Linac- and CyberKnife-based MRI-only treatment planning of prostate SBRT using an optimized synthetic CT calibration curve.

PURPOSE: CT Hounsfield Units (HUs) are converted to electron density using a calibration curve obtained from physical measurements of an electron density phantom. HU values assigned to an MRI-derived synthetic computed tomography (sCT) may present a different relationship with electron density compared to CT HU. Correct assignment of sCT HU values is critical for accurate dose calculation and delivery. The goals of this work were to develop a sCT calibration curve using patient data acquired on a clinically commissioned CT scanner and assess for CyberKnife- and volumetric modulated arc therapy (VMAT)-based MR-only treatment planning of prostate SBRT. METHODS: Same-day CT and MRI simulation in the treatment position were performed on 10 patients treated with SBRT to the prostate. Dixon in-phase and out-of-phase MRIs were acquired on a 3T scanner using a 3D T1-weighted gradient-echo sequence to generate sCTs using a commercial sCT algorithm. CT and sCT datasets were co-registered and HU values compared using mean absolute error (MAE). An optimized HU-to-density calibration curve was created based on average HU values across an institutional patient database for each of the four sCT tissue types. Clinical CyberKnife and VMAT treatment plans were generated on each patient CT and recomputed onto corresponding sCTs. Dose distributions computed using CT and sCT were compared using gamma criteria and dose-volume-histograms. RESULTS: For the optimized calibration curve, HU values were -96, 37, 204, and 1170 and relative electron densities were 0.95, 1.04, 1.1, and 1.7 for adipose, soft tissue, inner bone, and outer bone, respectively. The proposed sCT protocol produced total MAE of 94 ± 20HU. Gamma values mean ± std (min-max) were 98.9% ± 0.9% (97.1%-100%) and 97.7% ± 1.3% (95.3%-99.3%) for VMAT and CyberKnife plans, respectively. CONCLUSION: MRI-derived sCT using the proposed approach shows excellent dosimetric agreement with conventional CT simulation, demonstrating the feasibility of MRI-derived sCT for prostate SBRT treatment planning.

Predicting successful clinical candidates for fiducial-free lung tumor tracking with a deep learning binary classification model.

OBJECTIVES: The CyberKnife system is a robotic radiosurgery platform that allows the delivery of lung SBRT treatments using fiducial-free soft-tissue tracking. However, not all lung cancer patients are eligible for lung tumor tracking. Tumor size, density, and location impact the ability to successfully detect and track a lung lesion in 2D orthogonal X-ray images. The standard workflow to identify successful candidates for lung tumor tracking is called Lung Optimized Treatment (LOT) simulation, and involves multiple steps from CT acquisition to the execution of the simulation plan on CyberKnife. The aim of the study is to develop a deep learning classification model to predict which patients can be successfully treated with lung tumor tracking, thus circumventing the LOT simulation process. METHODS: Target tracking is achieved by matching orthogonal X-ray images with a library of digital radiographs reconstructed from the simulation CT scan (DRRs). We developed a deep learning model to create a binary classification of lung lesions as being trackable or untrackable based on tumor template DRR extracted from the CyberKnife system, and tested five different network architectures. The study included a total of 271 images (230 trackable, 41 untrackable) from 129 patients with one or multiple lung lesions. Eighty percent of the images were used for training, 10% for validation, and the remaining 10% for testing. RESULTS: For all five convolutional neural networks, the binary classification accuracy reached 100% after training, both in the validation and the test set, without any false classifications. CONCLUSIONS: A deep learning model can distinguish features of trackable and untrackable lesions in DRR images, and can predict successful candidates for fiducial-free lung tumor tracking.

Influence of respiratory motion management technique on radiation pneumonitis risk with robotic stereotactic body radiation therapy.

PURPOSE/OBJECTIVES: For lung stereotactic body radiation therapy (SBRT), real-time tumor tracking (RTT) allows for less radiation to normal lung compared to the internal target volume (ITV) method of respiratory motion management. To quantify the advantage of RTT, we examined the difference in radiation pneumonitis risk between these two techniques using a normal tissue complication probability (NTCP) model. MATERIALS/METHOD: 20 lung SBRT treatment plans using RTT were replanned with the ITV method using respiratory motion information from a 4D-CT image acquired at the original simulation. Risk of symptomatic radiation pneumonitis was calculated for both plans using a previously derived NTCP model. Features available before treatment planning that identified significant increase in NTCP with ITV versus RTT plans were identified. RESULTS: Prescription dose to the planning target volume (PTV) ranged from 22 to 60 Gy in 1-5 fractions. The median tumor diameter was 3.5 cm (range 2.1-5.5 cm) with a median volume of 14.5 mL (range 3.6-59.9 mL). The median increase in PTV volume from RTT to ITV plans was 17.1 mL (range 3.5-72.4 mL), and the median increase in PTV/lung volume ratio was 0.46% (range 0.13-1.98%). Mean lung dose and percentage dose-volumes were significantly higher in ITV plans at all levels tested. The median NTCP was 5.1% for RTT plans and 8.9% for ITV plans, with a median difference of 1.9% (range 0.4-25.5%, pairwise P < 0.001). Increases in NTCP between plans were best predicted by increases in PTV volume and PTV/lung volume ratio. CONCLUSIONS: The use of RTT decreased the risk of radiation pneumonitis in all plans. However, for most patients the risk reduction was minimal. Differences in plan PTV volume and PTV/lung volume ratio may identify patients who would benefit from RTT technique before completing treatment planning.

Investigating the clinical advantages of a robotic linac equipped with a multileaf collimator in the treatment of brain and prostate cancer patients.

The purpose of this study was to evaluate the performance of a commercially avail-able CyberKnife system with a multileaf collimator (CK-MLC) for stereotactic body radiotherapy (SBRT) and standard fractionated intensity-modulated radiotherapy (IMRT) applications. Ten prostate and ten intracranial cases were planned for the CK-MLC. Half of these cases were compared with clinically approved SBRT plans generated for the CyberKnife with circular collimators, and the other half were compared with clinically approved standard fractionated IMRT plans generated for conventional linacs. The plans were compared on target coverage, conformity, homogeneity, dose to organs at risk (OAR), low dose to the surrounding tissue, total monitor units (MU), and treatment time. CK-MLC plans generated for the SBRT cases achieved more homogeneous dose to the target than the CK plans with the circular collimators, for equivalent coverage, conformity, and dose to OARs. Total monitor units were reduced by 40% to 70% and treatment time was reduced by half. The CK-MLC plans generated for the standard fractionated cases achieved prescription isodose lines between 86% and 93%, which was 2%-3% below the plans generated for conventional linacs. Compared to standard IMRT plans, the total MU were up to three times greater for the prostate (whole pelvis) plans and up to 1.4 times greater for the intracranial plans. Average treatment time was 25min for the whole pelvis plans and 19 min for the intracranial cases. The CK-MLC system provides significant improvements in treatment time and target homogeneity compared to the CK system with circular collimators, while main-taining high conformity and dose sparing to critical organs. Standard fractionated plans for large target volumes (> 100 cm3) were generated that achieved high prescription isodose levels. The CK-MLC system provides more efficient SRS and SBRT treatments and, in select clinical cases, might be a potential alternative for standard fractionated treatments.

End-to-end validation of fiducial tracking accuracy in robotic radiosurgery using MRI-only simulation imaging.

BACKGROUND: Image-guided radiation-therapy (IGRT)-based robotic radiosurgery using magnetic resonance imaging (MRI)-only simulation could allow for improved target definition with highly conformal radiotherapy treatments. Fiducial marker (FM)-based alignment is used with robotic radiosurgery treatments of sites such as the prostate because it aids in accurate target localization. Synthetic CT (sCT) images are generated in the MRI-only workflow but FMs used for IGRT appear as signal voids in MRIs and do not appear in MR-generated sCTs, hindering the ability to use sCTs for fiducial-based IGRT. PURPOSE: In this study we evaluate the fiducial tracking accuracy for a novel artificial fiducial insertion method in sCT images that allows for fiducial marker tracking in robotic radiosurgery, using MRI-only simulation imaging (MRI-only workflow). METHODS: Artificial fiducial markers were inserted into sCT images at the site of the real marker implantation as visible in MRI. Two phantoms were used in this study. A custom anthropomorphic pelvis phantom was designed to validate the tracking accuracy for a variety of artificial fiducials in an MRI-only workflow. A head phantom containing a hidden target and orthogonal film pair inserts was used to perform end-to-end tests of artificial fiducial configurations inserted in sCT images. The setup and end-to-end targeting accuracy of the MRI-only workflow were compared to the computed tomography (CT)-based standard. Each phantom had six FMs implanted with a minimum spacing of 2 cm. For each phantom a bulk-density sCT was generated, and artificial FMs were inserted at the implantation location. Several methods of FM insertion were tested including: (1) replacing HU with a fixed value (10000HU) (voxel-burned); (2) using a representative fiducial image derived from a linear combination of fiducial templates (composite-fiducial); (3) computationally simulating FM signal voids using a digital phantom containing FMs and inserting the corresponding signal void into sCT images (simulated-fiducial). All tests were performed on a CyberKnife system (Accuray, Sunnyvale, CA). Treatment plans and digital-reconstructed-radiographs were generated from the original CT and sCTs with embedded fiducials and used to align the phantom on the treatment couch. Differences in the initial phantom alignment (3D translations/rotations) and tracking parameters between CT-based plans and sCT-based plans were analyzed. End-to-end plans for both scenarios were generated and analyzed following our clinical protocol. RESULTS: For all plans, the fiducial tracking algorithm was able to identify the fiducial locations. The mean FM-extraction uncertainty for the composite and simulated FMs was below 48% for fiducials in both the anthropomorphic pelvis and end-to-end phantoms, which is below the 70% treatment uncertainty threshold. The total targeting error was within tolerance (<0.95 mm) for end-to-end tests of sCT images with the composite and head-on simulated FMs (0.26, 0.44, and 0.35 mm for the composite fiducial in sCT, head-on simulated fiducial in sCT, and fiducials in original CT, respectively. CONCLUSIONS: MRI-only simulation for robotic radiosurgery could potentially improve treatment accuracy and reduce planning margins. Our study has shown that using a composite-derived or simulated FM in conjunction with sCT images, MRI-only workflow can provide clinically acceptable setup accuracy in line with CT-based standards for FM-based robotic radiosurgery.

Deep match: A zero-shot framework for improved fiducial-free respiratory motion tracking.

BACKGROUND AND PURPOSE: Motion management is essential to reduce normal tissue exposure and maintain adequate tumor dose in lung stereotactic body radiation therapy (SBRT). Lung SBRT using an articulated robotic arm allows dynamic tracking during radiation dose delivery. Two stereoscopic X-ray tracking modes are available - fiducial-based and fiducial-free tracking. Although X-ray detection of implanted fiducials is robust, the implantation procedure is invasive and inapplicable to some patients and tumor locations. Fiducial-free tracking relies on tumor contrast, which challenges the existing tracking algorithms for small (e.g., <15 mm) and/or tumors obscured by overlapping anatomies. To markedly improve the performance of fiducial-free tracking, we proposed a deep learning-based template matching algorithm - Deep Match. METHOD: Deep Match consists of four self-definable stages - training-free feature extractor, similarity measurements for location proposal, local refinements, and uncertainty level prediction for constructing a more trustworthy and versatile pipeline. Deep Match was validated on a 10 (38 fractions; 2661 images) patient cohort whose lung tumor was trackable on one X-ray view, while the second view did not offer sufficient conspicuity for tumor tracking using existing methods. The patient cohort was stratified into subgroups based on tumor sizes (<10 mm, 10-15 mm, and >15 mm) and tumor locations (with/without thoracic anatomy overlapping). RESULTS: On X-ray views that conventional methods failed to track the lung tumor, Deep Match achieved robust performance as evidenced by >80 % 3 mm-Hit (detection within 3 mm superior/inferior margin from ground truth) for 70 % of patients and <3 mm superior/inferior distance (SID) ∼1 mm standard deviation for all the patients. CONCLUSION: Deep Match is a zero-shot learning network that explores the intrinsic neural network benefits without training on patient data. With Deep Match, fiducial-free tracking can be extended to more patients with small tumors and with tumors obscured by overlapping anatomy.

A digital male pelvis phantom series showing anatomical variations over the course of fractionated radiotherapy treatment.

BACKGROUND: Daily IGRT images show day-to-day anatomical variations in patients undergoing fractionated prostate radiotherapy. This is of particular importance in particle beam treatments. PURPOSE: To develop a digital phantom series showing variation in pelvic anatomy for evaluating treatment planning and IGRT procedures in particle radiotherapy. METHODS: A pelvic phantom series was developed from the planning MRI and kVCT (planning CT) images along with six of the daily serial MVCT images taken of a single patient treated with a full bladder on a Tomotherapy unit. The selected patient had clearly visible yet unexceptional internal anatomy variation. Prostate, urethra, bladder, rectum, bowel, bowel gas, bone and soft tissue were contoured and a single Hounsfield Unit was assigned to each region. Treatment plans developed on the kVCT for photon, proton and carbon beams were recalculated on each phantom to demonstrate a clinical application of the series. Proton plans were developed with and without robust optimization. RESULTS: Limited to axial slices with prostate, the bladder volume varied from 6 to 46 cm3, the rectal volume (excluding gas) from 22 to 52 cm3, and rectal gas volume from zero to 18 cm3. The water equivalent path length to the prostate varied by up to 1.5 cm . The variations resulted in larger changes in the RBE-weighted Dose Volume Histograms of the non-robust proton plan and the carbon plan compared to the robust proton plan, the latter similar to the photon plan. The prostate coverage (V100%) decreased by an average of 18% in the carbon plan, 16% in the non-robust proton plan, 1.8% in the robust proton plan, and 4.4% in the photon plan. The volume of rectum receiving 75% of the prescription dose (V75%) increased by an average of 3.7 cm3, 4.7 cm3, 1.9 cm3, and 0.6 cm3 in those four plans, respectively. CONCLUSIONS: The digital pelvic phantom series provides for quantitative investigation of IGRT procedures and new methods for improving accuracy in particle therapy and may be used in cross-institutional comparisons for clinical trial quality assurance.

Improving plan quality and consistency by standardization of dose constraints in prostate cancer patients treated with CyberKnife.

Treatment plans for prostate cancer patients undergoing stereotactic body radiation therapy (SBRT) are often challenging due to the proximity of organs at risk. Today, there are no objective criteria to determine whether an optimal treatment plan has been achieved, and physicians rely on their personal experience to evaluate the plan's quality. In this study, we propose a method for determining rectal and bladder dose constraints achievable for a given patient's anatomy. We expect that this method will improve the overall plan quality and consistency, and facilitate comparison of clinical outcomes across different institutions. The 3D proximity of the organs at risk to the target is quantified by means of the expansion-intersection volume (EIV), which is defined as the intersection volume between the target and the organ at risk expanded by 5 mm. We determine a relationship between EIV and relevant dosimetric parameters, such as the volume of bladder and rectum receiving 75% of the prescription dose (V75%). This relationship can be used to establish institution-specific criteria to guide the treatment planning and evaluation process. A database of 25 prostate patients treated with CyberKnife SBRT is used to validate this approach. There is a linear correlation between EIV and V75% of bladder and rectum, confirming that the dose delivered to rectum and bladder increases with increasing extension and proximity of these organs to the target. This information can be used during the planning stage to facilitate the plan optimization process, and to standardize plan quality and consistency. We have developed a method for determining customized dose constraints for prostate patients treated with robotic SBRT. Although the results are technology specific and based on the experience of a single institution, we expect that the application of this method by other institutions will result in improved standardization of clinical practice.

Improved accuracy of relative electron density and proton stopping power ratio through CycleGAN machine learning.

Objective. Kilovoltage computed tomography (kVCT) is the cornerstone of radiotherapy treatment planning for delineating tissues and towards dose calculation. For the former, kVCT provides excellent contrast and signal-to-noise ratio. For the latter, kVCT may have greater uncertainty in determining relative electron density (ρe) and proton stopping power ratio (SPR). Conversely, megavoltage CT (MVCT) may result in superior dose calculation accuracy. The purpose of this work was to convert kVCT HU to MVCT HU using deep learning to obtain higher accuracyρeand SPR.Approach. Tissue-mimicking phantoms were created to compare kVCT- and MVCT-determinedρeand SPR to physical measurements. Using 100 head-and-neck datasets, an unpaired deep learning model was trained to learn the relationship between kVCTs and MVCTs, creating synthetic MVCTs (sMVCTs). Similarity metrics were calculated between kVCTs, sMVCTs, and MVCTs in 20 test datasets. An anthropomorphic head phantom containing bone-mimicking material with known composition was scanned to provide an independent determination ofρeand SPR accuracy by sMVCT.Main results. In tissue-mimicking bone,ρeerrors were 2.20% versus 0.19% and SPR errors were 4.38% versus 0.22%, for kVCT versus MVCT, respectively. Compared to MVCT,in vivomean difference (MD) values were 11 and 327 HU for kVCT and 2 and 3 HU for sMVCT in soft tissue and bone, respectively.ρeMD decreased from 1.3% to 0.35% in soft tissue and 2.9% to 0.13% in bone, for kVCT and sMVCT, respectively. SPR MD decreased from 1.8% to 0.24% in soft tissue and 6.8% to 0.16% in bone, for kVCT and sMVCT, respectively. Relative to physical measurements,ρeand SPR error in anthropomorphic bone decreased from 7.50% and 7.48% for kVCT to <1% for both MVCT and sMVCT.Significance. Deep learning can be used to map kVCT to sMVCT, suggesting higher accuracyρeand SPR is achievable with sMVCT versus kVCT.

Improved contrast and noise of megavoltage computed tomography (MVCT) through cycle-consistent generative machine learning.

PURPOSE: Megavoltage computed tomography (MVCT) has been implemented on many radiation therapy treatment machines as a tomographic imaging modality that allows for three-dimensional visualization and localization of patient anatomy. Yet MVCT images exhibit lower contrast and greater noise than its kilovoltage CT (kVCT) counterpart. In this work, we sought to improve these disadvantages of MVCT images through an image-to-image-based machine learning transformation of MVCT and kVCT images. We demonstrated that by learning the style of kVCT images, MVCT images can be converted into high-quality synthetic kVCT (skVCT) images with higher contrast and lower noise, when compared to the original MVCT. METHODS: Kilovoltage CT and MVCT images of 120 head and neck (H&N) cancer patients treated on an Accuray TomoHD system were retrospectively analyzed in this study. A cycle-consistent generative adversarial network (CycleGAN) machine learning, a variant of the generative adversarial network (GAN), was used to learn Hounsfield Unit (HU) transformations from MVCT to kVCT images, creating skVCT images. A formal mathematical proof is given describing the interplay between function sensitivity and input noise and how it applies to the error variance of a high-capacity function trained with noisy input data. Finally, we show how skVCT shares distributional similarity to kVCT for various macro-structures found in the body. RESULTS: Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were improved in skVCT images relative to the original MVCT images and were consistent with kVCT images. Specifically, skVCT CNR for muscle-fat, bone-fat, and bone-muscle improved to 14.8 ± 0.4, 122.7 ± 22.6, and 107.9 ± 22.4 compared with 1.6 ± 0.3, 7.6 ± 1.9, and 6.0 ± 1.7, respectively, in the original MVCT images and was more consistent with kVCT CNR values of 15.2 ± 0.8, 124.9 ± 27.0, and 109.7 ± 26.5, respectively. Noise was significantly reduced in skVCT images with SNR values improving by roughly an order of magnitude and consistent with kVCT SNR values. Axial slice mean (S-ME) and mean absolute error (S-MAE) agreement between kVCT and MVCT/skVCT improved, on average, from -16.0 and 109.1 HU to 8.4 and 76.9 HU, respectively. CONCLUSIONS: A kVCT-like qualitative aid was generated from input MVCT data through a CycleGAN instance. This qualitative aid, skVCT, was robust toward embedded metallic material, dramatically improves HU alignment from MVCT, and appears perceptually similar to kVCT with SNR and CNR values equivalent to that of kVCT images.

Stereotactic Body Radiation Therapy and High-Dose-Rate Brachytherapy Boost in Combination With Intensity Modulated Radiation Therapy for Localized Prostate Cancer: A Single-Institution Propensity Score Matched Analysis.

PURPOSE: To perform a propensity-score matched analysis comparing stereotactic body radiation therapy (SBRT) boost and high-dose-rate (HDR) boost for localized prostate cancer. METHODS AND MATERIALS: A single-institution retrospective chart review was conducted of men treated with pelvic external beam radiation therapy (EBRT) and SBRT boost (21 Gy and 19 Gy in 2 fractions) to the prostate for prostate cancer. A cohort treated at the same institution with HDR brachytherapy boost (19 Gy in 2 fractions) was compared. Propensity-score (PS) matching and multivariable Cox regression were used for analysis. Outcomes were biochemical recurrence freedom (BCRF) and metastasis freedom (MF). RESULTS: One hundred thirty-one men were treated with SBRT boost and 101 with HDR boost with median follow-up of 73.4 and 186.0 months, respectively. In addition, 68.8% of men had high-risk and 26.0% had unfavorable-intermediate disease, and 94.3% received androgen deprivation therapy. Five- and 10-year unadjusted BCRF was 88.8% and 85.3% for SBRT and 91.8% and 74.6% for HDR boost (log-rank P = .3), and 5- and 10-year unadjusted MF was 91.7% and 84.3% for SBRT and 95.8% and 82.0% for HDR (log-rank P = .8). After adjusting for covariates, there was no statistically significant difference in BCRF (hazard ratio [HR] 0.81; 95% confidence interval [CI], 0.37-1.79; P = .6) or MF (HR 1.07; 95% CI, 0.44-2.57; P = .9) between SBRT and HDR boost. Similarly, after PS matching, there was no statistically significant difference between SBRT and HDR (BCRF: HR 0.66, 0.27-1.62, P = .4; MF: HR 0.84, 0.31-2.26, P = .7). Grade 3+ genitourinary and gastrointestinal toxicity in the SBRT cohort were 4.6% and 1.5%, and 3.0% and 0.0% in the HDR cohorts (P = .4, Fisher exact test). CONCLUSIONS: SBRT boost plus pelvic EBRT for prostate cancer resulted in similar BCRF and MF to HDR boost in this single institution, PS matched retrospective analysis. Toxicity was modest. Prospective evaluation of SBRT boost for the treatment of unfavorable-intermediate and high-risk prostate cancer is warranted.

Irradiation of Nf1 mutant mouse models of spinal plexiform neurofibromas drives pathologic progression and decreases survival.

BACKGROUND: Genetically susceptible individuals can develop malignancies after irradiation of normal tissues. In the context of therapeutic irradiation, it is not known whether irradiating benign neoplasms in susceptible individuals promotes neoplastic transformation and worse clinical outcomes. Individuals with Neurofibromatosis 1 (NF1) are susceptible to both radiation-induced second malignancies and spontaneous progression of plexiform neurofibromas (PNs) to malignant peripheral nerve sheath tumors (MPNSTs). The role of radiotherapy in the treatment of benign neoplasms such as PNs is unclear. METHODS: To test whether radiotherapy promotes neoplastic progression of PNs and reduces overall survival, we administered spinal irradiation (SI) to conditional knockout mouse models of NF1-associated PNs in 2 germline contexts: Nf1 fllfl ; PostnCre + and Nf1 fl/- ; PostnCre + . Both genotypes develop extensive Nf1 null spinal PNs, modeling PNs in NF1 patients. A total of 101 mice were randomized to 0 Gy, 15 Gy (3 Gy × 5), or 30 Gy (3 Gy × 10) of spine-focused, fractionated SI and aged until signs of illness. RESULTS: SI decreased survival in both Nf1 fllfl mice and Nf1 fl/- mice, with the worst overall survival occurring in Nf1 fl/- mice receiving 30 Gy. SI was also associated with increasing worrisome histologic features along the PN-MPNST continuum in PNs irradiated to higher radiation doses. CONCLUSIONS: This preclinical study provides experimental evidence that irradiation of pre-existing PNs reduces survival and may shift PNs to higher grade neoplasms.

Technical Note: Performance of CyberKnife® tracking using low-dose CT and kV imaging.

PURPOSE: To investigate the effects of CT protocol and in-room x-ray technique on CyberKnife® (Accuray Inc.) tracking accuracy by evaluating end-to-end tests. METHODS: End-to-end (E2E) tests were performed for the different tracking methods (6D skull, fiducial, spine, and lung) using an anthropomorphic head phantom (Accuray Inc.) and thorax phantom (CIRS Inc.). Bolus was added to the thorax phantom to simulate a large patient and to evaluate the performance of lung tracking in a more realistic condition. The phantoms were scanned with a Siemens Sensation Open 24 slice CT at low dose (120 kV, 70 mAs, 1.5 mm slice thickness) and high dose (120 kV, 700 mAs, 1.5 mm slice thickness) to generate low-dose and high-dose digitally reconstructed radiographs (DRRs). The difference in initial phantom alignment, Δ(Align), and in total targeting accuracy, E2E, were obtained for all tracking methods with low- and high-dose DRRs. Additionally, Δ(Align) was determined for different in-room x-ray imaging techniques (0.5 to 50 mAs and 100 to 140 kV) using a low-dose lung tracking plan. RESULTS: Low-dose CT scans produced images with high noise; however, for these phantoms the targets could be easily delineated on all scans. End-to-end results were less than 0.95 mm for all tracking methods and all plans. The greatest difference in initial alignment Δ(Align) and E2E results between low- and high-dose CT protocols was 0.32 and 0.24 mm, respectively. Similar results were observed with a large thorax phantom. Tracking using different in-room x-ray imaging techniques (mAs) corresponding to low exposures (resulting in high image noise) or high exposure (resulting in image saturation) had alignment accuracy Δ(Align) greater than 1 mm. CONCLUSIONS: End-to-end targeting accuracy within tolerance (<0.95 mm) was obtained for all tracking methods using low-dose CT protocols, suggesting that CT protocol should be set by target contouring needs. Additionally, high tracking accuracy was achieved for in-room x-ray imaging techniques that produce high-quality images.

DoseGAN: a generative adversarial network for synthetic dose prediction using attention-gated discrimination and generation.

Deep learning algorithms have recently been developed that utilize patient anatomy and raw imaging information to predict radiation dose, as a means to increase treatment planning efficiency and improve radiotherapy plan quality. Current state-of-the-art techniques rely on convolutional neural networks (CNNs) that use pixel-to-pixel loss to update network parameters. However, stereotactic body radiotherapy (SBRT) dose is often heterogeneous, making it difficult to model using pixel-level loss. Generative adversarial networks (GANs) utilize adversarial learning that incorporates image-level loss and is better suited to learn from heterogeneous labels. However, GANs are difficult to train and rely on compromised architectures to facilitate convergence. This study suggests an attention-gated generative adversarial network (DoseGAN) to improve learning, increase model complexity, and reduce network redundancy by focusing on relevant anatomy. DoseGAN was compared to alternative state-of-the-art dose prediction algorithms using heterogeneity index, conformity index, and various dosimetric parameters. All algorithms were trained, validated, and tested using 141 prostate SBRT patients. DoseGAN was able to predict more realistic volumetric dosimetry compared to all other algorithms and achieved statistically significant improvement compared to all alternative algorithms for the V100 and V120 of the PTV, V60 of the rectum, and heterogeneity index.

Physical performance and image optimization of megavoltage cone-beam CT.

Megavoltage cone-beam CT (MVCBCT) is the most recent addition to the in-room CT systems developed for image-guided radiation therapy. The first generation MVCBCT system consists of a 6 MV treatment x-ray beam produced by a conventional linear accelerator equipped with a flat panel amorphous silicon detector. The objective of this study was to evaluate the physical performance of MVCBCT in order to optimize the system acquisition and reconstruction parameters for image quality. MVCBCT acquisitions were performed with the clinical system but images were reconstructed and analyzed with a separate research workstation. The geometrical stability and the positioning accuracy of the system were evaluated by comparing geometrical calibrations routinely performed over a period of 12 months. The beam output and detector intensity stability during MVCBCT acquisition were also evaluated by analyzing in-air acquisitions acquired at different exposure levels. Several system parameters were varied to quantify their impact on image quality including the exposure (2.7, 4.5, 9.0, 18.0, and 54.0 MU), the craniocaudal imaging length (2, 5, 15, and 27.4 cm), the voxel size (0.5, 1, and 2 mm), the slice thickness (1, 3, and 5 mm), and the phantom size. For the reconstruction algorithm, the study investigated the effect of binning, averaging and diffusion filtering of raw projections as well as three different projection filters. A head-sized water cylinder was used to measure and improve the uniformity of MVCBCT images. Inserts of different electron densities were placed in a water cylinder to measure the contrast-to-noise ratio (CNR). The spatial resolution was obtained by measuring the point-spread function of the system using an iterative edge blurring technique. Our results showed that the geometric stability and accuracy of MVCBCT were better than 1 mm over a period of 12 months. Beam intensity variations per projection of up to 35.4% were observed for a 2.7 MU MVCBCT acquisition. These variations did not cause noticeable reduction in the image quality. The results on uniformity suggest that the cupping artifact occurring with MVCBCT is mostly due to off-axis response of the detector and not scattered radiation. Simple uniformity correction methods were developed to nearly eliminate this cupping artifact. The spatial resolution of the baseline MVCBCT reconstruction protocol was approximately 2 mm. An optimized reconstruction protocol was developed and showed an improvement of 75% in CNR with a penalty of only 8% in spatial resolution. Using this new reconstruction protocol, large adipose and muscular structures were differentiated at an exposure of 9 MU. A reduction of 36% in CNR was observed on a larger (pelvic-sized) phantom. This study demonstrates that soft-tissue visualization with MVCBCT can be substantially improved with proper system settings. Further improvement is expected from the next generation MVCBCT system with an optimized megavoltage imaging beamline.

Whole-procedure clinical accuracy of gamma knife treatments of large lesions.

The mechanical accuracy of Gamma Knife radiosurgery based on single-isocenter measurement has been established to within 0.3 mm. However, the full delivery accuracy for Gamma Knife treatments of large lesions has only been estimated via the quadrature-sum analysis. In this study, the authors directly measured the whole-procedure accuracy for Gamma Knife treatments of large lesions to examine the validity of such estimation. The measurements were conducted on a head-phantom simulating the whole treatment procedure that included frame placement, computed tomography imaging, treatment planning, and treatment delivery. The results of the measurements were compared with the dose calculations from the treatment planning system. Average agreements of 0.1-1.6 mm for the isodose lines ranging from 25% to 90% of the maximum dose were found despite potentially large contributing uncertainties such as 3-mm imaging resolution, 2-mm dose grid size, 1-mm frame registration, multi-isocenter deliveries, etc. The results of our measurements were found to be significantly smaller (>50%) than the calculated value based on the quadrature-sum analysis. In conclusion, Gamma Knife treatments of large lesions can be delivered much more accurately than predicted from the quadrature-sum analysis of major sources of uncertainties from each step of the delivery chain.

Effect of composite sector collimation on average dose fall-off for Gamma Knife Perfexion.

OBJECT: The new capability of composite sector collimation in Gamma Knife Perfexion produces complex, nonspherical, and nonelliptical dose distributions. In this study, the authors investigated the effect of composite sector collimation on average dose fall-off compared with the previous Gamma Knife model. METHODS: A general formalism was derived to describe the peripheral dose distribution of all Gamma Knife models in the form of (V/V(0)) = (D/D(0))(gamma), where V is the volume of the peripheral isodose line with the value of D, V(0) is the reference prescription isodose volume, D(0) is the prescription dose, and gamma is the fitting parameter that determines how fast the dose falls off near the target. Based on this formula, the authors compared 40 cases involving patients treated with Gamma Knife Perfexion with 40 similar cases involving patients treated with Gamma Knife model 4C. The cases were grouped based on the use of the sector collimators in the treatment planning process. For each group as well as all cases combined, the mean gamma values were compared by means of the Student t-test for varying ranges of the peripheral dose distribution-from 100% of the prescription dose to 75, 50, and 25% of the prescription dose. RESULTS: The fit of general formula to the data was excellent for both Gamma Knife Perfexion and Gamma Knife 4C with R(2)> 0.99 for all the cases. The overall gamma values (mean +/- 2 standard deviations) were as follows: gamma = -1.74 +/- 0.47 (Model 4C) versus -1.77 +/- 0.40 (Perfexion) within 100-75% of the prescription dose; gamma = -1.57 +/- 0.26 (Model 4C) versus -1.58 +/- 0.25 (Perfexion) within 100-50% of the prescription dose; gamma = -1.47 +/- 0.18 (Model 4C) versus -1.50 +/- 0.16 (Perfexion) within 100-25% of the prescription dose. No statistical significance between the mean differences for Gamma Knife Perfexion and Model 4C was found within these ranges. The probability values were 0.65, 0.84, and 0.22, respectively. CONCLUSIONS: The use of composite sector collimators in Gamma Knife Perfexion demonstrated no statistically significant effects on the volume-averaged dose fall-off near a target periphery for typical treatment cases.

DoseNet: a volumetric dose prediction algorithm using 3D fully-convolutional neural networks.

The goal of this study is to demonstrate the feasibility of a novel fully-convolutional volumetric dose prediction neural network (DoseNet) and test its performance on a cohort of prostate stereotactic body radiotherapy (SBRT) patients. DoseNet is suggested as a superior alternative to U-Net and fully connected distance map-based neural networks for non-coplanar SBRT prostate dose prediction. DoseNet utilizes 3D convolutional downsampling with corresponding 3D deconvolutional upsampling to preserve memory while simultaneously increasing the receptive field of the network. DoseNet was implemented on 2 Nvidia 1080 Ti graphics processing units and utilizes a 3 phase learning protocol to help achieve convergence and improve generalization. DoseNet was trained, validated, and tested with 151 patients following Kaggle completion rules. The dosimetric quality of DoseNet was evaluated by comparing the predicted dose distribution with the clinically approved delivered dose distribution in terms of conformity index, heterogeneity index, and various clinically relevant dosimetric parameters. The results indicate that the DoseNet algorithm is a superior alternative to U-Net and fully connected methods for prostate SBRT patients. DoseNet required ~50.1 h to train, and ~0.83 s to make a prediction on a 128 × 128 × 64 voxel image. In conclusion, DoseNet is capable of making accurate volumetric dose predictions for non-coplanar SBRT prostate patients, while simultaneously preserving computational efficiency.

A continuous arc delivery optimization algorithm for CyberKnife m6.

PURPOSE: This study aims to reduce the delivery time of CyberKnife m6 treatments by allowing for noncoplanar continuous arc delivery. To achieve this, a novel noncoplanar continuous arc delivery optimization algorithm was developed for the CyberKnife m6 treatment system (CyberArc-m6). METHODS AND MATERIALS: CyberArc-m6 uses a five-step overarching strategy, in which an initial set of beam geometries is determined, the robotic delivery path is calculated, direct aperture optimization is conducted, intermediate MLC configurations are extracted, and the final beam weights are computed for the continuous arc radiation source model. This algorithm was implemented on five prostate and three brain patients, previously planned using a conventional step-and-shoot CyberKnife m6 delivery technique. The dosimetric quality of the CyberArc-m6 plans was assessed using locally confined mutual information (LCMI), conformity index (CI), heterogeneity index (HI), and a variety of common clinical dosimetric objectives. RESULTS: Using conservative optimization tuning parameters, CyberArc-m6 plans were able to achieve an average CI difference of 0.036 ± 0.025, an average HI difference of 0.046 ± 0.038, and an average LCMI of 0.920 ± 0.030 compared with the original CyberKnife m6 plans. Including a 5 s per minute image alignment time and a 5-min setup time, conservative CyberArc-m6 plans achieved an average treatment delivery speed up of 1.545x ± 0.305x compared with step-and-shoot plans. CONCLUSIONS: The CyberArc-m6 algorithm was able to achieve dosimetrically similar plans compared to their step-and-shoot CyberKnife m6 counterparts, while simultaneously reducing treatment delivery times.

Dose calculation using megavoltage cone-beam CT.

PURPOSE: To demonstrate the feasibility of performing dose calculation on megavoltage cone-beam CT (MVCBCT) of head-and-neck patients in order to track the dosimetric errors produced by anatomic changes. METHODS AND MATERIALS: A simple geometric model was developed using a head-size water cylinder to correct an observed cupping artifact occurring with MVCBCT. The uniformity-corrected MVCBCT was calibrated for physical density. Beam arrangements and weights from the initial treatment plans defined using the conventional CT were applied to the MVCBCT image, and the dose distribution was recalculated. The dosimetric inaccuracies caused by the cupping artifact were evaluated on the water phantom images. An ideal test patient with no observable anatomic changes and a patient imaged with both CT and MVCBCT before and after considerable weight loss were used to clinically validate MVCBCT for dose calculation and to determine the dosimetric impact of large anatomic changes. RESULTS: The nonuniformity of a head-size water phantom ( approximately 30%) causes a dosimetric error of less than 5%. The uniformity correction method developed greatly reduces the cupping artifact, resulting in dosimetric inaccuracies of less than 1%. For the clinical cases, the agreement between the dose distributions calculated using MVCBCT and CT was better than 3% and 3 mm where all tissue was encompassed within the MVCBCT. Dose-volume histograms from the dose calculations on CT and MVCBCT were in excellent agreement. CONCLUSION: MVCBCT can be used to estimate the dosimetric impact of changing anatomy on several structures in the head-and-neck region.