Equivalent-quality unflattened photon beam modeling, planning, and delivery.

The clinical application of the flattening filter-free photon beam (FFF) has enjoyed greater use due to its advantage of reduced treatment time because of the increased dose rate. Its unique beam characteristics, along with the very high-dose rate, require a thorough knowledge of the capability and accuracy in FFF beam modeling, planning, and delivery. This work verifies the feasibility of modeling an equivalent quality unflattened photon beam (eqUF), and the dosimetric accuracy in eqUF beam planning and delivery. An eqUF beam with a beam quality equivalent to a conventional 6 MV photon beam with the filter in place (WF) was modeled for the Pinnacle3 TPS and the beam model quality was evaluated by gamma index test. Results showed that the eqUF beam modeling was similar to that of the WF beam, as the overall passing rate of the 2%/2 mm gamma index test was 99.5% in the eqUF beam model and 96% in the WF beam model. Hypofractionated IMRT plans were then generated with the same constraints using both WF and eqUF beams, and the similarity was evaluated by DVH comparison and generalized 3D gamma index test. The WF and eqUF plans showed no clinically significant differences in DVH comparison and, on average > 98% voxels passed the 3%/3 mm 3D gamma index test. Dosimetric accuracy in gated phantom delivery was verified by ion chamber and film measurements. All ion chamber measurements at the isocenter were within 1% of calculated values and film measurements passed the 3 mm/3% gamma index test with an overall passing rate > 95% in the high-dose and low-gradient region in both WF and eqUF cases. Treatment plan quality assurance (QA), using either measurement-based or independent calculation-based methods of ten clinically treated eqUF IMRT plans were analyzed. In both methods, the point dose differences were all within 2% difference. In the relative 2D dose distribution comparison, >95% points were within 3% dose difference or 3 mm DTA.

Irradiator Commissioning and Dosimetry for Assessment of LQ α and ß Parameters, Radiation Dosing Schema, and in vivo Dose Deposition.

Radiation dosimetry is critical in the accurate delivery and reproducibility of radiation schemes in preclinical models for high translational relevance. Prior to performing any in vitro or in vivo experiments, the specific dose output for the irradiator and individual experimental designs must be assessed. Using an ionization chamber, electrometer, and solid water setup, the dose output of wide fields at isocenter can be determined. Using a similar setup with radiochromic films in the place of the ionization chamber, dose rates for smaller fields at different depths can also be determined. In vitro clonogenic survival assays of cancer cells in response to radiation treatment are inexpensive experiments that provide a measure of inherent radio-sensitivity of cell lines by fitting these data with the traditional linear-quadratic model. Model parameters estimated from these assays, combined with the principles of biologic effective doses, allows one to develop varying fractionation schedules for radiation treatment that provide equivalent effective doses in tumor-bearing animal experiments. This is an important factor to consider and correct for in comparing in vivo radiation therapy schedules to eliminate potential confounding of results due to variance in the delivered effective doses. Taken together, this article provides a general method for dose output verification preclinical animal and cabinet irradiators, in vitro assessment of radio-sensitivity, and verification of radiation delivery in small living organisms.

Optimized removal of the tongue-and-groove underdose via constrained partial synchronization and variable depth recursion.

Full synchronization (FS) leaf sequencing removes tongue-and-groove underdosages (TGU) but increases the number of segments. Constrained partial synchronization (CPS) uses a minimum tongue-and-groove ratio (TGR) to reduce the number of segments while achieving acceptable TGUs. TGR is the ratio of non-overlapping intensities that irradiate a common junction. For TGRs of 1, 1.5 and 2, the TGUs were measured as 18%, 4% and 0%, respectively, for a 6 MV beam and a Siemens 82 leaf MLC. The extraction and sweep processes of the variable depth recursion (VDR) leaf-sequencing algorithm were constrained to satisfy a minimum TGR. For a Siemens MLC and 15 clinical cases, VDR with a TGR = 1.5 produced 7% more segments than the unsynchronized VDR, while a fully synchronized sweeping window algorithm produces 62% more segments. For random intensity maps, VDR with CPS produced significantly fewer segments than an unsynchronized sweeping window. Similar results can be obtained for MLCs that interdigitate. This has implications for direct aperture optimization algorithms (DAO) that use the sweeping window as a starting point (Pinnacle), for which a significant TGU has been observed. The concept of CPS can be applied to DAO by choosing appropriate levels for each of the segments in DAO.

Deriving motion from megavoltage localization cone beam computed tomography scans.

Cone beam computed tomography (CBCT) projection data consist of views of a moving point (e.g. diaphragm apex). The point is selected in identification views of extreme motion (two inhale, two exhale). The room coordinates of the extreme points are determined by source-to-view ray tracing intersections. Projected to other views, these points become opposite corners of a motion-bounding box. The view coordinates of the point, relative to the box, are used to interpolate between extreme room coordinates. Along with the views' time stamps, this provides the point's room coordinates as a function of time. CBCT-derived trajectories of a tungsten pin, moving 3 cm cranio-caudally and 1 cm elsewhere, deviate from expected ones by at most 1.06 mm. When deviations from the ideal imaging geometry are considered, mean errors are less than 0.2 mm. While CBCT-derived cranio-caudal positions are insensitive to the choice of identification views, the bounding box determination requires view separations between 15 and 163 degrees . Inhale views with the two largest amplitudes should be used, though corrections can account for different amplitudes. The information could be used to calibrate motion surrogates, adaptively define phase triggers immediately before gated radiotherapy and provide phase and amplitude sorting for 4D CBCT.

Variable depth recursion algorithm for leaf sequencing.

The processes of extraction and sweep are basic segmentation steps that are used in leaf sequencing algorithms. A modified version of a commercial leaf sequencer changed the way that the extracts are selected and expanded the search space, but the modification maintained the basic search paradigm of evaluating multiple solutions, each one consisting of up to 12 extracts and a sweep sequence. While it generated the best solutions compared to other published algorithms, it used more computation time. A new, faster algorithm selects one extract at a time but calls itself as an evaluation function a user-specified number of times, after which it uses the bidirectional sweeping window algorithm as the final evaluation function. To achieve a performance comparable to that of the modified commercial leaf sequencer, 2-3 calls were needed, and in all test cases, there were only slight improvements beyond two calls. For the 13 clinical test maps, computation speeds improved by a factor between 12 and 43, depending on the constraints, namely the ability to interdigitate and the avoidance of the tongue-and-groove under dose. The new algorithm was compared to the original and modified versions of the commercial leaf sequencer. It was also compared to other published algorithms for 1400, random, 15 X 15, test maps with 3-16 intensity levels. In every single case the new algorithm provided the best solution.

Lung diaphragm tracking in CBCT images using spatio-temporal MRF.

In EBRT in order to monitor the intra fraction motion of thoracic and abdominal tumors, one of the standard approaches is to use the lung diaphragm apex as an internal marker. However, tracking the position of the apex from image based observations is a challenging problem, as it undergoes both position and shape variation. The purpose of this paper is to propose an alternative method for tracking the ipsi-lateral hemidiaphragm apex (IHDA) position on Cone Beam Computed Tomography (CBCT) projection images. A hierarchical method is proposed to track the IHDA position across the frames. The diaphragm state is modeled as a spatio-temporal Markov Random Field (MRF). The likelihood function is derived from the votes based on 4D-Hough space. The optimal state of the diaphragm is obtained by solving the associated energy minimization problem using graph-cuts. A heterogeneous GPU implementation is provided for the method using CUDA framework and the performance is compared with that of CPU implementation. The method was tested using 15 clinical CBCT images. The results demonstrate that the MRF formulation outperforms the full search method in terms of accuracy. The GPU based heterogeneous implementation of the proposed algorithm takes about 25s, which is 16% improvement over the existing benchmark. The proposed MRF formulation considers all the possible combinations from the 4D-Hough space and therefore results in better tracking accuracy. The GPU based implementation exploits the inherent parallelism in our algorithm to accelerate the performance thereby increasing the viability of the approach for clinical use.

Modifications to the IMFAST leaf sequencing optimization algorithm.

The optimizing leaf sequencer IMFAST minimizes intensity modulated treatment times. However, algorithm modifications can yield improved results. Currently, during segment extraction, the largest extract for a given number of levels is chosen. The modification chooses an extract that yields the fewest segments and levels when the rod-pushing algorithm is applied to the difference between the original map and the extract. Also, successive optimization parameter values are now allowed to increase. These modifications reduced the number of segments and the relative fluence from the original algorithm by an average of 7%-11% and 8%-17%, respectively, depending on whether interdigitation and/or tongue-and-groove constraints were considered. The tests were done on two clinical head and neck intensity modulated radiation therapy cases. Compared to the sweeping window algorithm, a reduction of 35%-55% of the number of segments is possible with a change in the relative fluence of -9% - 16%, depending on the constraints. Compared to other previously published algorithms that deal with the constraints tested here, the modified IMFAST algorithm provides the greatest reduction in the number of segments with the minimum increase in the relative fluence.

Quality assurance in IMRT: importance of the transmission through the jaws for an accurate calculation of absolute doses and relative distributions.

The goal of IMRT is to achieve an isodose distribution conformed to the tumor while avoiding the organs at risk. For these tasks several gantry angles are selected, each one containing a series of different leaf configurations for the multileaf collimator (MLC) (segments). Verifying the relative distributions as well as the absolute doses is an important step for quality assurance issues. We have observed that an accurate modeling of the transmission of the primary x-ray fluence through the jaws and MLC as well as the head scatter is crucial for a precise calculation of relative doses and monitor units. Also, an inaccurate calculation of the output factor for small size segments can lead to important differences in the absolute dose for points under these segments. Incorrect models could lead to systematic errors of around 5% to 10% in the calculated monitor units and a shift in the isodose curves.

A new method of diaphragm apex motion detection from 2D projection images of mega-voltage cone beam CT.

To present a new method of estimating 3D positions of the ipsi-lateral hemi-diaphragm apex (IHDA) from 2D projection images of mega-voltage cone beam CT (MVCBCT). The detection framework reconstructs a 3D volume from all the 2D projection images. An initial estimated 3D IHDA position is determined in this volume based on an imaging processing pipeline, including Otsu thresholding, connected component labeling and template matching. This initial position is then projected onto each 2D projection image to create a region of interest (ROI). To accurately detect the IHDA position in 2D projection space, two methods, dynamic Hough transform (DHT) and a tracking approach based on a joint probability density function (PDF) are developed. Both methods utilize a double-parabola model to fit the 2D diaphragm boundary. The 3D IHDA motion in the superior-inferior (SI) direction is estimated from the initial static 3D position and the detected 2D positions in projection space. The two Hough-based detection methods are tested on 35 MVCBCT scans from 15 patients. The detection is compared to manually identified IHDA positions in 2D projection space by three clinicians. An average and standard deviation of 4.252 ± 3.354 and 2.485 ± 1.750 mm was achieved for DHT and tracking-based approaches respectively, compared with the inter-expert variance among three experts of 1.822 ± 1.106 mm. Based on the results of the scans, the PDF tracking-based approach appears more robust than the DHT. The combination of the automatic ROI localization and the tracking-based approach is a quicker and more accurate method of extracting 3D IHDA motion from 2D projection images.

Motion-compensated mega-voltage cone beam CT using the deformation derived directly from 2D projection images.

This paper presents a novel method for respiratory motion compensated reconstruction for cone beam computed tomography (CBCT). The reconstruction is based on a time sequence of motion vector fields, which is generated by a dynamic geometrical object shape model. The dynamic model is extracted from the 2D projection images of the CBCT. The process of the motion extraction is converted into an optimal 3D multiple interrelated surface detection problem, which can be solved by computing a maximum flow in a 4D directed graph. The method was tested on 12 mega-voltage (MV) CBCT scans from three patients. Two sets of motion-artifact-free 3D volumes, full exhale (FE) and full inhale (FI) phases, were reconstructed for each daily scan. The reconstruction was compared with three other motion-compensated approaches based on quantification accuracy of motion and size. Contrast-to-noise ratio (CNR) was also quantified for image quality. The proposed approach has the best overall performance, with a relative tumor volume quantification error of 3.39 ± 3.64% and 8.57 ± 8.31% for FE and FI phases, respectively. The CNR near the tumor area is 3.85 ± 0.42 (FE) and 3.58 ± 3.33 (FI). These results show the clinical feasibility to use the proposed method to reconstruct motion-artifact-free MVCBCT volumes.

3D lung tumor motion model extraction from 2D projection images of mega-voltage cone beam CT via optimal graph search.

In this paper, we propose a novel method to convert segmentation of objects with quasi-periodic motion in 2D rotational cone beam projection images into an optimal 3D multiple interrelated surface detection problem, which can be solved by a graph search framework. The method is tested on lung tumor segmentation in projection images of mega-voltage cone beam CT (MVCBCT). A 4D directed graph is constructed based on an initialized tumor mesh model, where the cost value for this graph is computed from the point location of a silhouette outline of projected tumor mesh in 2D projection images. The method was first evaluated on four different sized phantom inserts (all above 1.9 cm in diameter) with a predefined motion of 3.0 cm to mimic the imaging of lung tumors. A dice coefficient of 0.87 +/- 0.03 and a centroid error of 1.94 +/- 1.31 mm were obtained. Results based on 12 MVCBCT scans from 3 patients obtained 0.91 +/- 0.03 for dice coefficient and 1.83 +/- 1.31 mm for centroid error, compared with a difference between two sets of independent manual contours of 0.89 +/- 0.03 and 1.61 +/- 1.19 mm, respectively.