Tim Fox

Integer Programming Applied to Intensity-Modulated Radiation Therapy Treatment Planning

In intensity-modulated radiation therapy (IMRT) not only is the shape of the beam controlled, but combinations of open and closed multileaf collimators modulate the intensity as well. In this paper, we offer a mixed integer programming approach which allows optimization over beamlet fluence weights as well as beam and couch angles. Computational strategies, including a constraint and column generator, a specialized set-based branching scheme, a geometric heuristic procedure, and the use of disjunctive cuts, are described. Our algorithmic design thus far has been motivated by clinical cases. Numerical tests on real patient cases reveal that good treatment plans are returned within 30 minutes. The MIP plans consistently provide superior tumor coverage and conformity, as well as dose homogeneity within the tumor region while maintaining a low irradiation to important critical and normal tissues.

MR‐based attenuation correction for hybrid PET‐MR brain imaging systems using deformable image registration

PURPOSEnRealization of combined positron emission tomography (PET)--magnetic resonance (MR) scanners has the potential to significantly change healthcare and revolutionize clinical practice as it allows, simultaneously, visualization of molecular imaging and anatomical imaging. PET-MR, acquired in one imaging study, will likely become the advanced imaging modality of choice for neurological studies, certain forms of cancer, stroke, and the emerging study of stem cell therapy. A challenge toward the implementation and operation of combined PET-MR scanners is that attenuation corrections maps are not directly available due to space and cost constraints. This article presents a method to obtain accurate patient-specific PET attenuation coefficients maps in head imaging by warping an atlas computed tomography (CT) data set to the patient-specific MR data set using a deformable registration model.nnnMETHODSnA multimodality optical flow deformable model has been developed that establishes a voxel-to-voxel correspondence between the CT atlas and patient MR images. Once the mapping is established, the atlas is warped with the deformation field obtained by the registration to create a simulated CT image study that matches the patient anatomy, which could be used for attenuation correction.nnnRESULTSnTo evaluate the accuracy of the deformable-based attenuation correction, 17 clinical brain tumor cases were studied using acquired MR-CT images. A simulated CT was compared to the patients true CT to assess geometrical accuracy of the deformation module as well as voxel-to-voxel comparison of Hounsfield units (HUs). In all cases, mapping from the atlas CT to the individual MR was achieved with geometrical accuracy as judged using quantitative inspection tools. The mean distance between simulated and true CT external contour and bony anatomy was 1.26 and 2.15 mm, respectively. In terms of HU unit comparison, the mean voxel-to-voxel difference was less than 2 HU for all cases.nnnCONCLUSIONSnAttenuation correction for hybrid PET-MR scanners was easily achieved by individualizing an atlas CT to the MR data set using a deformable model without requiring user interaction. The method provided clinical accuracy while eliminating the need for an additional CT scan for PET attenuation correction.

Patient‐specific quality assurance method for VMAT treatment delivery

Volumetric modulated arc therapy (VMAT) is a system for intensity-modulated radiotherapy treatment delivery that achieves high dose conformality by optimizing the dose rate, gantry speed, and the leaf positions of the dynamic multileaf collimator (DMLC). The aim of this work is to present a practical approach for patient-specific volumetric reconstruction of the dose delivered of a VMAT treatment using the DMLC and treatment controller log (Dynalog) files. The accuracy of VMAT delivery was analyzed for five prostate patients. For each patient, a clinical treatment was delivered and values recorded in the log files for the gantry angle, dose rate, and leaf positions were converted to a new DICOM-compliant plan using a custom-developed software system. The plan was imported in a treatment planning system and the dose distribution was recreated on the original CT by simply recomputing the dose. Using the standard evaluation tools, it is straightforward to assess if reconstructed dose meets clinical endpoints, as well as to compare side-by-side reconstructed and original plans. The study showed that log files can be directly used for dose reconstruction without resorting to phantom measurements or setups. In all cases, analysis of the leaf positions showed a maximum error of -0.26 mm (mean of 0.15 mm). Gantry angle deviation was less than 1degree and the total MU was within 0.5 from the planned value. Differences between the reconstructed and the intended dose matrices were less than 1.46% for all cases. Measurements using the MATRIXX system in a phantom were used to validate the dosimetric accuracy of the proposed method, with an agreement of at least 96% in all pixels as measured using the gamma index. The methodology provides a volumetric evaluation of the dose reconstructed by VMAT plans which is easily achieved by automated analysis of Dynalog files without additional measurements or phantom setups. This process provides a valuable platform for adaptive therapy in the future.

Evaluation of Automatic Atlas-Based Lymph Node Segmentation for Head-and-Neck Cancer

PURPOSEnTo evaluate if automatic atlas-based lymph node segmentation (LNS) improves efficiency and decreases inter-observer variability while maintaining accuracy.nnnMETHODS AND MATERIALSnFive physicians with head-and-neck IMRT experience used computed tomography (CT) data from 5 patients to create bilateral neck clinical target volumes covering specified nodal levels. A second contour set was automatically generated using a commercially available atlas. Physicians modified the automatic contours to make them acceptable for treatment planning. To assess contour variability, the Simultaneous Truth and Performance Level Estimation (STAPLE) algorithm was used to take collections of contours and calculate a probabilistic estimate of the true segmentation. Differences between the manual, automatic, and automatic-modified (AM) contours were analyzed using multiple metrics.nnnRESULTSnCompared with the true segmentation created from manual contours, the automatic contours had a high degree of accuracy, with sensitivity, Dice similarity coefficient, and mean/max surface disagreement values comparable to the average manual contour (86%, 76%, 3.3/17.4 mm automatic vs. 73%, 79%, 2.8/17 mm manual). The AM group was more consistent than the manual group for multiple metrics, most notably reducing the range of contour volume (106-430 mL manual vs. 176-347 mL AM) and percent false positivity (1-37% manual vs. 1-7% AM). Average contouring time savings with the automatic segmentation was 11.5 min per patient, a 35% reduction.nnnCONCLUSIONSnUsing the STAPLE algorithm to generate true contours from multiple physician contours, we demonstrated that, in comparison with manual segmentation, atlas-based automatic LNS for head-and-neck cancer is accurate, efficient, and reduces interobserver variability.

Quantitative evaluation of a cone-beam computed tomography-planning computed tomography deformable image registration method for adaptive radiation therapy

Deformable (non‐rigid) registration is an essential tool in both adaptive radiation therapy and image‐guided radiation therapy to account for soft‐tissue changes during the course of treatment. The evaluation method most commonly used to assess the accuracy of deformable image registration is qualitative human evaluation. Here, we propose a method for systematically measuring the accuracy of an algorithm in recovering artificially introduced deformations in cases of rigid geometry, and we use that method to quantify the ability of a modified basis spline (B‐Spline) registration algorithm to recover artificially introduced deformations. The evaluation method is entirely computer‐driven and eliminates biased interpretation associated with human evaluation; it can be applied to any chosen method of image registration. Our method involves using planning computed tomography (PCT) images acquired with a conventional CT simulator and cone‐beam computed tomography (CBCT) images acquired daily by a linear accelerator–mounted kilovoltage image system in the treatment delivery room. The deformation that occurs between the PCT and daily CBCT images is obtained using a modified version of the B‐Spline deformable model designed to overcome the low soft‐tissue contrast and the artifacts and distortions observed in CBCT images. Clinical CBCT images and contours of phantom and central nervous system cases were deformed (warped) with known random deformations. In registering the deformed with the non‐deformed image sets, we tracked the algorithms ability to recover the original, non‐deformed set. Registration error was measured as the mean and maximum difference between the original and the registered surface contours from outlined structures. Using this approach, two sets of tests can be devised. To measure the residual error related to the optimizers convergence performance, the warped CBCT image is registered to the unwarped version of itself, eliminating unknown factors such as noise and positioning errors. To study additional errors introduced by artifacts and noise in the CBCT image, the warped CBCT image is registered to the original PCT image. Using a B‐Spline deformable image registration algorithm, mean residual error introduced by the algorithms performance on noise‐free images was less than 1 mm, with a maximum of 2 mm. The chosen deformable image registration model was capable of accommodating significant variability in structures over time, because the artificially introduced deformation magnitude did not significantly influence the residual error. On the second type of test, noise and artifacts reduced registration accuracy to a mean of 1.33 mm and a maximum of 4.86 mm. The accuracy of deformable image registration can be easily and consistently measured by evaluating the algorithms ability to recover artificially introduced deformations in rigid cases in which the true solution is known a priori. The method is completely automated, applicable to any chosen registration algorithm, and does not require user interaction of any kind. PACS numbers: 87.57.Gg, 87.57.Ce, 87.62.+n

Optimization of radiosurgery treatment planning via mixed integer programming.

An automated optimization algorithm based on mixed integer programming techniques is presented for generating high-quality treatment plans for LINAC radiosurgery treatment. The physical planning in radiosurgery treatment involves selecting among a large collection of beams with different physical parameters an optimal beam configuration (geometries and intensities) to deliver the clinically prescribed radiation dose to the tumor volume while sparing the nearby critical structure and normal tissue. The proposed mixed integer programming models incorporate strict dose restrictions on tumor volume, and constraints on the desired number of beams, isocenters, couch angles, and gantry angles. The model seeks to deliver full prescription dose coverage and uniform radiation dose to the tumor volume while minimizing the excess radiation to the periphery normal tissue. In particular, it ensures that proximal normal tissues receive minimal dose via rapid dose fall-off. Preliminary numerical tests on a single patient case indicate that this approach can produce exceptionally high-quality plans in a fraction of the time required using the procedure currently employed by clinicians. The resulting plans provide highly uniform prescription dose to the tumor volume while drastically reducing the irradiation received by the proximal critical normal tissue.

Dosimetric performance of the new high‐definition multileaf collimator for intracranial stereotactic radiosurgery

The objective was to evaluate the performance of a high‐definition multileaf collimator (MLC) of 2.5 mm leaf width (MLC2.5) and compare to standard 5 mm leaf width MLC (MLC5) for the treatment of intracranial lesions using dynamic conformal arcs (DCA) technique with a dedicated radiosurgery linear accelerator. Simulated cases of spherical targets were created to study solely the effect of target volume size on the performance of the two MLC systems independent of target shape complexity. In addition, 43 patients previously treated for intracranial lesions in our institution were retrospectively planned using DCA technique with MLC2.5 and MLC5 systems. The gross tumor volume ranged from 0.07 to 40.57u2009cm3 with an average volume of 5.9u2009cm3. All treatment parameters were kept the same for both MLC‐based plans. The plan evaluation was performed using figures of merits (FOM) for a rapid and objective assessment on the quality of the two treatment plans for MLC2.5 and MLC5. The prescription isodose surface was selected as the greatest isodose surface covering ≥95% of the target volume and delivering 95% of the prescription dose to 99% of target volume. A Conformity Index (CI) and conformity distance index (CDI) were used to quantifying the dose conformity to a target volume. To assess normal tissue sparing, a normal tissue difference (NTD) was defined as the difference between the volume of normal tissue receiving a certain dose utilizing MLC5 and the volume receiving the same dose using MLC2.5. The CI and normal tissue sparing for the simulated spherical targets were better with the MLC2.5 as compared to MLC5. For the clinical patients, the CI and CDI results indicated that the MLC2.5 provides better treatment conformity than MLC5 even at large target volumes. The CIs range was 1.15 to 2.44 with a median of 1.59 for MLC2.5 compared to 1.60–2.85 with a median of 1.71 for MLC5. Improved normal tissue sparing was also observed for MLC2.5 over MLC5, with the NTD always positive, indicating improvement, and ranging from 0.1 to 8.3 for normal tissue receiving 50% (NTV50), 70% (NTV70) and 90% (NTV90) of the prescription dose. The MLC2.5 has a dosimetric advantage over the MLC5 in Linac‐based radiosurgery using DCA method for intracranial lesions, both in treatment conformity and normal tissue sparing when target shape complexity increases. PACS number: 87.56J‐, 87.56 jk

A registration-based approach for 4D cardiac micro-CT using combined prospective and retrospective gating

Recent advances in murine cardiac studies with three-dimensional cone beam micro-computed tomography (CT) have used either prospective or retrospective gating technique. While prospective gating ensures the best image quality and the highest resolution, it involves longer sampling times and higher radiation dose. Sampling is faster and the radiation dose can be reduced with retrospective gating but the image quality is affected by the limited number of projections with an irregular angular distribution which complicate the reconstruction process, causing significant streaking artifacts. This work involves both prospective and retrospective gating in sampling. Deformable registration is used between a high quality image set acquired with prospective gating with the multiple data sets during the cardiac cycle obtained using retrospective gating. Tests were conducted on a four-dimensional (4D) cardiac mouse phantom and after optimization, the method was applied to in vivo cardiac micro-CT data. Results indicate that, by using our method, the sampling time can be reduced by a factor of 2.5 and the radiation dose can be reduced 35% compared to the prospective sampling while the image quality can be maintained. In conclusion, we proposed a novel solution to 4D cine cardiac micro-CT based on a combined prospective with retrospective gating in sampling and deformable registration post reconstruction that mixed the advantages of both strategies.

Six degrees of freedom CBCT‐based positioning for intracranial targets treated with frameless stereotactic radiosurgery

Frameless radiosurgery is an attractive alternative to the framed procedure if it can be performed with comparable precision in a reasonable time frame. Here, we present a positioning approach for frameless radiosurgery based on in‐room volumetric imaging coupled with an advanced six‐degrees‐of‐freedom (6 DOF) image registration technique which avoids use of a bite block. Patient motion is restricted with a custom thermoplastic mask. Accurate positioning is achieved by registering a cone‐beam CT to the planning CT scan and applying all translational and rotational shifts using a custom couch mount. System accuracy was initially verified on an anthropomorphic phantom. Isocenters of delineated targets in the phantom were computed and aligned by our system with an average accuracy of 0.2 mm, 0.3 mm, and 0.4 mm in the lateral, vertical, and longitudinal directions, respectively. The accuracy in the rotational directions was 0.1°, 0.2°, and 0.1° in the pitch, roll, and yaw, respectively. An additional test was performed using the phantom in which known shifts were introduced. Misalignments up to 10 mm and 3° in all directions/rotations were introduced in our phantom and recovered to an ideal alignment within 0.2 mm, 0.3 mm, and 0.4 mm in the lateral, vertical, and longitudinal directions, respectively, and within 0.3° in any rotational axis. These values are less than couch motion precision. Our first 28 patients with 38 targets treated over 63 fractions are analyzed in the patient positioning phase of the study. Mean error in the shifts predicted by the system were less than 0.5 mm in any translational direction and less than 0.3° in any rotation, as assessed by a confirmation CBCT scan. We conclude that accurate and efficient frameless radiosurgery positioning is achievable without the need for a bite block by using our 6 DOF registration method. This system is inexpensive compared to a couch‐based 6 DOF system, improves patient comfort compared to systems that utilize a bite block, and is ideal for the treatment of pediatric patients with or without general anesthesia, as well as of patients with dental issues. From this study, it is clear that only adjusting for 4 DOF may, in some cases, lead to significant compromise in PTV coverage. Since performing the additional match with 6 DOF in our registration system only adds a relatively short amount of time to the overall process, we advocate making the precise match in all cases. PACS number: 87.55.tm; 87.55.Qr; 87.57.nj

Performance evaluation of an automated image registration algorithm using an integrated kilovoltage imaging and guidance system.

Image‐guided radiation therapy delivery may be used to assess the position of the tumor and anatomical structures within the body as opposed to relying on external marks. The purpose of this manuscript is to evaluate the performance of the image registration software for automatically detecting and repositioning a 3D offset of a phantom using a kilovoltage onboard imaging system. Verification tests were performed on both a geometric rigid phantom and an anthropomorphic head phantom containing a humanoid skeleton to assess the precision and accuracy of the automated positioning system. From the translation only studies, the average deviation between the detected and known offset was less than 0.75 mm for each of the three principal directions, and the shifts did not show any directional sensitivity. The results are given as the measurement with standard deviation in parentheses. The combined translations and rotations had the greatest average deviation in the lateral, longitudinal, and vertical directions. For all dimensions, the magnitude of the deviation does not appear to be correlated with the magnitude of the actual translation introduced. The On‐Board Imager™ (OBI) system has been successfully integrated into a feasible online radiotherapy treatment guidance procedure. Evaluation of each patients resulting automatch should be performed by therapists before each treatment session for adequate clinical oversight. PACS numbers: 87.53.‐j, 87.53.Xd, 87.57.Gg