H Kang

Spatial and temporal performance of 3D optical surface imaging for real‐time head position tracking

PURPOSE The spatial and temporal tracking performance of a commercially available 3D optical surface imaging system is evaluated for its potential use in frameless stereotactic radiosurgery head tracking applications. METHODS Both 3D surface and infrared (IR) marker tracking were performed simultaneously on a head phantom mounted on an xyz motion stage and on four human subjects. To allow spatial and temporal comparison on human subjects, three points were simultaneously monitored, including the upper facial region (3D surface), a dental plate (IR markers), and upper forehead (IR markers). RESULTS For both static and dynamic phantom studies, the 3D surface tracker was found to have a root mean squared error (RMSE) of approximately 0.30 mm for region-of-interest (ROI) surface sizes greater than 1000 vertex points. Although, the processing period (1/fps) of the 3D surface system was found to linearly increase as a function of the number of ROI vertex points, the tracking accuracy was found to be independent of ROI size provided that the ROI was sufficiently large and contained features for registration. For human subjects, the RMSE between 3D surface tracking and IR marker tracking modalities was 0.22 mm left-right (x-axis), 0.44 mm superior-inferior (y-axis), 0.27 mm anterior-posterior (z-axis), 0.29° pitch (around x-axis), 0.18° roll (around y-axis), and 0.15° yaw (around z-axis). CONCLUSIONS 3D surface imaging has the potential to provide submillimeter level head motion tracking. This is provided that a highly accurate camera-to-LINAC frame of reference calibration can be performed and that the reference ROI is of sufficient size and contains suitable surface features for registration.

Assessment of interfractional variation of the breast surface following conventional patient positioning for whole-breast radiotherapy

The purpose of this study was to quantify the variability of the breast surface position when aligning whole‐breast patients to bony landmarks based on MV portal films or skin marks alone. Surface imaging was used to assess the breast surface position of 11 whole‐breast radiotherapy patients, but was not used for patient positioning. On filmed fractions, AlignRT v5.0 was used to capture the patients surface after initial positioning based on skin marks (28 “preshifts” surfaces), and after treatment couch shifts based on MV films (41 “postshifts” surfaces). Translations and rotations based on surface captures were recorded, as well as couch shifts based on MV films. For nonfilmed treatments, “daily” surface images were captured following positioning to skin marks alone. Group mean and systematic and random errors were calculated for all datasets. Pearson correlation coefficients, setup margins, and 95% limits of agreement (LOA) were calculated for preshifts translations and MV film shifts. LOA between postshifts surfaces and the filmed treatment positions were also computed. All the surface captures collected were retrospectively compared to both a DICOM reference surface created from the planning CT and to an AlignRT reference surface. All statistical analyses were performed using the DICOM reference surface dataset. AlignRT reference surface data was only used to calculate the LOA with the DICOM reference data. This helped assess any outcome differences between both reference surfaces. Setup margins for preshifts surfaces and MV films range between 8.3–12.0 mm and 5.4–13.4 mm, respectively. The largest margin is along the left–right (LR) direction for preshift surfaces, and along craniocaudal (CC) for films. LOA ranges between the preshifts surfaces and MV film shifts are large (12.6–21.9 mm); these decrease for postshifts surfaces (9.8–18.4 mm), but still show significant disagreements between the two modalities due to their focus on different anatomical landmarks (patients topography versus bony anatomy). Pearsons correlation coefficients further support this by showing low to moderate correlations in the anterior–posterior (AP) and LR directions (0.47–0.69) and no correlation along CC(<0.15). The use of an AlignRT reference surface compared to the DICOM reference surface does not significantly affect the LOA. Alignment of breast patients based solely on bony alignment may lead to interfractional inconsistencies in the breast surface position. The use of surface imaging tools highlights these discrepancies, and allows the radiation oncology team to better assess the possible effects on treatment quality. PACS number: 87

An EPID based method for performing high accuracy calibration between an optical external marker tracking device and the LINAC reference frame

PURPOSE With the increasing use of external 3D optical tracking cameras to guide modern radiation therapy procedures, it has become vitally important to have an accurate camera to linear accelerator (LINAC) reference frame calibration. To eliminate errors present in current calibration procedures based on the manual hand alignment of a device using the light field crosshairs and in room guidance lasers, a semiautomated quantitative calibration approach requiring only use of an electronic portal imaging device (EPID) was developed. METHODS A phantom comprised of seven highly IR reflective plastic BBs was placed on the LINAC treatment couch and imaged with both a 3D stereoscopic IR imager and the on board megavoltage (MV) EPID imager. Having knowledge of the optically determined 3D positions and projected EPID images of the BBs, simulated annealing was used to optimize the location of the BBs in the LINAC frame using four different optimization functions. Singular value decomposition was then used to calculate the transformation matrix between the camera and LINAC reference frames. Results were then compared to a traditional camera calibration method for overall accuracy. RESULTS Using modeled data, the simulated annealing process was able to determine the actual locations of the BBs with a RMSE of 0.23 mm. Using projection images acquired with an MV imager, the process was able to determine locations of BBs within .26 mm. The results depend on the choice of optimization function. CONCLUSIONS Results show that the method can be used to provide highly accurate spatial registration between an external 3D imaging reference frame and the LINAC frame. The experimental MV imager results, while not as precise as the simulated results, exceed 1 mm accuracy and the current accepted AAPM TG-142 standard of ≤2 mm positioning accuracy.

Development of an automated region of interest selection method for 3D surface monitoring of head motion.

PURPOSE To simplify the often complex and user-dependent manual region of interest (ROI) selection process for head motion monitoring, an automatic ROI selection method was developed. METHODS The automatic ROI selection algorithm calculated the displacements and velocities of 3D surface points between a temporally correlated 3D image series and a reference image. Only facial surfaces satisfying certain spatial and temporal criteria were selected. The algorithm was tested on five healthy volunteers instructed to perform different types of facial movements for a total of 27 real-time image sets (40-120 images for each image set). RESULTS The algorithm detected and excluded surface areas affected by different types of local facial movements that were independent of actual net head motion. Eye, eyebrow, and mandible motion were most commonly detected as being independent of head motion and were excluded from the final ROI. For 3D images taken with substantial facial or whole head motion, either most of the facial area was excluded or only small areas with random patterns were included in the final ROI. Surface image registration using iterative closest point (ICP) methods showed more stable real-time head tracking using the automatically selected ROI than manual user defined ROIs. CONCLUSIONS The automatic selection method successfully found ROIs stable over time for tracking head motion by excluding locally varying facial motions. By automating the ROI selection process, it is feasible that the time and complexity of current ROI definition can be reduced, together with user-dependent registration errors.

SU‐FF‐J‐107: Extraction of Internal and External Marker 3D‐Motion in Liver Patients with Compression Belt Using KV Cone‐Beam Radiographic Projections

Purpose: To study correlation of internal implanted vs. external skin markers for tracking respiratory motion in liver patients using radiographic projections from on‐board kV cone‐beam scans. Material and Method: Cone‐beam projections were analyzed to extract 3D‐motion of internal and external markers for five liver patients receiving hypofractionated radiotherapy. Patients were immobilized in a stereotactic body frame and an abdominal compression belt was used to constrain respiratory motion. Marker motion was derived using a tracking algorithm and analysis of a sequence of 650 cone‐beam projections acquired during a 1 minute scan. Corrections were made for imager rotation and sag. Results: External and internal markers had the same frequency of respiratory motion, however, the amplitude of external marker motion is smaller. Internal and external marker motion was also out‐of‐phase in some patients. Internal marker motion is greater in the superior‐inferior direction than in anterior‐posterior and right‐left directions, which is due to compression belt constraining of respiratory motion in these direction. Two patients showed small or no motion of the external marker, whereas, internal marker motion was as large as 1.0 cm, which may be due to the proximity of the external marker to the compression belt. Conclusions: Although, the motions of internal and external markers are usually correlated and have similar motion frequency, the amplitude of marker motion may differ significantly and in some patients markers may move our‐of‐phase. The abdominal compression belt suppresses respiratory motion strongly normal to patient skin and may contribute to phase differences. The external markers motion for monitoring internal changes of respiration should be used with caution. Marker 3D‐motion from cone‐beam projections provides real time tumor trajectory that can be used to determine accuracy of PTV margins with no extra dose other than that used in CBCTimaging.Conflict of Interest: Supported by NCI Grant P01‐CA59017.

Automated calculation of point A coordinates for CT-based high-dose-rate brachytherapy of cervical cancer

Purpose The goal is to develop a stand-alone application, which automatically and consistently computes the coordinates of the dose calculation point recommended by the American Brachytherapy Society (i.e., point A) based solely on the implanted applicator geometry for cervical cancer brachytherapy. Material and methods The application calculates point A coordinates from the source dwell geometries in the computed tomography (CT) scans, and outputs the 3D coordinates in the left and right directions. The algorithm was tested on 34 CT scans of 7 patients treated with high-dose-rate (HDR) brachytherapy using tandem and ovoid applicators. A single experienced user retrospectively and manually inserted point A into each CT scan, whose coordinates were used as the “gold standard” for all comparisons. The gold standard was subtracted from the automatically calculated points, a second manual placement by the same experienced user, and the clinically used point coordinates inserted by multiple planners. Coordinate differences and corresponding variances were compared using nonparametric tests. Results Automatically calculated, manually placed, and clinically used points agree with the gold standard to < 1 mm, 1 mm, 2 mm, respectively. When compared to the gold standard, the average and standard deviation of the 3D coordinate differences were 0.35 ± 0.14 mm from automatically calculated points, 0.38 ± 0.21 mm from the second manual placement, and 0.71 ± 0.44 mm from the clinically used point coordinates. Both the mean and standard deviations of the 3D coordinate differences were statistically significantly different from the gold standard, when point A was placed by multiple users (p < 0.05) but not when placed repeatedly by a single user or when calculated automatically. There were no statistical differences in doses, which agree to within 1-2% on average for all three groups. Conclusions The study demonstrates that the automated algorithm calculates point A coordinates consistently, while reducing inter-user variability. Point placement using the algorithm expedites the planning process and minimizes associated potential human errors.

SU-F-J-19: Robust Region-Of-Interest (ROI) for Consistent Registration On Deteriorated Surface Images

PURPOSE For African-American patients receiving breast radiotherapy with a bolus, skin darkening can affect the surface visualization when using optical imaging for daily positioning and gating at deep-inspiration breath holds (DIBH). Our goal is to identify a region-of-interest (ROI) that is robust against deteriorating surface image quality due to skin darkening. METHODS We study four patients whose post-mastectomy surfaces are imaged daily with AlignRT (VisionRT, UK) for DIBH radiotherapy and whose surface image quality is degraded toward the end of treatment. To simulate the effects of skin darkening, surfaces from the first ten fractions of each patient are systematically degraded by 25-35%, 40-50% and 65-75% of the total area of the clinically used ROI_ipsilateral-chestwall. The degraded surfaces are registered to the reference surface in six degrees-of-freedom. To identify a robust ROI, three additional reference ROIs - ROI_chest+abdomen, ROI_bilateral-chest and ROI_extended-ipsilateral-chestwall are created and registered to the degraded surfaces. Differences in registration using these ROIs are compared to that using ROI_ipsilateral-chestwall. RESULTS For three patients, the deviations in the registrations to ROI_ipsilateral-chestwall are > 2.0, 3.1 and 7.9mm on average for 25-35%, 40-50% and 65-75% degraded surfaces, respectively. Rotational deviations reach 11.1° in pitch. For the last patient, registration is consistent to within 2.6mm even on the 65-75% degraded surfaces, possibly because the surface topography has more distinct features. For ROI_bilateral-chest and ROI_extended-ipsilateral-chest registrations deviate in a similar pattern. However, registration on ROI_chest+abdomen is robust to deteriorating image qualities to within 4.2mm for all four patients. CONCLUSION Registration deviations using ROI_ipsilateral-chestwall can reach 9.8mm on the 40-50% degraded surfaces. Caution is required when using AlignRT for patients experiencing skin darkening since the accuracy of AlignRT registration deteriorates. To avoid this inaccuracy, we recommend use of ROI_chest+abdomen, on which registration is consistent within 4.2mm even for highly degraded surfaces.

SU-F-J-123: CT-Based Determination of DIBH Variability and Its Dosimetric Impact On Post-Mastectomy Plus Regional Nodal Radiation Therapy

PURPOSE Breast cancer radiotherapy delivered using voluntary deep inspiration breath-hold (DIBH) requires reproducible breath holds, particularly when matching supraclavicular fields to tangential fields. We studied the impact of variation in DIBHs on CTV and OAR dose metrics by comparing the dose distribution computed on two DIBH CT scans taken at the time of simulation. METHODS Ten patients receiving 50Gy in 25 fractions to the left chestwall and regional lymph nodes were studied. Two simulation CT scans were taken during separate DIBHs along with a free-breathing (FB) scan. The treatment was planned using one DIBH CT. The dose was recomputed on the other two scans using adaptive planning (Pinnacle 9.10) in which the scans are registered using a cross-correlation algorithm. The chestwall, lymph nodes and OARs were contoured on the scans following the RTOG consensus guidelines. The overall translational and rotational variation between the DIBH scans was used to estimate positional variation between breath-holds. Dose metrics between plans were compared using paired t-tests (p < 0.05) and means and standard deviations were reported. RESULTS The registration parameters were sub-millimeter and sub-degree. Although DIBH significantly reduced mean heart dose by 2.4Gy compared to FB (p < 0.01), no significant changes in dose were observed for targets or OARs between the two DIBH scans. Nodal coverage as assessed by V90% was 90%±8% and 89%±8% for supraclavicular and 99%±2% and 97%±22% for IM nodes. Though a significant decrease (10.5%±12.4%) in lung volume in the second DIBH CT was observed, the lung V20Gy was unchanged (14±2% and 14±3%) between the two DIBH scans. CONCLUSION While the lung volume often varied between DIBHs, the CTV and OAR dose metrics were largely unchanged. This indicates that manual DIBH has the potential to provide consistent dose delivery to the chestwall and regional nodes targets when using matched fields.

SU-E-T-141: Automated Dose Point Placement for Cervical Cancer Brachytherapy Using Tandem and Ovoid Applicators

Purpose: To develop a standalone application, which automatically and consistently calculates the coordinates of points A and H based solely on the implanted applicator geometry for cervical cancer HDR brachytherapy. Methods: Manchester point A and ABS point H are both located 2cm lateral from the central tandem plane. While both points are located 2cm above the cervical os, surrogates for the os differ. Point A is defined relative to the anatomical cervical os. Point H is defined relative to the intersection of the tandem with the superior aspects of the ovoids. The application takes an input text file generated by the treatment planning system (TPS, BrachyVision, Varian) that specifies the source geometries. It then outputs the 3D coordinates of points A and H in both the left and right directions. The algorithm was implemented and tested on 34 CT scans of 7 patients treated with HDR brachytherapy delivered using tandem and ovoids. A single experienced user retrospectively and manually placed points A and H on the CT scans, whose coordinates were used as the gold standard for the comparison to the automatically calculated points. Results: The automatically calculated coordinates of points A and H agree within 0.7mm with the gold standard. The averages and standard deviations of the 3D coordinate difference between points placed by the two methods are 0.3±0.1 and 0.4±0.1mm for points A and H, respectively. The maximum difference in 3D magnitude is 0.7mm. Conclusion: The algorithm consistently calculates dose point coordinates independently of the planner for cervical cancer brachytherapy treated with tandem and ovoids. Automated point placement based on the geometry of the implanted applicators agrees in sub-millimeter with careful manual placements by an experienced user. This algorithm expedites the planning process and eliminates dependencies on either user input or TPS visualization tools.

SU‐E‐T‐543: The Use of a Proportional‐Integral‐Derivative Design for Optimized Real‐Time Head Motion Correction in Frameless SRS

Purpose: While less invasive, the immobilization devices used in frameless SRS techniques generally allow for more head movement than a frame, leading to a larger PTV. To maintain traditional SRS accuracies, we have developed a robotic 3D head motion stage to correct for these sub‐ millimeter head deviations occurring during treatment. To achieve optimal head motion management we report on our use of a proportional‐integral‐ derivative (PID) feedback loop to monitor and compensate for real‐time patient motion during frameless SRS. Methods: A PID controller consists of three corrective parameters: a proportional term to account for the error difference between the desired patient position and the actual position; an integral term that accelerates the position correction toward the desired setpoint value; and a derivative term that accounts for the instantaneous rate of change of the patient position error. These parameters were tuned according to manual and heuristic Zielger‐Nichols methods. The PID algorithm was integrated into existing Labview software designed to control patient motion via a computer‐controlled 3D stepper motor stage. For realtime head position feedback either a Polaris external marker tracker or VisionRT system was used. The PID correction was tested both on phantom and healthy volunteers. Results: With volunteers in a relaxed supine position, the PID method maintained head motion to under 0.25 mm as based on the optical monitoring system in all three directions for greater than 90% of the time. Artificial introduction of extreme 1–5 mm head shifts by manual couch moving were corrected 2–3 times faster by PID when compared to our prior constant motor velocity approach Conclusions: In the event of patient head deviations, a properly tuned PID controller can offer significant time sparing advantages. This would allow for quicker initial setup after coarse couch positioning and a higher duty cycle during treatment.