Volume 36, Issue 6, June 2009
Index of content:
- Joint Imaging/Therapy Scientific Session: Ballroom C
- Interfraction Motion and Margin Assessment
TU‐D‐BRC‐01: Repeated Measures Analysis of Variance of Patient Specific Daily Margins to Assess Interfraction Motion for Cervical Cancer Patients Undergoing IMRT Using Daily CBCT Imaging36(2009); http://dx.doi.org/10.1118/1.3182376View Description Hide Description
Purpose: To analyze interfraction motion in cervical cancer patients based on repeated measures analysis of variance of patient‐specific daily uniform margins obtained using daily CBCTimaging.Methods and Materials: Ten cervical cancer patients (stage IB‐IVA) treated with IMRT were analyzed for this study. A CTV consisting of the regional lymph nodes, upper vagina, parametria, cervix and uterus was contoured on the planning CT and daily CBCTimages. Twenty‐five CBCT scans, corresponding to each fraction were planned and attempted for each patient. The acquired CBCTimages for all the fractions were rigidly registered to the planning CT with respect to bony anatomy. The initial planning CTV was then cast onto the registered daily CBCT and were modified to reflect changes due to organ motion and deformation. Patient‐specific daily margins covering 100% of daily CTV volumes were calculated. Repeated measures analysis of variance (ANOVA) was performed on the patient specific daily margins to quantify inter‐patient and intra‐patient variations in estimated margins. Mean values within the subjects were imputed for the missing fractions. F statistics was used to test the null hypothesis that day of the treatment or fraction number had no effect on daily margins. Results: Patient specific daily uniform margins were used to generate an ANOVA table. The data shows a grand mean of 15.3 mm among all patients and fractions which represents the estimate of the average population margin. In addition, the inter‐patient standard deviation in margins is 5.1 mm and intra‐patient standard deviation is 35.4 mm indicating that the variation between patients is significant. The F statistics (F = 0.654) was significant indicating that the day of the treatment or fraction number had no influence on margin. Conclusion: Margins varied substantially between patients indicating adaptive RT approaches are needed for the treatment of cervical cancer patients.
TU‐D‐BRC‐02: Determination of Skin Mark Based Patient Setup Errors Using OBI and CBCT for Head and Neck Patients and Investigation of the Dosimetric Impact of the Errors On IMRT Treatment36(2009); http://dx.doi.org/10.1118/1.3182377View Description Hide Description
Purpose: To determine the skin‐mark‐based patient setup error using On Board Imaging (OBI) and cone‐beam CT(CBCT) for head and neck patients, and to determine the CTV‐PTV margin necessary to compensate for this setup error for IMRTtreatments.Method and Materials: A total of 50 patients were analyzed for this study. All patients were first set up by aligning skin marks with lasers. Before the delivery of first treatment, a CBCTimage was acquired to determine the final treatment position. OBI images were acquired for the subsequent 4 fractions. This weekly image acquisition protocol was followed for all patients until treatments were completed. Shift data were measured for a total of 1064 OBI fractions and 313 CBCT fractions. These data were compared statistically. Correlations between the CBCT and OBI data and among the OBI data were analyzed. The top 10 patients showing the largest setup error were selected to investigate the dosimetric impact of setup error using a 5 mm CTV‐PTV margin. Results: The mean ± sd of the shifts measured by OBI in the SI, lateral and AP direction were 0.6±2.9, −1.6±3.5 and −0.7±2.9 mm and those measured by CBCT were 0.1±2.9, −1.9±3.3 and 0.3±3.3 mm respectively. The correlation indices between the CBCT and OBI were 0.43, 0.59, and 0.38, and those among the OBI were 0.46, 0.34, and 0.53 along SI, lateral and AP directions respectively. For 9 of the 10 patients investigated, at least 99.5% of CTV volume received more than 95% of the prescription dose after accounting for the setup errors. Conclusion: There is no significant difference between the patient shifts measured by OBI and CBCT. No correlation was found between the shifts for different fractions. A 5 mm CTV‐PTV margin was found to be adequate to account for the setup errors measured in this study.
TU‐D‐BRC‐03: Assessment of Interobserver and Intraobserver Surgical Cavity Contour Variability in Accelerated Partial Breast Irradiation Through the Use of a Representative Surgical Cavity Contour36(2009); http://dx.doi.org/10.1118/1.3182378View Description Hide Description
Purpose: To quantify interobserver and intraobserver variations in definition of the surgical cavity (SC) in the planning of accelerated partial breast irradiation (APBI). Method and Materials: Eight prospectively accrued patients underwent four CT scans each: one planning CT and three repeat CTs during treatment. Three radiation oncologists contoured the SC on each scan for all patients and repeated the contours two times on each scan for three patients. Analysis of contour variations was performed by combining the contours to create a representative surgical cavity (RSC); the volume and spatial extent of each contour was compared with the RSC. Agreement of the delineated volumes was quantified by comparing the volume enclosed by all contours with the volume common to all contours. The standard deviation at each point where the RSC was calculated was also used to assess spatial variations. Fields from the original treatment plans were applied to the repeat CTs and dose distributions in the SCs were evaluated using the equivalent uniform dose approach. Results: The average interobserver volume difference was larger than the intraobserver difference (3.09 versus 0.86 cm3). The average agreement of the delineated volumes was 15.4% better in intraobserver than in interobserver comparisons. Likewise, the average interobserver spatial differences were larger than the intraobserver differences (0.34 versus 0.11 cm in the superior‐inferior direction, 0.37 versus 0.23 cm in the anterior‐posterior direction and 0.62 versus 0.28 cm in the left‐right direction). RSC standard deviations tended to be larger for interobserver rather than intraobserver variations. Despite interobserver variations, margins used for planning appeared sufficient to achieve clinically acceptable coverage of the SC. Conclusion: Interobserver variations are most responsible for SC contour variability in APBI. Future studies investigating changes in shape or position of the SC should focus more effort accounting for interobserver rather than intraobserver variability.
TU‐D‐BRC‐04: Evaluation of Positioning and Margin Selection in Image‐Guided Whole‐Pelvis Prostate IMRT36(2009); http://dx.doi.org/10.1118/1.3182379View Description Hide Description
Purpose: To evaluate the interplay of margin selection and alignment strategies on image‐guided whole pelvis prostate IMRT using serial CTimage studies and deformable image registration.Method and Materials: Forty‐two CT scans of 7 patients are studied — each patient has one initial and 5 serial CT scans. Three PTVs are created on the initial CT scan using three different margins ranging from 3 to 13mm. Each PTV is used to create a unique IMRT plan, which is registered to the serial CT scans, by bony alignment or by prostate‐based alignment. The dose computed on the 5 serial CTimage sets is accumulated onto the initial CT using deformable image registration with mesh method (Pinnacle 8.1x). Results: Prostate‐based alignment introduces statistically significant increase of dose to rectum, prostate and seminal vesicles (SV), and reduction of dose to bladder, for all the three planning margins. There is no statistically significant difference in doses to vascular spaces expanded by 7mm or 10mm under either alignment. IMRT plans with small margins and soft tissue alignment compared to large margins with bony alignment show marginal statistically significant difference in rectal DVHs, prostate or SV coverage, and statistically significant increases in doses to bladder and vascular spaces. The reduction of margins around prostate and SV with soft tissue alignment will results in statistically significant reduction of dose to rectum, bladder, prostate and SV, but no statistically significant difference of dose to LN. Conclusion: Comparing the recommended planning strategies —(a) tight margins with prostate‐based alignment and, (b) large margins with bony anatomy alignment— we find that similar doses to rectum and targets are achieved, and a slight reduction in bladder doses occurs under strategy a. From these we recommend small margins with prostate‐based alignment to maximize sparing and target coverage.
36(2009); http://dx.doi.org/10.1118/1.3182380View Description Hide Description
Purpose: To evaluate the uncertainty of computed proton range in radiotherapytreatment planning which is attributed to random component in CT numbers. Method and Materials: We utilize a random number generator to simulate a white Gaussian noise in CT numbers along the proton pathlength. The proton range is computed using continuous slowing down approximation which is valid for most of proton range. To simulate the statistical straggling of computed proton range, this procedure is iteratively repeated to obtain convergence of proton range PDF which is approaching a Gaussian. The FWHM (full‐width at half maximum) of the range PDF is used as a measure of uncertainty. Results: We investigate parameters which affect the proton range uncertainty in the presence of CTimagenoise. These parameters may include 1) initial proton energy, 2) noise period and 3) noise amplitude. The FWHM of range PDF increases linearly with the noise period. These results indicate that low frequency fluctuations in CTimagenoise can significantly increase the range uncertainty. We have also computed the range PDF as a function of initial proton energy. The FWHM of range PDF increases linearly with the initial proton energy. For the maximum proton energy of 250 MeV, the FWHM of proton range PDF can achieve a value of 5 mm in the presence of CTimagenoise. We note that the ratio FWHM/range increases as the proton range decreases; therefore, the relative range uncertainty is larger for smaller ranges. Conclusions: Range uncertainties due to CTimagenoise can be significant and comparable to the uncertainties attributed to the calibration of CT numbers. The relative range uncertainty increases as the range decreases. Noise reduction in CTimages using smoothing and denoising algorithms can be recommended to reduce the standard deviation of range PDF.
TU‐D‐BRC‐06: Towards Online Image Guided Radiotherapy for Cervical Cancer: Accurate Cervix‐Uterus Prediction Based On Measured Bladder Volumes36(2009); http://dx.doi.org/10.1118/1.3182381View Description Hide Description
Purpose: To investigate whether variable bladder filling CT‐scans can be used to predict the cervix‐uterus shape and position based on measured bladder volumes and to determine the number of CT‐scans required for an accurate prediction. Methods and Materials: Two series of CT‐scans were acquired for eleven patients in prone position, the first before EBRT and the second after 40 Gy. Each series consisted of a full bladder CT‐scan and four subsequent CT‐scans with a naturally filling bladder (empty to full). The cervix‐uterus and bladder were manually contoured and 3D cervix‐uterus surfaces were generated. For each patient non‐rigid registration was used to generate corresponding points on all ten surfaces. Patient‐specific models were built by fitting the coordinates of the corresponding points of a variable number of first series surfaces to linear functions of the bladder volume. Each model was used to predict, based on bladder volume the cervix‐uterus surfaces excluded from the model generation. The prediction error was quantified by the margin required around the predicted to accommodate 95% of the observed surface. Results: The maximum cervix‐uterus displacement range was 14–49 mm at planning and 16–72 mm after 40 Gy. The prediction error moderately increased with the decrease of the number of input surfaces (from 5 to 7 mm at planning and from 8 to 9 mm after 40 Gy for 4 to 2 input surfaces). For 9/11 patients the bladder vs. cervix‐uterus relationship was hardly influenced by radiotherapy (error range 6–7 mm). Conclusion: This work demonstrates the potential for accurate cervix‐uterus localization by using a prediction model based on measured bladder volumes. For most patients the prediction error was well below the extent of motion of the cervix‐uterus, even if only two CT‐scans were included in the model. The model could be used to facilitate the adaptation of treatment plans.
TU‐D‐BRC‐07: An Online Replanning Technique with Deformation‐Based Aperture Morphing and Weight Optimization36(2009); http://dx.doi.org/10.1118/1.3182382View Description Hide Description
Purpose: We have previously proposed a fast online replanning method using contour‐based segment aperture morphing (SAM) and segment weight optimization (SWO), that was shown to be effective in correcting interfractional variations. Here we propose to enhance the replanning method by developing new SAM algorithm using deformation field. Methods: A new software tool was developed to incorporate the full 3D deformation field between the planning and daily CTs to morph the planed apertures based on the daily CT. The previously reported deformable registration algorithm based on a fast symmetric Demons method with the use of masks was utilized to generate the deformation vector field. The vector field was used to generate daily contours and to morph aperture shapes. The SWO was applied to further improve the dosimetry for the morphed apertures. The resulting doses were compared with those obtained with the full‐scope reoptimization (the golden standard), rigid‐body‐registration based repositioning (the current IGRT practice), and the previous contour‐based SAM. The new deformation‐based SAM and SWO process was tested for selected prostate cases with large daily organ deformation. Results: The deformable image registration took 1 minute, and the entire deformation‐based SAM and SWO process was completed within 5 minutes. The plans generated by the new deformation‐based SAM were considerably better than those from the repositioning, and were comparable to the contour‐based SAM. SWO further improved the plan quality, resulting in the final plans that are comparable to the full‐blown reoptimization plan. Conclusion: An algorithm using deformation field for aperture morphing was successfully developed and tested. The replanning process including the new algorithm is effective and can be completed within a reasonable timeframe (5–10 minutes) for practical implementation. The new replanning process can be fully automated and can effectively correct for interfraction errors, even for complex cases with multiple targets and overlapping critical structures.
TU‐D‐BRC‐08: Quantifying Uncertainty in Dose Deposition Resulting From Patient Breathing Variability36(2009); http://dx.doi.org/10.1118/1.3182383View Description Hide Description
Purpose: To obtain more accurate predictions of dose deposition and elucidate discrepancies in planned and deliveredtreatments resulting from respiratory‐induced target motion, we develope a stochastic model of dose deposition and a computational framework to determine the effects of patient breathing variability on stereotactic Body Radiation Therapy(SBRT)dose calculations. Method and Materials: We developed a model to characterize the underlying day‐to‐day variability in patient breathing patterns and incorporated this model into a calculation of dose deposition. To accomplish this, we fitted Gaussian Mixture Models (GMM) to Realtime Position Management (RPM) breathing trace amplitude recordings from the same patient on different days. Using Principal Component Analysis, we identified the modes of greatest variation in the GMM fits. We then modeled each principal mode as an independent and uncorrelated random variable. Applying the stochastic computational methods, we determined statistical moments (e.g. mean and variance) of the predicted dose for a particular SBRT patient. Results: For the patient study, the standard deviation in dose resulting from day‐to‐day breathing variation is as large as 7\% of the maximum prescribed dose. High‐dose level variation is observed to occur near the boundaries of the lesion corresponding to regions of high dose gradient that experience large respiratory‐induced organ deformation. Such areas are important because they indicate regions in which the planned dose may differ significantly from the true deposition during treatment and are likely candidates for over‐or under‐dosing. Additionally, the maximum calculated target dose resulting from incorporation of motion is 12.4\% lower than the maximum treatment‐planning‐predicteddose, indicating that organ motion can significantly alter the overall dose deposition for a treatment.Conclusion: Our framework provides an accurate and efficient means of calculating the effect of respiratory‐induced organ motion on SBRTdose deposition.
TU‐D‐BRC‐09: Is Respiratory Gating More Prone to Dosimetric Errors Due to Irregular Respiratory Motion?36(2009); http://dx.doi.org/10.1118/1.3182384View Description Hide Description
Purpose: The aim of this study is to investigate the dosimetric effects of the irregular respiratory motion of lungtumors treated using respiratory gating techniques. Several treatment parameters such as the gating window, average target motion, and the inclusion of respiratory coaching techniques are also investigated for their effect on target dosimetric coverage. Methods and Materials: Respiratory motion traces of 11 lungcancer patients are used to reflect the typical respiratory irregularities. From the entire respiratory motion data, an average respiratory cycle is calculated to represent the corresponding regular motion for the patient. Using the regular motion amplitude, target volume is expanded by the amount of residual motion observed within the gating window and a treatment plan is generated. The target is then subjected to both regular and irregular motions without affecting anatomical topology. Target dosimetry under regular and irregular motion conditions is compared for quantitative evaluations Results: The reduction in Dmin due to irregular motion was less than 1% (4%) for a mean target motion of 1.0 cm (2.0 cm) when no gating technologies utilized. The drop in target Dmin increased as the shorter gating window sizes are implemented. For gated treatments with 50% duty cycle, largest change in Dmin was 2% and 9% for 1.0 and 2.0 cm mean target motions respectively. More importantly, gated treatments were observed to undermine target dosimetry in cases where the synchronization between the target motion and the external respiratory signal doesn't stay constant as observed by other investigators. Within the parameter space investigated here, up to 35% reduction in target Dmin was observed. Conclusion: Being a more aggressive treatment strategy, respiratory gated treatments were observed to be more susceptible to dosimetric uncertainties as compared to traditional delivery techniques in presence of irregular respiratory motion.
- Volumetric and Surface Imaging for Radiotherapy Guidance
TU‐E‐BRC‐01: Non‐Axisymmetric Permanent Magnet Design Optimization for a Coupled Medical Linear Accelerator and Magnetic Resonance Imager36(2009); http://dx.doi.org/10.1118/1.3182415View Description Hide Description
Purpose: To develop a robust permanent magnet design optimization scheme capable of producing non‐axisymmetric pole plate designs resulting in greater levels of magnetic field homogeneity achieved with rose‐ring designs as currently found in industry. To permit the reduction of the overall dimensions of a magnet structure to be used in a coupled medicallinear accelerator and magnetic resonanceimager while maintaining acceptable levels of magnetic field homogeneity. Method and Materials: A steepest descent optimization program is scripted in MATLAB and integrated with finite element analysis software COMSOL Multiphysics. The pole plate surface is parameterized with a 2D grid distribution of mesh independent control points. The optimization is completed in two stages, the first yielding axisymmetric designs about the static field direction, the second permitting non‐axisymmetric designs. A model of a 0.2T biplanar magnet structure with a rose‐ring design is used to quantify the increase in PPM inhomogeneity over a 30cm and 40cm DSV at isocenter when the lateral dimensions of the structure are decreased. The optimization scheme is executed on the reduced structure to obtain a modified pole plate design. Results:FEM analysis indicates an increase in inhomogeneity of 184ppm to 4313ppm over a 30cm DSV, and 344ppm to 7649ppm over a 40cm DSV, as the lateral dimensions of the structure are reduced. The optimization results in an axisymmetric pole plate design with inhomogeneities of 200ppm and 1163ppm over a 30cm and 40cm DSV, respectively. Further optimization results in a non‐axisymmetric design with inhomogeneities of 49ppm and 271ppm over a 30cm and 40cm DSV, respectively. Conclusion: A permanent magnet optimization scheme is demonstrated to yield novel pole plate designs with greater levels of field homogeneity than presently achieved. The method permits reducing the size of permanent magnet structures while maintaining suitable levels of field homogeneity over typical sized imaging volumes.
TU‐E‐BRC‐02: Combined Use of CT and MVCBCT for Optimal Dose Calculation in Presence of High‐Z Material36(2009); http://dx.doi.org/10.1118/1.3182416View Description Hide Description
Purpose: High atomic number (Z) materials such as dental fillings complicate the use of conventional CTimages and application of the superposition dose calculation algorithm. This work investigates the accuracy of current planning methods for patients with dental work and proposes to combine the strengths of two imaging modalities in order to obtain a more realistic dose distribution in presence of high‐Z materials. Method and Materials: Dose simulations were performed on conventional CT and megavoltage cone‐beam CT (MVCBCT) images calibrated for electron density. MVCBCT images were acquired on a phantom and patient using two MVCBCT beam lines having substantially different energy spectra. Dose measurements on a cylindrical phantom with an ion chamber and MOSFETdetectors were compared to simulations performed in Pinnacle and with Monte Carlo.Monte Carlo simulations used EGSnrc/BEAMnrc code with a treatment head model that produced a close match to measured data. Phantom and patient CTimages were simulated with MCRTP using a voxel resolution of 3mm. Results: MVCBCT produced much less high‐Z artifact than CT. MVCBCT images from the highest energy beam with low energy photons filtered out had the least artifact but the lowest soft‐tissue resolution. Dose measurements on phantom were much closer to simulations performed with Monte Carlo (2–5%) than Pinnacle (7–14%). Simulations indicate that Pinnacle underestimates dose in soft tissue adjacent to fillings as well as attenuation through the high‐Z material. This could result in a hotter gum dose and underdosage of the target volume. Conclusion: Combined CT‐MVCBCT images of head and neck patients are being used with Monte Carlo to quantify the dosimetric impact of dental fillings for IMRT and 3DCRT. In absence of corrective measure, constraints on dental contours could be used to limit the number of IMRT segments entering through the high‐Z material before reaching the tumor.
TU‐E‐BRC‐03: A Comparison Between 4D and 3D PET‐Based Tumor Volume Definition for Lung Cancer Patients36(2009); http://dx.doi.org/10.1118/1.3182417View Description Hide Description
Purpose: Utilizing an automatic segmentation technique, the definition of tumor volume on clinical 4D (gated) and 3D (non‐gated) PET scans was compared and the potential benefits of gated PET/CT acquisition for lungcancer patients were explored. Method and Materials: Data from patients that underwent 3D and 4D PET/CT scans prior to radiation therapy were analyzed. A belt (Anzai Medical Systems Inc) was used to monitor patient breathing and gate the 4D acquisition. The raw data were then binned into five breathing phases. On the reconstructedPETimages, an automatic segmentation technique based on a Gaussian Mixture Model (GMM) was applied to each phase individually and five gross tumor volumes (4D‐GTVs) were defined. The individual 4D‐GTVs were grouped together to create an internal target volume (ITV). The GMM method was also applied to the 3D data and a 3D‐GTV was defined. The resulting ITV and GTVs were compared and the contours were superimposed on the CT portion of the PET/CT scan to visually assess tumor coverage. Results: The ITV was 8% to 148% greater than the 3D‐GTV with an average of 45%. Visual inspection of the ITV and 3D‐GTV contours superimposed on the CT revealed better coverage of the CT‐defined tumor with the ITV. This was especially apparent in the superior and inferior portions of tumors exhibiting respiratory motion close to 1cm and small lung nodules with moderate FDG uptake. Deformations in the FDG uptake between different breathing phases were observed, demonstrating the possibility of deformations in the tumor shape. Conclusion: This is one of the first studies comparing tumor volume definition on 4D and 3D PET/CT clinical data. The results suggest that ITV obtained from 4D PET/CT covers the full extent of moving tumors better than a 3D‐GTV obtained from 3D PET/CT.
36(2009); http://dx.doi.org/10.1118/1.3182418View Description Hide Description
Purpose: Integrated MRI+linac systems can potentially yield complete spatio‐temporal knowledge of the irradiated anatomy during beam‐on — representing the ideal guidance strategy for 4D radiotherapy delivery. In this work, we investigate rapid imaging strategies for such devices to enable real‐time, MR‐guided, motion‐adaptive radiation delivery. Method and Materials: SNR in MR images is described by: SNR μ B0×(Dx×Dy×Dz)× Tacq, where 0 B is the primary field, Dx, Dy and Dz are voxel dimensions and acq T is the acquisition time. We investigated trade‐offs between SNR and two important design and operational parameters for MRI+linac systems — (i) 0 B, which impacts design complexity, and (ii) acq T, which impacts the spatio‐temporal accuracy of real‐time guidance. In the first study, SSFP and SPGR sequences were employed to acquire 3D volumes (1.2 s/volume) and 2D coronal slices (0.3 s/slice) of the thoracic region from five human subjects (1.5T scanner). To simulate lower fieldstrength, imageSNR was progressively degraded by adding increasing levels of gaussian noise. A fat deposit on the diaphragm was segmented in the noise‐free and the degraded images and the error in the estimated position was computed. In the second study, faster acquisition through partial k‐space scanning was simulated. A cylindrical water‐filled phantom containing seven oilfilled cylinders was imaged and a progressively increasing number of mid‐frequency phase encode lines were zeroed prior to reconstruction. The centers of oil‐filled cylinders were auromatically segmented in the full and partial k‐space acquisitions, and the positional error was computed with respect to the full k‐space image.Results: Positional errors of the anatomic feature were within 1.5 mm for a factor‐of‐6 SNR degradation, corresponding to B0 = 0.25T. Partial k‐space acquisition could be performed to increase acquisition speed by over a factor of 5, while maintaining sub‐1 mm accuracy. Conclusion: These initial studies indicate the feasibility of low‐field, real‐time MRI for intrafraction motion management using integrated MRI+linac systems.
TU‐E‐BRC‐05: First MR Images Obtained During Megavoltage Photon Irradiation From An Integrated Linac‐MR System36(2009); http://dx.doi.org/10.1118/1.3182419View Description Hide Description
Purpose: To prove whether our hybrid linac‐MR design can sufficiently decrease mutual magnetic and radio frequency (RF) interferences to produce a MRimage during linac‐irradiation that has similar geometric accuracy and image contrast to an MRimage obtained without linac‐irradiation. Method and Materials: A hybrid prototype has been built consisting of a 6 MV linac mounted on the open end of a bi‐planar 0.2 T permanent‐magnet MR system. Both the linac waveguide and the MR system are mounted onto a single gantry that would rotate around a subject. The permanent magnet poles (82×82 cm2) are rigidly held apart to give 27.9cm pole‐to‐pole opening with flat gradients (40 mT/m max) running under a TMX NRC console (Canada). The linac components are comprised of salvaged magnetron‐based Varian 6 MV 600C decommissioned system. The distance from linac target to MR isocentre is 80 cm. Magnetic and RF shielding calculations were performed by finite element analysis and confirmed with appropriate measurements. Faraday cage shielding, typical of all MR installations, was also provided. The testing phantom was an acrylic rectangular cube, 15.95 × 15.95 × 25.4 mm with holes of diameters 2.52 mm, 3.45 mm and 4.78 mm drilled parallel to the length (25.4 mm) of rectangular cube. The cube was immersed in a 10 mM solution of within a plastic container 22.5 mm inner diameter. The container with the cube insert were placed inside an inductively tuned solenoid RF coil with an integrated pin‐diode transmit/receive switch. Results:Images (128×128) were obtained in about 38 secs using raw gradient echo sequences. Compared to no‐radiation, MRimages during linac‐radiation had no geometric distortion but slightly reduced SNR. Conclusion: Our linac‐MR design produces MRimages obtained during or without linac irradiation that are sufficiently similar to implement real‐time MR‐guided radiotherapy.
36(2009); http://dx.doi.org/10.1118/1.3182420View Description Hide Description
Purpose: To investigate the feasibility of using three‐dimensional surfaceimagingcameras as an external surrogate of target motion through a temporal synchronization with kV imaging.Method and Materials: A gate controller, connected between kV fluoroscopy and AlignRT (Vision RT Ltd, London) computers, was triggered with kV beam‐on and reflected in AlignRT output data. First, phantom experiments were performed using a programmable respiratory motion platform (sinusoidal curves, 12–18 BPM). The platform included a chest‐wall component (A‐P amplitude = 1 cm) tracked with the surfacecamera, while the object translated superior‐inferior was fluoroscopically tracked (300 frames, frequency ∼5.5 fps). Accuracy of tracking the chest‐wall platform was assessed. Increasing the complexity of experiments, tumor displacement curves from three patients were simulated. Our approach was further validated by imaging a free‐breathing lungcancer patient over 11 fractions with simultaneous AlignRT and kV (300 frames/session, ∼55 seconds). Results: For simple sinusoidal curves, measured amplitude (peak‐to‐peak) was 1.005 ± 0.003, 1.013 ± 0.003 cm, 1.003 ± 0.005 cm for 12, 15, and 18 BPM, respectively, demonstrating excellent agreement with the expected chest platform amplitude of 1.0 cm. Period measurements were within 0.2% of expected for the surfacecameras, and within 0.9 % of expected for fluoroscopy. Using simulated patient data and the motion platform, the latency between the surfacetracking system and fluoroscopy was determined by performing linear regression between the peak time from AlignRT and fluoroscopy. For the three patient cases studied, latency was found to be 0.65 ± 0.03 seconds. Moderate agreement was observed between the diaphragm, tumor, and abdomen tracked for the patient studied. Conclusion: The utility of surfaceimagingcameras for tracking while synchronized with fluoroscopy has been demonstrated. Additional patient studies are warranted for further validation, which may facilitate gated radiation therapy treatments using surfaceimagingcameras as the external surrogate.
36(2009); http://dx.doi.org/10.1118/1.3182421View Description Hide Description
Purposes: Scatter and noise are well‐known factors that degrade the cone‐beam CT(CBCT)image quality. In addition to these factors, the fundamental reconstruction process adopted by most commercial systems is also imperfect. The reconstructions are mostly based on the Feldkamp formula. Such algorithm gives the accurate solution for the central slice but approximations for the remaining slices. The error increases when the slices are further away from the center due to “incompleteness” of the projection data. To combat this problem, we developed in this work a new reconstruction algorithm, which improved the CBCTimage quality significantly. Methods and Materials: The reconstruction algorithm integrated the Feldkamp formula into an iterative optimization process. For validation, a three‐dimensional (3D) digital thoracic phantom was constructed based on a patient CTimage and used to simulate CBCT projections. Noise and scatter were excluded from the simulation in order to isolate the basic theoretical issues in reconstruction. The projections were reconstructed with the Feldkamp algorithm, Simultaneous Algebraic Reconstruction Technique (SART), and our proposed “FDK‐ART” algorithm, respectively. The images were then systematically evaluated. Results: Artifacts became apparent for outer slices when Feldkamp algorithm was used. These artifacts were significantly reduced with the proposed algorithm. Although Feldkamp gives an accurate solution for the central slice, the root mean square error was shown to be reduced greatly by the proposed algorithm. By visual and profile comparison, it was found that the image resolution using our new method was clearly superior to that obtained with traditional Feldkamp algorithm or SART algorithm. The algorithm converged within 5 iterations for all sample cases tested. Conclusions: A new reconstruction algorithm has been developed for CBCTimaging. The new algorithm not only reduced the errors for outer slices, but also significantly improved the image resolution and the overall quantitative accuracy for every slice.
- Imaging for Therapy Assessment
36(2009); http://dx.doi.org/10.1118/1.3182469View Description Hide Description
Purpose: Uncertainty in the location of the distal dose edge is one of the main concerns about proton therapy. It leads to cautious treatment plans that partially neutralize the dosimetric advantage of protons. Vertebral bone marrow responds to radiation with fatty replacement that is visible on post‐treatment MRI. This presents a unique opportunity to visualize radiation effectsin vivo. We have developed a method that uses spine MRI changes to precisely localize the distal dose edge in spinal protonradiation patients. Method and Materials: We carefully registered treatment planningCT scans and follow‐up T1‐weighted MRI scans from 10 proton spinal radiation patients. A radiationdose‐MRI signal intensity curve was created using the lateral beam penumbra in the sacrum. This curve was then used to quantitatively examine possible systematic or spatially varying proton range errors. Results: In the lateral penumbra there was a gradual increase in signal intensity with higher dose throughout the full dose range of 0–37.5 Gy. In the distal dose fall‐off region, the beam appeared to penetrate farther than planned in the central part of the vertebral bodies. The mean overshoot in five patients was 2.71 mm (95% confidence interval 1.13–4.28 mm). These errors are probably not clinically significant with current treatment planning procedures. Conclusion: We have demonstrated in vivoproton range verification based on post‐treatment spine MRI changes. Our analysis indicates that if range errors occur in spine treatments, their magnitude is at most a few millimeters in the majority of patients. It may be possible to extend our technique to MRI sequences that show early bone marrow changes. It could then be used for adaptive modification of spine radiation plans in order to reduce radiationdose to bone marrow and other normal tissues.
WE‐C‐BRC‐02: Combined Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE‐MRI) and Magnetic Resonance Thermal Imaging (MRTI) for Optimal Hyperthermia Treatment of Advanced Extremity Sarcomas: Fifteen Patients Update36(2009); http://dx.doi.org/10.1118/1.3182470View Description Hide Description
Purpose: To deliver optimal hyperthermia (HT) treatments for patients with advanced extremity sarcomas using DCE‐MRI and MRTI Method and Materials: Patients were treated on a protocol using radiotherapy consisting of 45 Gy and once a week HT for 5 weeks in a setting that allows the use of a 1.5 T GE magnet for imaging. Pre‐ and post first HT treatment DCE‐MR images were acquired and analyzed with software from iCAD Inc. (Nashua, NH) based on a full time point pharmacokinetics (PK) analysis that measures tumor's permeability (Ktrans) and extracellular volume fraction (Ve). Multi‐slice temperature rise images were obtained using the proton resonance frequency shift technique during heating in a 140 MHz phased array HT applicator and compared to invasive temperature measurements. Thermal dose metrics were calculated and correlated with response and with the PK parameters. Results: Fifteen patients were enrolled on this protocol; ten patients were heated. Permeability maps derived from DCE‐MR images were used to guide the placement of the invasive catheter to span regions varying from high to low perfusion. The MRTI images were used to steer the power, with the intent of heating the tumor uniformly while maintaining surrounding normal tissue at lower temperatures. From the evaluable HT treatments, we achieved excellent correlation (ΔT<1°C) between the MRTI and invasive measurements. As a trend, patients that were either pathological complete or partial responders had a decrease in Ktrans. The Ve parameter showed no clear trend with pathological response or % necrosis. As expected, tumors that were more vascularized (higher Ktrans) heated less than tumors with a high degree of necrosis or fluid pockets. Conclusion: DCE‐MRI coupled with MRTI provides information on tumor environment that can improve planning, delivery, and evaluation of HT treatments.
Supported by a grant from the NCI CA42745.
WE‐C‐BRC‐03: Evaluating Extent of Cell Death in 3D Mid‐To‐High Frequency Ultrasound by Registration with Whole‐Mount Tumor Histopathology36(2009); http://dx.doi.org/10.1118/1.3182471View Description Hide Description
Purpose: In this study, we investigate the precision and accuracy of mid‐to‐high frequency ultrasound imaging to assess non‐invasively cell death for incorporation of this method in pre‐clinical and clinical practice to characterize tumor response to radiotherapy.Method and Materials:Tumor xenografts (n=8) of head and neck cancer were exposed to radiationdoses of 2, 4 and 8Gy. Ultrasoundimages were collected with an ultrasoundscanner using frequencies of 15–35MHz before and 24 hours after exposure to radiation. Irradiated tumors exhibited large hyperechoic regions in ultrasoundsimages 24 hours after exposure to radiation that corresponded to areas of cell death in histology. The ultrasoundimages were registered with the histological images of the tumor slices taken at regular intervals. The tumor was contoured on histological slices and ultrasoundimages, the regions of cell death were contoured on histological slices and the hyperechoic regions were contoured on ultrasoundimages. Each set of contours was converted to a surface mesh. The volume and center of mass were calculated for each representation determined by a surface mesh. Results: The average difference between the relative (to histology) volume representations in histology and ultrasound were 10.7%±8.9% for tumor and 21.7%±12.2% for cell death. The average differences between the relative (to the maximum dimension of the tumor) center of mass of volume representations in histology and ultrasoundimages were 2.7%±2.0% for tumor and 15.5%±8.9 % for cell death. Conclusion: The method provides the correspondence between the volumes of cell death assessed from histology and from ultrasound imaging and can be used to assess early tumor response to radiotherapy. Part of the differences associated with cell death representation in histological and ultrasoundimages (21.7%) was caused by the differences in tumor representation (10.7%) in these images. The effect of these uncertainties is the subject of ongoing investigation.
36(2009); http://dx.doi.org/10.1118/1.3182472View Description Hide Description
Purpose: Assessment of treatment response in solid tumors is essential for disease management but typical response measures are limited and non‐robust. We compared the predictive power of several PET‐based measures in cancer patients receiving therapy. Method and Materials: Fourteen adult patients with advanced solid malignancies were treated with sunitinib malate, a molecular targeted agent. Using the cellularproliferation marker FLT, whole‐body PET/CT scans were acquired pre‐, mid‐, and post‐treatment. Lesions were segmented on PETimages and treatment response was assessed via percent change of the following measures: SUVmean, SUVmax, SUVpeak, SUVtotal, and PET defined tumor size: unidimensional, bidimensional, and volume. PET response assessment and clinical endpoint of therapy were used independently to classify patients into response categories: partial response (PR), stable disease (SD), or progressive disease (PD). Predictive power and robustness of PET measures were tested by varying response category cutoffs to generate ROC curves and Matthews correlation coefficients (MCC), a classification quality measure. Results: SUV measures demonstrated improved predictive power over PET defined size measures (MCCaverage: 0.27 vs. 0.21). Of all measures, SUVpeak was the most robust response predictor (MCCaverage: 0.30, MCCmax: 1.0), especially for prediction of PR. Of PET defined size measures, bidimensional size was the most robust response predictor. Changes in SUV measures pre‐ to mid‐treatment showed improved predictive power over pre‐ and mid‐ to post‐treatment changes (MCCaverage: 0.32, 0.27, and 0.22 respectively). Prediction of PR was vastly superior to that of SD and PD (MCCaverage: 0.44, 0.16, and 0.14 respectively). ROC analysis supported MCC results. Conclusion: Comparison of PET‐based measures revealed that SUV measures were more predictive of response than PET defined size measures. SUVpeak, a rarely used functional measure, demonstrated the most robust predictive power. Future treatment response studies should test and implement more robust measures, like SUVpeak, for improved response assessment.