Weiguang Yao

Determining the optimal dosimetric leaf gap setting for rounded leaf‐end multileaf collimator systems by simple test fields

Individual QA for IMRT/VMAT plans is required by protocols. Sometimes plans cannot pass the institutes QA criteria. For the Eclipse treatment planning system (TPS) with rounded leaf‐end multileaf collimator (MLC), one practical way to improve the agreement of planned and delivered doses is to tune the value of dosimetric leaf gap (DLG) in the TPS from the measured DLG. We propose that this step may be necessary due to the complexity of the MLC system, including dosimetry of small fields and the tongue‐and‐groove (T&G) effects, and report our use of test fields to obtain linac‐specific optimal DLGs in TPSs. More than 20 original patient plans were reoptimized with the linac‐specific optimal DLG value. We examined the distribution of gaps and T&G extensions in typical patient plans and the effect of using the optimal DLG on the distribution. The QA pass rate of patient plans using the optimal DLG was investigated. The dose‐volume histograms (DVHs) of targets and organs at risk were checked. We tested three MLC systems (Varian millennium 120 MLC, high‐definition 120 MLC, and Siemens 160 MLC) installed in four Varian linear accelerators (linacs) (TrueBEAM STx, Trilogy, Clinac 2300 iX, and Clinac 21 EX) and 1 Siemens linac (Artiste). With an optimal DLG, the individual QA for all those patient plans passed the institutes criteria (95% in DTA test or gamma test with 3%/3 mm/10%), even though most of these plans had failed to pass QA when using original DLGs optimized from typical patient plans or from the optimization process (automodeler) of Pinnacle TPS. Using either our optimal DLG or one optimized from typical patient plans or from the Pinnacle optimization process yielded similar DVHs. PACS number: 87.55Qr

A robotic C-arm cone beam CT system for image-guided proton therapy: design and performance

OBJECTIVE A ceiling-mounted robotic C-arm cone beam CT (CBCT) system was developed for use with a 190° proton gantry system and a 6-degree-of-freedom robotic patient positioner. We report on the mechanical design, system accuracy, image quality, image guidance accuracy, imaging dose, workflow, safety and collision-avoidance. METHODS The robotic CBCT system couples a rotating C-ring to the C-arm concentrically with a kV X-ray tube and a flat-panel imager mounted to the C-ring. CBCT images are acquired with flex correction and maximally 360° rotation for a 53 cm field of view. The system was designed for clinical use with three imaging locations. Anthropomorphic phantoms were imaged to evaluate the image guidance accuracy. RESULTS The position accuracy and repeatability of the robotic C-arm was high (<0.5 mm), as measured with a high-accuracy laser tracker. The isocentric accuracy of the C-ring rotation was within 0.7 mm. The coincidence of CBCT imaging and radiation isocentre was better than 1 mm. The average image guidance accuracy was within 1 mm and 1° for the anthropomorphic phantoms tested. Daily volumetric imaging for proton patient positioning was specified for routine clinical practice. CONCLUSION Our novel gantry-independent robotic CBCT system provides high-accuracy volumetric image guidance for proton therapy. Advances in knowledge: Ceiling-mounted robotic CBCT provides a viable option than CT on-rails for partial gantry and fixed-beam proton systems with the added advantage of acquiring images at the treatment isocentre.

A correction scheme for a simplified analytical random walk model algorithm of proton dose calculation in distal Bragg peak regions.

The lateral homogeneity assumption is used in most analytical algorithms for proton dose, such as the pencil-beam algorithms and our simplified analytical random walk model. To improve the dose calculation in the distal fall-off region in heterogeneous media, we analyzed primary proton fluence near heterogeneous media and propose to calculate the lateral fluence with voxel-specific Gaussian distributions. The lateral fluence from a beamlet is no longer expressed by a single Gaussian for all the lateral voxels, but by a specific Gaussian for each lateral voxel. The voxel-specific Gaussian for the beamlet of interest is calculated by re-initializing the fluence deviation on an effective surface where the proton energies of the beamlet of interest and the beamlet passing the voxel are the same. The dose improvement from the correction scheme was demonstrated by the dose distributions in two sets of heterogeneous phantoms consisting of cortical bone, lung, and water and by evaluating distributions in example patients with a head-and-neck tumor and metal spinal implants. The dose distributions from Monte Carlo simulations were used as the reference. The correction scheme effectively improved the dose calculation accuracy in the distal fall-off region and increased the gamma test pass rate. The extra computation for the correction was about 20% of that for the original algorithm but is dependent upon patient geometry.

Feasibility study of range‐based registration using daily cone beam CT for intensity‐modulated proton therapy

PURPOSE Proton dose coverage is sensitive to proton beam range. The current practice of CT number-based registration for patient positioning focuses on matching the target and is not sufficient for proton therapy because the proton range depends on the medium traversed by the beam. Patient body deformations and anatomical changes result in range deviation in the target. We propose proton range-based registration to minimize the range deviation. METHODS The range was calculated from cone beam-computed tomography (CBCT) of the patient on couch, and the range deviation was the difference of the calculated range from that on the initial (day 1) CBCT. In the investigated prostate cases in which the main cause of range deviation was the rotation of femur bones, and in the investigated abdomen cases in which the main cause of range deviation was body growth and anatomic change, our range-based registration was used to obtain the optimal beam angle by minimizing the range deviation. The new angle was limited to be ±5° from that planned to prevent potentially increased dose to the organs at risk. To demonstrate the benefit of range-based registration, we investigated the range at the voxels on the surface of the target volume. The calculation error of range deviation due to CBCT scatter was investigated by using solid water phantoms with different thicknesses. Range-based registration using both CBCTs and CTs was performed in cases of two patients with pelvic rhabdomyosarcoma and one patient with upper abdominal tumor. The range was represented by the water-equivalent thickness to shorten the computation for online application purposes. RESULTS In the phantom study, the calculation error of range deviation due to CBCT scatter was within 2 mm for a 1-cm thickness change (the mean range deviation was 0.8 mm). In the CT study of the prostate cases, the range deviation (mean ± root-mean-square deviation) on the contour in each slice was efficiently reduced from 3.6 ± 2.8 mm to 2.1 ± 1.4 mm, with most slices being within 3 mm; in the CT study of the abdomen cases, the range deviation of the whole set was reduced from 4.4 ± 1.9 mm to 3.5 ± 2.1 mm. Both the mean and root-mean-square deviation of the range deviation on each treatment day were decreased. The dose coverage on the target was improved and the dose on the OARs was only slightly changed. CONCLUSION Range-based registration can efficiently mitigate range deviation due to patient positioning and anatomical changes. It can shorten patient positioning time and reduce the patients dose from CBCT.

SU‐F‐I‐09: Improvement of Image Registration Using Total‐Variation Based Noise Reduction Algorithms for Low‐Dose CBCT

PURPOSE To study the effect of total-variation based noise reduction algorithms to improve the image registration of low-dose CBCT for patient positioning in radiation therapy. METHODS In low-dose CBCT, the reconstructed image is degraded by excessive quantum noise. In this study, we developed a total-variation based noise reduction algorithm and studied the effect of the algorithm on noise reduction and image registration accuracy. To study the effect of noise reduction, we have calculated the peak signal-to-noise ratio (PSNR). To study the improvement of image registration, we performed image registration between volumetric CT and MV- CBCT images of different head-and-neck patients and calculated the mutual information (MI) and Pearson correlation coefficient (PCC) as a similarity metric. The PSNR, MI and PCC were calculated for both the noisy and noise-reduced CBCT images. RESULTS The algorithms were shown to be effective in reducing the noise level and improving the MI and PCC for the low-dose CBCT images tested. For the different head-and-neck patients, a maximum improvement of PSNR of 10 dB with respect to the noisy image was calculated. The improvement of MI and PCC was 9% and 2% respectively. CONCLUSION Total-variation based noise reduction algorithm was studied to improve the image registration between CT and low-dose CBCT. The algorithm had shown promising results in reducing the noise from low-dose CBCT images and improving the similarity metric in terms of MI and PCC.

SU-F-J-205: Effect of Cone Beam Factor On Cone Beam CT Number Accuracy

PURPOSE To examine the suitability of a Catphan™ 700 phantom for image quality QA of a cone beam computed tomography (CBCT) system deployed for proton therapy. METHODS Catphan phantoms, particularly Catphan™ 504, are commonly used in image quality QA for CBCT. As a newer product, Catphan™ 700 offers more tissue equivalent inserts which may be useful for generating the electron density - CT number curve for CBCT based treatment planning. The sensitometry-and-geometry module used in Catphan™ 700 is located at the end of the phantom and after the resolution line pair module. In Catphan™ 504 the line pair module is located at the end of the phantom and after the sensitometry-and-geometry module. To investigate the effect of difference in location on CT number accuracy due to the cone beam factor, we scanned the Catphan™ 700 with the central plane of CBCT at the center of the phantom, line pair and sensitometry-andgeometry modules of the phantom, respectively. The protocol head and thorax scan modes were used. For each position, scans were repeated 4 times. RESULTS For the head scan mode, the standard deviation (SD) of the CT numbers of each insert under 4 repeated scans was up to 20 HU, 11 HU, and 11 HU, respectively, for the central plane of CBCT located at the center of the phantom, line pair, and sensitometry-and-geometry modules of the phantom. The mean of the SD was 9.9 HU, 5.7 HU, and 5.9 HU, respectively. For the thorax mode, the mean of the SD was 4.5 HU, 4.4 HU, and 4.4 HU, respectively. The assessment of image quality based on resolution and spatial linearity was not affected by imaging location changes. CONCLUSION When the Catphan™ 700 was aligned to the center of imaging region, the CT number accuracy test may not meet expectations. We recommend reconfiguration of the modules.

TU-FG-BRB-11: Design and Evaluation of a Robotic C-Arm CBCT System for Image-Guided Proton Therapy

PURPOSE To describe the design and performance of a ceiling-mounted robotic C-arm CBCT system for image-guided proton therapy. METHODS Uniquely different from traditional C-arm CBCT used in interventional radiology, the imaging system was designed to provide volumetric image guidance for patients treated on a 190-degree proton gantry system and a 6 degree-of-freedom (DOF) robotic patient positioner. The mounting of robotic arms to the ceiling rails, rather than gantry or nozzle, provides the flexibility in imaging locations (isocenter, iso+27cm in X, iso+100cm in Y) in the room and easier upgrade as technology advances. A kV X-ray tube and a 43×43cm flat panel imager were mounted to a rotating C-ring (87cm diameter), which is coupled to the C-arm concentrically. Both C-arm and the robotic arm remain stationary during imaging to maintain high position accuracy. Source-to-axis distance and source-to-imager distance are 100 and 150cm, respectively. A 14:1 focused anti-scatter grid and a bowtie filer are used for image acquisition. A unique automatic collimator device of 4 independent blades for adjusting field of view and reducing patient dose has also been developed. RESULTS Sub-millimeter position accuracy and repeatability of the robotic C-arm were measured with a laser tracker. High quality CBCT images for positioning can be acquired with a weighted CTDI of 3.6mGy (head in 200° full fan mode: 100kV, 20mA, 20ms, 10fps)-8.7 mGy (pelvis in 360° half fan mode: 125kV, 42mA, 20ms, 10fps). Image guidance accuracy achieved <1mm (3D vector) with automatic 3D-3D registration for anthropomorphic head and pelvis phantoms. Since November 2015, 22 proton therapy patients have undergone daily CBCT imaging for 6 DOF positioning. CONCLUSION Decoupled from gantry and nozzle, this CBCT system provides a unique solution for volumetric image guidance with half/partial proton gantry systems. We demonstrated that daily CBCT can be integrated into proton therapy for pre-treatment position verification.

A multiscale filter for noise reduction of low-dose cone beam projections.

The Poisson or compound Poisson process governs the randomness of photon fluence in cone beam computed tomography (CBCT) imaging systems. The probability density function depends on the mean (noiseless) of the fluence at a certain detector. This dependence indicates the natural requirement of multiscale filters to smooth noise while preserving structures of the imaged object on the low-dose cone beam projection. In this work, we used a Gaussian filter, exp(-x2/2σ(2)(f)) as the multiscale filter to de-noise the low-dose cone beam projections. We analytically obtained the expression of σ(f), which represents the scale of the filter, by minimizing local noise-to-signal ratio. We analytically derived the variance of residual noise from the Poisson or compound Poisson processes after Gaussian filtering. From the derived analytical form of the variance of residual noise, optimal σ(2)(f)) is proved to be proportional to the noiseless fluence and modulated by local structure strength expressed as the linear fitting error of the structure. A strategy was used to obtain the reliable linear fitting error: smoothing the projection along the longitudinal direction to calculate the linear fitting error along the lateral direction and vice versa. The performance of our multiscale filter was examined on low-dose cone beam projections of a Catphan phantom and a head-and-neck patient. After performing the filter on the Catphan phantom projections scanned with pulse time 4 ms, the number of visible line pairs was similar to that scanned with 16 ms, and the contrast-to-noise ratio of the inserts was higher than that scanned with 16 ms about 64% in average. For the simulated head-and-neck patient projections with pulse time 4 ms, the visibility of soft tissue structures in the patient was comparable to that scanned with 20 ms. The image processing took less than 0.5 s per projection with 1024   ×   768 pixels.

SU‐C‐207‐05: A Comparative Study of Noise‐Reduction Algorithms for Low‐Dose Cone‐Beam Computed Tomography

Purpose: To study different noise-reduction algorithms and to improve the image quality of low dose cone beam CT for patient positioning in radiation therapy. Methods: In low-dose cone-beam CT, the reconstructed image is contaminated with excessive quantum noise. In this study, three well-developed noise reduction algorithms namely, a) penalized weighted least square (PWLS) method, b) split-Bregman total variation (TV) method, and c) compressed sensing (CS) method were studied and applied to the images of a computer–simulated “Shepp-Logan” phantom and a physical CATPHAN phantom. Up to 20% additive Gaussian noise was added to the Shepp-Logan phantom. The CATPHAN phantom was scanned by a Varian OBI system with 100 kVp, 4 ms and 20 mA. For comparing the performance of these algorithms, peak signal-to-noise ratio (PSNR) of the denoised images was computed. Results: The algorithms were shown to have the potential in reducing the noise level for low-dose CBCT images. For Shepp-Logan phantom, an improvement of PSNR of 2 dB, 3.1 dB and 4 dB was observed using PWLS, TV and CS respectively, while for CATPHAN, the improvement was 1.2 dB, 1.8 dB and 2.1 dB, respectively. Conclusion: Penalized weighted least square, total variation and compressed sensing methods were studied and compared for reducing the noise on a simulated phantom and a physical phantom scanned by low-dose CBCT. The techniques have shown promising results for noise reduction in terms of PSNR improvement. However, reducing the noise without compromising the smoothness and resolution of the image needs more extensive research.

SU‐F‐BRD‐09: A Random Walk Model Algorithm for Proton Dose Calculation

Purpose: To develop a random walk model algorithm for calculating proton dose with balanced computation burden and accuracy. Methods: Random walk (RW) model is sometimes referred to as a density Monte Carlo (MC) simulation. In MC proton dose calculation, the use of Gaussian angular distribution of protons due to multiple Coulomb scatter (MCS) is convenient, but in RW the use of Gaussian angular distribution requires an extremely large computation and memory. Thus, our RW model adopts spatial distribution from the angular one to accelerate the computation and to decrease the memory usage. From the physics and comparison with the MC simulations, we have determined and analytically expressed those critical variables affecting the dose accuracy in our RW model. Results: Besides those variables such as MCS, stopping power, energy spectrum after energy absorption etc., which have been extensively discussed in literature, the following variables were found to be critical in our RW model: (1) inverse squared law that can significantly reduce the computation burden and memory, (2) non-Gaussian spatial distribution after MCS, and (3) the mean direction of scatters at each voxel. In comparison to MC results, taken as reference, for a water phantom irradiated by mono-energetic proton beams from 75 MeV to 221.28 MeV, the gamma test pass rate was 100% for the 2%/2mm/10% criterion. For a highly heterogeneous phantom consisting of water embedded by a 10 cm cortical bone and a 10 cm lung in the Bragg peak region of the proton beam, the gamma test pass rate was greater than 98% for the 3%/3mm/10% criterion. Conclusion: We have determined key variables in our RW model for proton dose calculation. Compared with commercial pencil beam algorithms, our RW model much improves the dose accuracy in heterogeneous regions, and is about 10 times faster than MC simulations.