐treatment images and the pre‐treatment phase‐specific DRRs are registered and corrected for setup induced affine transformations. DRRs are made from each of the 10 phase bins from the 4DCT simulation scan. Principal Components Analysis is able to discriminate small differences between similar images by shifting the image space axes to cause significant differences. The multidimensional image space is generated using the collection of DRR and EPIDimages. Each image can be represented as a linear combination of the best principal components. Using the projection coefficients, a multidimensional distance between the images is computed to compare each in‐treatment image with the ten phase‐specific DRRs. The maximum similarity is found by computing the minimum distance between the images so the phase‐specific DRR and in‐treatment EPIDimage are matched. Results: We were able to attribute the phase information to at least 75 % of the acquired in‐treatment images. The methodology was tested on the DRRimages alone with 90 % phase recovery. Conclusion: We have developed a PCA‐based algorithm that enabled us to quantify the differences between the tumor motion within the planning 4DCT and the actual tumor motion during the treatment.Conflict of Interest: This work was partially supported by Varian Medical Systems, Inc.
35(2008); http://dx.doi.org/10.1118/1.2962481View Description Hide Description
Purpose: To develop algorithms that can track the invisible lungtumors by tracking relevant anatomic features in the fluoroscopic images based principle component analysis (PCA) and regression analysis.Method and Materials: For each patient 15 seconds of fluoroscopic images were taken before treatment and used as training dataset. A few regions‐of‐interest (ROIs) were manually selected in the first image frame that may contain anatomic features correlated with tumor motion. PCA was applied to reduce the dimensionality of each ROI to 3. Tumor positions were manually marked by an expert observer in all training images. Regression methods were applied to build the correlations between the tumor position and the PCA‐processed ROIs. Then the correlations were used to predict the tumor positions in the testingimages. Four regression methods were considered: 1‐degree and 2‐degree linear regression,artificial neural network (ANN), and support vector machine (SVM). Their accuracy was assessed by comparing the prediction results with the reference tumor locations manually determined by the expert observer. Results: 12 sequences of fluoroscopic images have been studied retrospectively. Results are reported in terms of mean and variation of localization error. Considering all regression methods, the mean localization errors (MLE) are smaller than 1 mm for most patients and in the worst case is still smaller than 2.5 mm. Variations are smaller than 1 mm most of the time, and the largest variation is about 9 mm. 1‐degree linear regression and ANN in general perform better than the other two methods. The other two methods tend to have overfitting problems. Conclusion: Based on PCA and regression analysis, we proposed a novel method that can track the invisible lungtumor in fluoroscopic images by tracking correlated anatomic features. The method has an accuracy of 1mm in most of cases and smaller than 3mm in the worst case.
TU‐C‐351‐10: Inference of Hysteretic Respiratory Tumour Motion From External Surrogates: A State Augmentation Approach35(2008); http://dx.doi.org/10.1118/1.2962485View Description Hide Description
Purpose: To infer internal respiration‐induced tumor motion from external surrogate. To systematically resolve mapping ambiguity caused by breathing hysteresis. Method and Materials: We propose a state‐augmentation approach to capture system dynamics. Concatenating real‐time surrogate observations with their time‐delayed records describes the state information in a higher‐dimensional state space. In such space, inhale and exhale “stages” are naturally separated due to the incorporated velocity contents. Any existing inference model migrates effortlessly into this framework. We illustrate the idea with simple polynomial inference models, and derive a closed‐form solution for optimal choice of model parameters. Choice of lag length is demonstrated empirically to be robust and may be chosen offline. This approach is tested on synchronized recordings of internal tumor trajectories and external fiducial marker readouts from eight lung patients (multiple fractions and readings) with a Mitsubishi real‐time radiotherapy (RTRT) system. Internal recording is obtained by fluoroscopic tracking of implanted 1.5mm‐diameter gold ball bearings around the tumor and external surrogates measure relative abdominal skin positions. Results: Examination of trajectories in the augmented state‐space suggests the existence of a consistent and unambiguous inference map. Empirical tests with clinical data show that using state augmentation decreases the 3D RMSE from 2.01mm to 1.74mm with the linear model and 1.93mm to 1.63 with the quadratic model. Paired student‐t tests with P‐values on the order of 10e‐13 indicate statistical significance of the improvement. Conclusion: We proposed a simple state‐augmentation approach to implicitly incorporate the hysteretic internal‐external response pattern into the estimation framework with any existing inference model. For the general class of correspondence models that are linear in their parameters, closed‐form solutions for the optimal parameter values and the error evaluations are derived. Tests with clinical data demonstrate statistically significant improvement over direct models.
This work is sponsored by NIH P01‐CA59827 and Barbour Scholarship.