Measuring uncertainty in dose delivered to the cochlea due to setup error during external beam treatment of patients with cancer of the head and neck.

PURPOSE: To use Cone Beam CT scans obtained just prior to treatments of head and neck cancer patients to measure the setup error and cumulative dose uncertainty of the cochlea. METHODS: Data from 10 head and neck patients with 10 planning CTs and 52 Cone Beam CTs taken at time of treatment were used in this study. Patients were treated with conventional fractionation using an IMRT dose painting technique, most with 33 fractions. Weekly radiographic imaging was used to correct the patient setup. The authors used rigid registration of the planning CT and Cone Beam CT scans to find the translational and rotational setup errors, and the spatial setup errors of the cochlea. The planning CT was rotated and translated such that the cochlea positions match those seen in the cone beam scans, cochlea doses were recalculated and fractional doses accumulated. Uncertainties in the positions and cumulative doses of the cochlea were calculated with and without setup adjustments from radiographic imaging. RESULTS: The mean setup error of the cochlea was 0.04 ± 0.33 or 0.06 ± 0.43 cm for RL, 0.09 ± 0.27 or 0.07 ± 0.48 cm for AP, and 0.00 ± 0.21 or -0.24 ± 0.45 cm for SI with and without radiographic imaging, respectively. Setup with radiographic imaging reduced the standard deviation of the setup error by roughly 1-2 mm. The uncertainty of the cochlea dose depends on the treatment plan and the relative positions of the cochlea and target volumes. Combining results for the left and right cochlea, the authors found the accumulated uncertainty of the cochlea dose per fraction was 4.82 (0.39-16.8) cGy, or 10.1 (0.8-32.4) cGy, with and without radiographic imaging, respectively; the percentage uncertainties relative to the planned doses were 4.32% (0.28%-9.06%) and 10.2% (0.7%-63.6%), respectively. CONCLUSIONS: Patient setup error introduces uncertainty in the position of the cochlea during radiation treatment. With the assistance of radiographic imaging during setup, the standard deviation of setup error reduced by 31%, 42%, and 54% in RL, AP, and SI direction, respectively, and consequently, the uncertainty of the mean dose to cochlea reduced more than 50%. The authors estimate that the effects of these uncertainties on the probability of hearing loss for an individual patient could be as large as 10%.

SU-E-J-125: Uncertainty in Accumulated Dose to the Cochlea Due to Setup Error during External Beam Treatment of Head & Neck Cancer Patients.

PURPOSE: Treatment uncertainties are not included in modeling dose response of hearing loss. We determine uncertainty of accumulated dose to cochleas of head & neck cancer patients due to setup error during treatments of external beam IMRT. METHODS: We studied 8 patients: 1 planning CT for each patient and total 40 cone beam CTs. Those patients were treated with 33 fractions. Treatments delivered 70 Gy to PTV, 50 to 60 Gy to nodal and sub-clinical disease using IMRT dose painting. Setup error was measured using 6-dimensional rigid registration between planning CT and cone beam scans obtained prior to treatment. Planning CTs were transformed and resliced to produce 'treatment CTs' according to the measured setup error matrix to simulate the cochlea position during treatment. We calculated dose delivered to cochleas at each treatment position. Treated dose to the cochlea from each fraction was accumulated on the planning CT using its registration with each 'treatment CT'. RESULTS: The RMS value of set error of left cochlea is 0.48pm0.28cm, right cochlea is 0.47pm0.26cm. Mean values of left and right cochleas are ∼zero. Uncertainty in the dose to cochlea depends on each treatment plan and relative positions of the cochlea and target volumes. Average uncertainty of mean dose to cochlea is 5.0%, or 5.2cGy per fraction. CONCLUSION: Patient setup error introduces uncertainty to the position of the cochlea and consequent uncertainties in accumulated dose. Our method calculates accumulated dose delivered to the same cochlea volume at different treatment positions due to setup error. We found uncertainty in cochlea doses was 5% of planned dose after 4 to 7 fractions. The largest uncertainty was 17 cGy per fraction. Our Results will be used to determine uncertainties in dose response of hearing loss in head and neck patients. NIH R01-CA129182.

SU-E-J-67: A Potential New Technique to Measure Respiratory Tidal Volume Using Optical Surface Imaging: A Geometric and Deformable Phantom Study.

PURPOSE: To develop a new method for accurate measurement of dynamic respiratory tidal volume, we investigate the feasibility of measuring torso volume change using optical surface imaging (OSI). METHODS: Based on a validated volume conservation theory, the tidal volume is equal to the volume change of the torso during quiet respiration (Li et al, PMB, 54:1693, 2009). A clinical OSI system was employed to acquire surface images of seven geometric phantoms and two 'deformable' torso phantoms. The mesh surface images were converted into contours for volume calculation using a treatment planning system. For geometric phantoms, their volumes under the incomplete surface images were calculated with aid of their symmetry. The results were compared with theoretical calculation and water containment experiments. For deformable torso phantoms, we created volume-controlled deformation stages by placing deformable PlayDoh (DPD) materials on top of rigid Rando/Thorax phantoms, mimicking respiration-induced torso surface elevation and volume change. The volume difference under the surfaces with and without the DPD padding was calculated with aid of a common posterior line to enclose the region of interest. Three different volumes of DPD padding (>500cc) were mounted on the torso phantoms and CT scanned for volume measurements. RESULTS: For geometric phantoms, the OSI measured volume had accuracy (±1s) of 0.0%±1.6% (vs. geometric volume calculation) and 0.6%±3.8% (vs. water containment experiment). For deformable torso phantoms, the volume change was measured using OSI with an accuracy of 1.5%±2.5% against the measured volume using CT imaging. Linear regression showed a one-to-one relationship between the OSI volumes and CT volumes with a slope of 1.003 (r2=0.999). CONCLUSIONS: The optical surface imaging system can accurately measure the volume of geometric phantoms and the volume change of deformable torso phantoms. The accuracy is about 3% against standard volume measurement methods. Further study on human subjects is under investigation. Memorial Sloan-Kettering Cancer Center has a reserach agreement with Vision RT, Inc.

SU-E-T-583: Feasibility of Constraining Dose to the Nausea Center (area Postrema and Dorsal Vagal Complex) in IMRT Treatment Planning of the Head and Neck.

PURPOSE: Nausea and vomiting have been known to occur in patients undergoing external beam radiation treatments for head&neck cancers. We sought to determine the feasibility of limiting the dose delivered to the nausea center, area postrema (AP) and dorsal vagal complex (DVC), for these patients without compromising target coverage and critical organ doses. METHODS: In a retrospective study 23 oropharyngeal cancer patients were identified as being treated with definitive or adjuvant radiotherapy at Memorial Sloan Kettering Cancer Center. Patients were treated solely with external beam radiation using intensity modulated radiation therapy (IMRT). The nausea center was carefully contoured in the treatment CT with the assistance of a board certified neuroradiologist. The doses delivered to the nausea center were calculated for each plan delivered. Cases were replanned offline to determine the lowest achievable nausea center dose that does not compromise the overall PTV coverage or critical structures doses, these being brainstem, spinal cord, cochleas, and temporal lobes. RESULTS: Patients reporting higher nausea grade had median AP and DVC doses of 38.7Gy and 40.4Gy, respectively. Patients reporting higher vomiting grade had median AP and DVC doses of 39.5Gy and 44.7Gy, respectively. Replanning resulted in reduced dose to AP by an average of 18% and to the DVC by an average of 16% while maintaining adequate target coverage and doses to the critical organs the same or decreased by 1-4% . We aim to achieve a max dose of 36Gy to AP and 38Gy to DVC for these cases. CONCLUSIONS: It is feasible to limit the doses to the nausea center without compromising target coverage or critical organ limits for oropharyngeal cancer patients undergoing IMRT treatment. Clinical results indicating an association between radiation dose to the nausea center and development of nausea and/or vomiting can potentially be addressed by implementing this technique.

Evaluation of an automated deformable image matching method for quantifying lung motion in respiration-correlated CT images.

We have evaluated an automated registration procedure for predicting tumor and lung deformation based on CT images of the thorax obtained at different respiration phases. The method uses a viscous fluid model of tissue deformation to map voxels from one CT dataset to another. To validate the deformable matching algorithm we used a respiration-correlated CT protocol to acquire images at different phases of the respiratory cycle for six patients with nonsmall cell lung carcinoma. The position and shape of the deformable gross tumor volumes (GTV) at the end-inhale (EI) phase predicted by the algorithm was compared to those drawn by four observers. To minimize interobserver differences, all observers used the contours drawn by a single observer at end-exhale (EE) phase as a guideline to outline GTV contours at EI. The differences between model-predicted and observer-drawn GTV surfaces at EI, as well as differences between structures delineated by observers at EI (interobserver variations) were evaluated using a contour comparison algorithm written for this purpose, which determined the distance between the two surfaces along different directions. The mean and 90% confidence interval for model-predicted versus observer-drawn GTV surface differences over all patients and all directions were 2.6 and 5.1 mm, respectively, whereas the mean and 90% confidence interval for interobserver differences were 2.1 and 3.7 mm. We have also evaluated the algorithm's ability to predict normal tissue deformations by examining the three-dimensional (3-D) vector displacement of 41 landmarks placed by each observer at bronchial and vascular branch points in the lung between the EE and EI image sets (mean and 90% confidence interval displacements of 11.7 and 25.1 mm, respectively). The mean and 90% confidence interval discrepancy between model-predicted and observer-determined landmark displacements over all patients were 2.9 and 7.3 mm, whereas interobserver discrepancies were 2.8 and 6.0 mm. Paired t tests indicate no significant statistical differences between model predicted and observer drawn structures. We conclude that the accuracy of the algorithm to map lung anatomy in CT images at different respiratory phases is comparable to the variability in manual delineation. This method has therefore the potential for predicting and quantifying respiration-induced tumor motion in the lung.

Use of EPID for leaf position accuracy QA of dynamic multi-leaf collimator (DMLC) treatment.

We describe in this paper an alternative method for routine dynamic multi-leaf collimator (DMLC) quality assurance (QA) using an electronic portal imaging device (EPID). Currently, this QA is done at our institution by filming an intensity-modulated radiotherapy (IMRT) test field producing a pattern of five 1-mm bands 2 cm apart and performing a visual spot-check for leaf alignment, motion lags, sticking and any other mechanical problems. In this study, we used an amorphous silicon aS500 EPID and films contemporaneously for the DMLC QA to test the practicality and efficacy of EPID vis-à-vis film. The EPID image was transformed to an integrated dose map by first converting the reading to dose using a calibration curve, and then multiplying by the number of averaged frames. The EPID dose map was then back-projected to the central axis plane and was compared to the film measurements which were scanned and converted to dose using a film dosimetry system. We determined the full-width half-maximum (FWHM) of each band for both images, and evaluated the dose to the valley between two peaks. We also simulated mechanical problems by increasing the band gap to 1.5 mm for some leaf pairs. Our results show that EPID is as good as the film in resolving the band pattern of the IMRT test field. Although the resolution of the EPID is lower than that of the film (0.78 mm/pixel vs 0.36 mm/pixel for the film), it is high enough to faithfully reproduce the band pattern without significant distortion. The FWHM of the EPID is 2.84 mm, slightly higher than the 2.01 mm for the film. The lowest dose to the valley is significantly lower for the EPID (15.5% of the peak value) than for the film (28.6%), indicating that EPID is less energy independent. The simulated leaf problem can be spotted by visual inspection of both images; however, it is more difficult for the film without being scanned and contrast-enhanced. EPID images have the advantage of being already digital and their analysis can easily be automated to flag leaf pairs outside tolerance limits of set parameters such as FWHM, peak dose values, peak location, and distance between peaks. This automation is a new feature that will help preempt MLC motion interlocks and decrease machine downtime during actual IMRT treatment. We conclude that since EPID images can be acquired, analyzed and stored much more conveniently than film, EPID is a good alternative to film for routine DMLC QA.

A parameter optimization algorithm for intensity-modulated radiotherapy prostate treatment planning.

An iterative algorithm has been developed to analytically determine patient specific input parameters for intensity-modulated radiotherapy prostate treatment planning. The algorithm starts with a generic set of inverse planning parameters that include dose and volume constraints for the target and surrounding critical structures. The overlap region between the target volume and the rectum is used to determine the optimized target volume coverage goal. Sequential iterations are performed to vary the numerous parameters individually or in sets while other parameters remain fixed. A coarse grid search is first used to avoid convergence on a local maximum. Linear interpolation is then used to define a region for a fine grid search. Selected parameters are also tested for possible improvements in target coverage. In several representative test cases investigated the coverage of the planning target volume improved with the use of the algorithm while still meeting the clinical acceptability criteria for critical structures. The algorithm avoids time-consuming random trial and error variations that are often associated with difficult cases and also eliminates lengthy user learning curves. The methodology described in this paper can be applied to any treatment planning system that requires the user to select the input optimization parameters.