Automated and Clinically Optimal Treatment Planning for Cancer Radiotherapy.

Each year, approximately 18 million new cancer cases are diagnosed worldwide, and about half must be treated with radiotherapy. A successful treatment requires treatment planning with the customization of penetrating radiation beams to sterilize cancerous cells without harming nearby normal organs and tissues. This process currently involves extensive manual tuning of parameters by an expert planner, making it a time-consuming and labor-intensive process, with quality and immediacy of critical care dependent on the planner's expertise. To improve the speed, quality, and availability of this highly specialized care, Memorial Sloan Kettering Cancer Center developed and applied advanced optimization tools to this problem (e.g., using hierarchical constrained optimization, convex approximations, and Lagrangian methods). This resulted in both a greatly improved radiotherapy treatment planning process and the generation of reliable and consistent high-quality plans that reflect clinical priorities. These improved techniques have been the foundation of high-quality treatments and have positively impacted over 4,000 patients to date, including numerous patients in severe pain and in urgent need of treatment who might have otherwise required longer hospital stays or undergone unnecessary surgery to control the progression of their disease. We expect that the wide distribution of the system we developed will ultimately impact patient care more broadly, including in resource-constrained countries.

Impact of a Centralized Database System on Radiation Therapy Quality Assurance Management at a Large Health Care Network: 5 Years' Experience.

PURPOSE: This study reports the impact of using a centralized database system for major equipment quality assurance (QA) at a large institution. METHODS AND MATERIALS: A centralized database system has been implemented for radiation therapy machine QA in our institution at 6 campuses with 11 computed tomographies and 22 linear accelerators (LINACs). The database system was customized to manage monthly and annual computed tomography and LINAC QA. This includes providing the same set of QA procedures across the enterprise, digitally storing all measurement records, and generating trend analyses. Compared with conventional methods (ie, paper forms), the effectiveness of the database system was quantified by changes in the compliance of QA tests and perceptions of staff to the efficiency of data retrieval and analyses. An anonymized questionnaire was provided to physicists enterprise-wide to assess workflow changes. RESULTS: With the implementation of the database system, the compliance of QA test completion improved from 80% to >99% for the entire institution. This resonates with the 56% of physicists who found the database system helpful in guiding them through QA, and 25% of physicists found the contrary, and 19% reported no difference (n = 16). Meanwhile, 40% of physicists reported longer times needed to record data using the database system compared with conventional methods, and another 40% suggested otherwise. In addition, 87% and 80% found the database more efficient to analyze and retrieve previous data, respectively. This was also reflected by the shorter time taken to generate year-end QA statistics using the software (5 vs 30 min per LINAC). Overall, 94% of physicists preferred the centralized database system over conventional methods and endorsed continued use of the system. CONCLUSIONS: A centralized database system is useful and can improve the effectiveness and efficiency of QA management in a large institution. With consistent data collection and proper data storage using a database, high-quality data can be obtained for failure modes and effects analyses as per TG 100.

Dosimetric evaluation of irradiation geometry and potential air gaps in an acrylic miniphantom used for external audit of absolute dose calibration for a hybrid 1.5 T MR-linac system.

INTRODUCTION: To investigate the impact of partial lateral scatter (LS), backscatter (BS) and presence of air gaps on optically stimulated luminescence dosimeter (OSLD) measurements in an acrylic miniphantom used for dosimetry audit on the 1.5 T magnetic resonance-linear accelerator (MR-linac) system. METHODS: The following irradiation geometries were investigated using OSLDs, A26 MR/A12 MR ion chamber (IC), and Monaco Monte Carlo system: (a) IC/OSLD in an acrylic miniphantom (partial LS, partial BS), (b) IC/OSLD in a miniphantom placed on a solid water (SW) stack at a depth of 1.5 cm (partial LS, full BS), (c) IC/OSLD placed at a depth of 1.5 cm inside a 3 cm slab of SW/buildup (full LS, partial BS), and (d) IC/OSLD centered inside a 3 cm slab of SW/buildup at a depth of 1.5 cm placed on top of a SW stack (full LS, full BS). Average of two irradiated OSLDs with and without water was used at each setup. An air gap of 1 and 2 mm, mimicking presence of potential air gap around the OSLDs in the miniphantom geometry was also simulated. The calibration condition of the machine was 1 cGy/MU at SAD = 143.5 cm, d = 5 cm, G90, and 10 × 10 cm2 . RESULTS: The Monaco calculation (0.5% uncertainty and 1.0 mm voxel size) for the four setups at the measurement point were 108.2, 108.1, 109.4, and 110.0 cGy. The corresponding IC measurements were 109.0 ± 0.03, 109.5 ± 0.06, 110.2 ± 0.02, and 109.8 ± 0.03 cGy. Without water, OSLDs measurements were ∼10% higher than the expected. With added water to minimize air gaps, the measurements were significantly improved to within 2.2%. The dosimetric impacts of 1 and 2 mm air gaps were also verified with Monaco to be 13.3% and 27.9% higher, respectively, due to the electron return effect. CONCLUSIONS: A minimal amount of air around or within the OSLDs can cause measurement discrepancies of 10% or higher when placed in a high b-field MR-linac system. Care must be taken to eliminate the air from within and around the OSLD.

Deep learning auto-segmentation and automated treatment planning for trismus risk reduction in head and neck cancer radiotherapy.

BACKGROUND AND PURPOSE: Reducing trismus in radiotherapy for head and neck cancer (HNC) is important. Automated deep learning (DL) segmentation and automated planning was used to introduce new and rarely segmented masticatory structures to study if trismus risk could be decreased. MATERIALS AND METHODS: Auto-segmentation was based on purpose-built DL, and automated planning used our in-house system, ECHO. Treatment plans for ten HNC patients, treated with 2 Gy × 35 fractions, were optimized (ECHO0). Six manually segmented OARs were replaced with DL auto-segmentations and the plans re-optimized (ECHO1). In a third set of plans, mean doses for auto-segmented ipsilateral masseter and medial pterygoid (MIMean, MPIMean), derived from a trismus risk model, were implemented as dose-volume objectives (ECHO2). Clinical dose-volume criteria were compared between the two scenarios (ECHO0 vs. ECHO1; ECHO1 vs. ECHO2; Wilcoxon signed-rank test; significance: p < 0.01). RESULTS: Small systematic differences were observed between the doses to the six auto-segmented OARs and their manual counterparts (median: ECHO1 = 6.2 (range: 0.4, 21) Gy vs. ECHO0 = 6.6 (range: 0.3, 22) Gy; p = 0.007), and the ECHO1 plans provided improved normal tissue sparing across a larger dose-volume range. Only in the ECHO2 plans, all patients fulfilled both MIMean and MPIMean criteria. The population median MIMean and MPIMean were considerably lower than those suggested by the trismus model (ECHO0: MIMean = 13 Gy vs. ≤42 Gy; MPIMean = 29 Gy vs. ≤68 Gy). CONCLUSIONS: Automated treatment planning can efficiently incorporate new structures from DL auto-segmentation, which results in trismus risk sparing without deteriorating treatment plan quality. Auto-planning and deep learning auto-segmentation together provide a powerful platform to further improve treatment planning.

Clinical experience and workflow challenges with magnetic resonance-only radiation therapy simulation and planning for prostate cancer.

BACKGROUND AND PURPOSE: Magnetic Resonance (MR)-only planning has been implemented clinically for radiotherapy of prostate cancer. However, fewer studies exist regarding the overall success rate of MR-only workflows. We report on successes and challenges of implementing MR-only workflows for prostate. MATERIALS AND METHODS: A total of 585 patients with prostate cancer underwent an MR-only simulation and planning between 06/2016-06/2018. MR simulation included images for contouring, synthetic-CT generation and fiducial identification. Workflow interruptions occurred that required a backup CT, a re-simulation or an update to our current quality assurance (QA) process. The challenges were prospectively evaluated and classified into syn-CT generation, motion/artifacts in the MRs, fiducial QA and bowel preparation guidelines. RESULTS: MR-only simulation was successful in 544 (93.2 %) patients. . In seventeen patients (2.9%), reconstruction of synthetic-CT failed due to patient size, femur angulation, or failure to determine the body contour. Twenty-four patients (4.1%) underwent a repeat/backup CT scan because of artifacts on the MR such as image blur due to patient motion or biopsy/surgical artifacts that hampered identification of the implanted fiducial markers. In patients requiring large coverage due to nodal involvement, inhomogeneity artifacts were resolved by using a two-stack acquisition and adaptive inhomogeneity correction. Bowel preparation guidelines were modified to address frequent rectum/gas issues due to longer MR scan time. CONCLUSIONS: MR-only simulation has been successfully implemented for a majority of patients in the clinic. However, MR-CT or CT-only pathway may still be needed for patients where MR-only solution fails or patients with MR contraindications.

A phantom study to evaluate three different registration platform of 3D/3D, 2D/3D, and 3D surface match with 6D alignment for precise image-guided radiotherapy.

PURPOSE: To evaluate two three-dimensional (3D)/3D registration platforms, one two-dimensional (2D)/3D registration method, and one 3D surface registration method (3DS). These three technologies are available to perform six-dimensional (6D) registrations for image-guided radiotherapy treatment. METHODS: Fiducial markers were asymmetrically placed on the surfaces of an anthropomorphic head phantom (n = 13) and a body phantom (n = 8), respectively. The point match (PM) solution to the six-dimensional (6D) transformation between the two image sets [planning computed tomography (CT) and cone beam CT (CBCT)] was determined through least-square fitting of the fiducial positions using singular value decomposition (SVD). The transformation result from SVD was verified and was used as the gold standard to evaluate the 6D accuracy of 3D/3D registration in Varian's platform (3D3DV), 3D/3D and 2D/3D registration in the BrainLab ExacTrac system (3D3DE and 2D3D), as well as 3DS in the AlignRT system. Image registration accuracy from each method was quantitatively evaluated by root mean square of target registration error (rmsTRE) on fiducial markers and by isocenter registration error (IRE). The Wilcoxon signed-rank test was utilized to compare the difference of each registration method with PM. A P < 0.05 was considered significant. RESULTS: rmsTRE was in the range of 0.4 mm/0.7 mm (cranial/body), 0.5 mm/1 mm, 1.0 mm/1.5 mm, and 1.0 mm/1.2 mm for PM, 3D3D, 2D3D, and 3DS, respectively. Comparing to PM, the mean errors of IRE were 0.3 mm/1 mm for 3D3D, 0.5 mm/1.4 mm for 2D3D, and 1.6 mm/1.35 mm for 3DS for the cranial and body phantoms respectively. Both of 3D3D and 2D3D methods differed significantly in the roll direction as compared to the PM method for the cranial phantom. The 3DS method was significantly different from the PM method in all three translation dimensions for both the cranial (P = 0.003-P = 0.03) and body (P < 0.001-P = 0.008) phantoms. CONCLUSION: 3D3D using CBCT had the best image registration accuracy among all the tested methods. 2D3D method was slightly inferior to the 3D3D method but was still acceptable as a treatment position verification device. 3DS is comparable to 2D3D technique and could be a substitute for X-ray or CBCT for pretreatment verification for treatment of anatomical sites that are rigid.

Strict bladder filling and rectal emptying during prostate SBRT: Does it make a dosimetric or clinical difference?

BACKGROUND: To evaluate inter-fractional variations in bladder and rectum during prostate stereotactic body radiation therapy (SBRT) and determine dosimetric and clinical consequences. METHODS: Eighty-five patients with 510 computed tomography (CT) images were analyzed. Median prescription dose was 40 Gy in 5 fractions. Patients were instructed to maintain a full bladder and empty rectum prior to simulation and each treatment. A single reviewer delineated organs at risk (OARs) on the simulation (Sim-CT) and Cone Beam CTs (CBCT) for analyses. RESULTS: Bladder and rectum volume reductions were observed throughout the course of SBRT, with largest mean reductions of 86.9 mL (19.0%) for bladder and 6.4 mL (8.7%) for rectum noted at fraction #5 compared to Sim-CT (P < 0.01). Higher initial Sim-CT bladder volumes were predictive for greater reduction in absolute bladder volume during treatment (ρ = - 0.69; P < 0.01). Over the course of SBRT, there was a small but significant increase in bladder mean dose (+ 4.5 ± 12.8%; P < 0.01) but no significant change in the D2cc (+ 0.8 ± 4.0%; P = 0.28). The mean bladder trigone displacement was in the anterior direction (+ 4.02 ± 6.59 mm) with a corresponding decrease in mean trigone dose (- 3.6 ± 9.6%; P < 0.01) and D2cc (- 6.2 ± 15.6%; P < 0.01). There was a small but significant increase in mean rectal dose (+ 7.0 ± 12.9%, P < 0.01) but a decrease in rectal D2cc (- 2.2 ± 10.1%; P = 0.04). No significant correlations were found between relative bladder volume changes, bladder trigone displacements, or rectum volume changes with rates of genitourinary or rectal toxicities. CONCLUSIONS: Despite smaller than expected bladder and rectal volumes at the time of treatment compared to the planning scans, dosimetric impact was minimal and not predictive of detrimental clinical outcomes. These results cast doubt on the need for excessively strict bladder filling and rectal emptying protocols in the context of image guided prostate SBRT and prospective studies are needed to determine its necessity.

Clinical Experience of Automated SBRT Paraspinal and Other Metastatic Tumor Planning With Constrained Hierarchical Optimization.

PURPOSE: We report on the clinical performance of a fully automated approach to treatment planning based on a Pareto optimal, constrained hierarchical optimization algorithm, named Expedited Constrained Hierarchical Optimization (ECHO). METHODS AND MATERIALS: From April 2017 to October 2018, ECHO produced 640 treated plans for 523 patients who underwent stereotactic body radiation therapy (RT) for paraspinal and other metastatic tumors. A total of 182 plans were for 24 Gy in a single fraction, 387 plans were for 27 Gy in 3 fractions, and the remainder were for other prescriptions or fractionations. Of the plans, 84.5% were for paraspinal tumors, with 69, 302, and 170 in the cervical, thoracic, and lumbosacral spine, respectively. For each case, after contouring, a template plan using 9 intensity modulated RT fields based on disease site and tumor location was sent to ECHO through an application program interface plug-in from the treatment planning system. ECHO returned a plan that satisfied all critical structure hard constraints with optimal target volume coverage and the lowest achievable normal tissue doses. Upon ECHO completion, the planner received an e-mail indicating the plan was ready for review. The plan was accepted if all clinical criteria were met. Otherwise, a limited number of parameters could be adjusted for another ECHO run. RESULTS: The median planning target volume size was 84.3 cm3 (range, 6.9-633.2). The median time to produce 1 ECHO plan was 63.5 minutes (range, 11-340 minutes) and was largely dependent on the field sizes. Of the cases, 79.7% required 1 run to produce a clinically accepted plan, 13.3% required 1 additional run with minimal parameter adjustments, and 7.0% required ≥2 additional runs with significant parameter modifications. All plans met or bettered the institutional clinical criteria. CONCLUSIONS: We successfully implemented automated stereotactic body RT paraspinal and other metastatic tumors planning. ECHO produced high-quality plans, improved planning efficiency and robustness, and enabled expedited treatment planning at our clinic.

Couch and multileaf collimator tracking: A clinical feasibility study for pancreas and liver treatment.

PURPOSE: Real-time tumor tracking through active correction by the multileaf collimator or treatment couch offers a promising strategy to mitigate delivery uncertainty due to intrafractional tumor motion. This study evaluated the performance of MLC and couch tracking using the prototype iTools Tracking system in TrueBeam Developer Mode and the application for abdominal cancer treatments. METHODS: Experiments were carried out using a phantom with embedded Calypso transponders and a motion simulation platform. Geometric evaluations were performed using a circular conformal field with sinusoidal traces and pancreatic tumor motion traces. Geometric tracking accuracy was retrospectively calculated by comparing the compensational MLC or couch motion extracted from machine log files to the target motion reconstructed from real-time MV and kV images. Dosimetric tracking accuracy was measured with radiochromic films using clinical abdominal VMAT plans and pancreatic tumor traces. RESULTS: Geometrically, the root-mean-square errors for MLC tracking were 0.5 and 1.8 mm parallel and perpendicular to leaf travel direction, respectively. Couch tracking, in contrast, showed an average of 0.8 mm or less geometric error in all directions. Dosimetrically, both MLC and couch tracking reduced motion-induced local dose errors compared to no tracking. Evaluated with five pancreatic tumor motion traces, the average 2%/2 mm global gamma pass rate of eight clinical abdominal VMAT plans was 67.4% (range: 26.4%-92.7%) without tracking, which was improved to 86.0% (range: 67.9%-95.6%) with MLC tracking, and 98.1% (range: 94.9%-100.0%) with couch tracking. In 16 out of 40 deliveries with different plans and motion traces, MLC tracking did not achieve clinically acceptable dosimetric accuracy with 3%/3mm gamma pass rate below 95%. CONCLUSIONS: This study demonstrated the capability of MLC and couch tracking to reduce motion-induced dose errors in abdominal cases using a prototype tracking system. Clinically significant dose errors were observed with MLC tracking for certain plans which could be attributed to the inferior MLC tracking accuracy in the direction perpendicular to leaf travel, as well as the interplay between motion tracking and plan delivery for highly modulated plans. Couch tracking outperformed MLC tracking with consistently high dosimetric accuracy in all plans evaluated, indicating its clinical potential in the treatment of abdominal cancers.

Impact of varying air cavity on planning dosimetry for rectum patients treated on a 1.5 T hybrid MR-linac system.

PURPOSE: To investigate the dosimetric impact of magnetic (B) field on varying air cavities in rectum patients treated on the hybrid 1.5 T MR-linac. METHODS: Artificial air cavities of varying diameters (0.0, 1.0, 1.5, 2.0, 2.5, 3.0, and 5.0 cm) were created for four rectum patients (two prone and two supine). A total of 56 plans using a 7 MV flattening filter-free beam were generated with and without B-field. Reference intensity-modulated radiation therapy treatment plans without air cavity in the presence and absence of B-field were generated to a total dose of 45/50 Gy. The reference plans were copied and recalculated for the varying air cavities. D95 (PTV45 -PTV50 ), D95 (PTV50 -aircavity), V50 (PTV50 -aircavity), Dmax (PTV50 -aircavity), and V110% (PTV50 -aircavity) were extracted for each patient. Annulus rings of 1-mm-diameter step size were generated for one of the air cavity plans (3.0 cm) for all four patients to determine Dmax (%) and V110% (cc) within each annulus. RESULTS: In the presence of B-field, hot spots at the cavity interface start to become visible at ~1 cm air cavity in both supine and prone positioning due to electron return effect (ERE). In the presence of B-field Dmax and V110% varied from 5523 ± 49 cGy and 0.09 ± 0.16 cc for 0 cm air cavity size to 6050 ± 109 cGy and 11.6 ± 6.7 cc for 5 cm air cavity size. The hot spots were located within 3 mm inside the rectal-air interface, where Dmax increased from 110.4 ± 0.5% without B-field to 119.2 ± 0.8 % with B-field. CONCLUSIONS: Air cavities inside rectum affects rectum plan dosimetry due ERE. Location and magnitude of hot spots are dependent on the size of the air cavity.

Prostate SBRT With Intrafraction Motion Management Using a Novel Linear Accelerator-Based MV-kV Imaging Method.

PURPOSE: This study reports clinical experience using a linear accelerator-based MV-kV imaging system for intrafraction motion management during prostate stereotactic body radiation therapy (SBRT). METHODS AND MATERIALS: From June 2016 to August 2018, 193 prostate SBRT patients were treated using MV-kV motion management (median dose 40 Gy in 5 fractions). Patients had 3 fiducials implanted then simulated and treated with a full bladder and empty rectum. Pretreatment orthogonal kVs and cone beam computed tomography were used to position patients and evaluate internal anatomy. Motion was tracked during volumetric modulated arc therapy delivery using simultaneously acquired kV and MV images from standard on-board systems. Treatment was interrupted to reposition patients when motion >1.5-2 mm was detected. Motion traces were analyzed and compared with Calypso traces from a previously treated similar patient cohort. To evaluate "natural motion" (ie, if we had not interrupted treatment and repositioned), intrafraction couch corrections were removed from all traces. Clinical effectiveness of the MV-kV system was explored by evaluating toxicity (Common Terminology Criteria for Adverse Events v3.0) and biochemical recurrence rates (nadir + 2 ng/mL). RESULTS: Median number of interruptions for patient repositioning was 1 per fraction (range, 0-9). Median overall treatment time was 8.2 minutes (range, 4.2-44.8 minutes). Predominant motion was inferior and posterior, and probability of motion increased with time. Natural motion >3 mm and >5 mm in any direction was observed in 32.3% and 10.2% of fractions, respectively. Calypso monitoring (n = 50) demonstrated similar motion results. In the 151 MV-kV patients with ≥3-month follow-up (median, 9.5 months; range, 3-26.5 months), grade ≥2 acute genitourinary/gastrointestinal and late genitourinary/gastrointestinal toxicity was observed in 9.9%/2.0% and 11.9%/2.7%, respectively. Biochemical control was 99.3% with a single failure in a high-risk patient. CONCLUSIONS: The MV-kV system is an effective method to manage intrafraction prostate motion during SBRT, offering the opportunity to correct for prostate clinical target volume displacements that would have otherwise extended beyond typical planning target volume margins.

Technical Note: A custom-designed flexible MR coil array for spine radiotherapy treatment planning.

PURPOSE: To assess the performance and optimize the MR image quality when using a custom-built flexible radiofrequency (RF) spine coil array fitted between the immobilization device and the patient for spine radiotherapy treatment planning. METHODS: A 32 channel flexible custom-designed receive-only coil array has been developed for spine radiotherapy simulation for a 3 T Philips MR scanner. Coil signal-to-noise performance and interactions with standard vendor hardware were assessed. In four volunteers, immobilization molds were created with a dummy version of the array within the mold, and subjects were scanned using the custom array in the mold. Phantoms and normal volunteers were scanned with both the custom spine coil array and the vendor's FDA-approved in-table posterior coil array to compare performance. RESULTS: The superior-inferior field of view for the custom spine array was ~30 cm encompassing at least 10 vertebrae. A noise correlation matrix showed at least 25 dB isolation between all coil elements. Signal-to-noise ratio (SNR) calculated on a phantom scan at the depth of the spinal cord was a factor of 3 higher with the form-fit spine array as compared to the vendor's posterior coil array. The body coil B1 transmit map was equivalent with and without the spine array in place demonstrating that the elements are decoupled from the body coil. Volunteer imaging showed improved SNR as compared to the vendor's posterior coil array. The custom array permitted a high degree of acceleration making possible the acquisition of isotropic high-resolution 1.1 × 1.1 × 1.1 mm3 three-dimensional data set over a 30-cm section of the spine in less than 5 min. CONCLUSION: The custom-designed form-fitting flexible spine coil array provided enhanced SNR and increased acceleration compared to the vendor's posterior array. Future studies will assess MR-based spinal cord imaging with the custom coil in comparison to CT myelogram.

Efficiency and safety increases after the implementation of a multi-institutional automated plan check tool at our institution.

PURPOSE: The plan check tool (PCT) is the result of a multi-institutional collaboration to jointly develop a flexible automated plan checking framework designed with the versatility to be shared across collaborating facilities while supporting the individual differences between practices. We analyze the effect that PCT has had on the efficiency and effectiveness of initial chart checks at our institution. METHODS AND MATERIALS: Data on errors identified during initial chart checks were acquired during two time periods: before the introduction of PCT in the clinic (6/24/2015 to 7/31/2015, 187 checks) and post-clinical release (4/14/2016 to 5/2/2016, 186 checks). During each time period, human plan checkers were asked to record all issues that they either manually detected or that were detected by PCT as well as the amount of time, less breaks, or interruptions, it took to check each plan. RESULTS: After the clinical release of PCT, there was a statistically significant decrease in the number of issues recorded by the human plan checkers both related to checks explicitly performed by PCT (13 vs 50, P < 0.001) and in issues identified overall (127 vs 200, P < 0.001). The mean and medium time for a plan check decreased by 20%. CONCLUSIONS: The use of a multi-institutional, configurable, automated plan checking tool has resulted in both substantial gains in efficiency and moving error detection to earlier points in the planning process, decreasing their likelihood that they reach the patient. The sizeable startup effort needed to create this tool from scratch was mitigated by the sharing, and subsequent co-development, of software code from a peer institution.

Technical Note: 3D localization of lung tumors on cone beam CT projections via a convolutional recurrent neural network.

PURPOSE: To design a convolutional recurrent neural network (CRNN) that calculates three-dimensional (3D) positions of lung tumors from continuously acquired cone beam computed tomography (CBCT) projections, and facilitates the sorting and reconstruction of 4D-CBCT images. METHOD: Under an IRB-approved clinical lung protocol, kilovoltage (kV) projections of the setup CBCT were collected in free-breathing. Concurrently, an electromagnetic signal-guided system recorded motion traces of three transponders implanted in or near the tumor. Convolutional recurrent neural network was designed to utilize a convolutional neural network (CNN) for extracting relevant features of the kV projections around the tumor, followed by a recurrent neural network for analyzing the temporal patterns of the moving features. Convolutional recurrent neural network was trained on the simultaneously collected kV projections and motion traces, subsequently utilized to calculate motion traces solely based on the continuous feed of kV projections. To enhance performance, CRNN was also facilitated by frequent calibrations (e.g., at 10° gantry rotation intervals) derived from cross-correlation-based registrations between kV projections and templates created from the planning 4DCT. Convolutional recurrent neural network was validated on a leave-one-out strategy using data from 11 lung patients, including 5500 kV images. The root-mean-square error between the CRNN and motion traces was calculated to evaluate the localization accuracy. RESULT: Three-dimensional displacement around the simulation position shown in the Calypso traces was 3.4 ± 1.7 mm. Using motion traces as ground truth, the 3D localization error of CRNN with calibrations was 1.3 ± 1.4 mm. CRNN had a success rate of 86 ± 8% in determining whether the motion was within a 3D displacement window of 2 mm. The latency was 20 ms when CRNN ran on a high-performance computer cluster. CONCLUSIONS: CRNN is able to provide accurate localization of lung tumors with aid from frequent recalibrations using the conventional cross-correlation-based registration approach, and has the potential to remove reliance on the implanted fiducials.

Early Tolerance and Tumor Control Outcomes with High-dose Ultrahypofractionated Radiation Therapy for Prostate Cancer.

BACKGROUND: Studies using stereotactic body radiotherapy (SBRT) dose escalation in in low- and intermediate-risk prostate cancer patients have indicated favorable outcomes. OBJECTIVE: To evaluate tolerance and tumor control outcomes in low- and intermediate-risk prostate cancer patients treated with high-dose SBRT following our phase 1 trial. DESIGN, SETTING, AND PARTICIPANTS: A total of 551 patients with low- or intermediate-risk prostate cancer were treated with SBRT. INTERVENTION: Treatment with 37.5-40Gy SBRT in five fractions directed to the prostate and seminal vesicles. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Outcome measurements included acute toxicities (<3 mo after radiotherapy [RT]) and late toxicities (>3 mo after RT) and tumor control evaluation (prostate-specific antigen [PSA] levels at 3-6-mo intervals and post-treatment prostate biopsy at 2yr). RESULTS AND LIMITATIONS: Acute grade 2 gastrointestinal (GI) toxicities occurred in 1.8% of patients, and late grade 2 and 3 GI toxicities were observed in 3.4% and 0.4% of patients, respectively. Acute grade 2 genitourinary (GU) toxicities occurred in 10% of patients, and grade 3 acute GU toxicities were observed in 0.7% of patients. Late grade 2 and 3 GU toxicities were observed in 21.1% and 2.5% of patients, respectively. The use of a hydrogel rectal spacer was significantly associated with reduced late GI toxicity and lower odds of developing late GU toxicity. The median follow-up was 17 mo, and 53% of those with at least 2yr of follow-up (103/193) had a biopsy performed. The 5-yr cumulative incidence of PSA failure was 2.1%, and the incidence of a positive 2-yr treatment biopsy was 12%. Limitations to this report include its retrospective nature and short follow-up time. CONCLUSIONS: Favorable short-term outcomes were achieved with high-dose SBRT for low- and intermediate-risk disease. Severe late toxicities were observed and favorable tumor control was found. PATIENT SUMMARY: We utilized stereotactic body radiotherapy, a form of external beam radiotherapy that delivers highly targeted high-dose treatment to the prostate, to treat over 500 localized prostate cancer patients in five sessions over 1.5 wk. Treatments were well tolerated without significant urinary or rectal side effects. Nearly 90% of those who underwent biopsies after treatment did not demonstrate residual active disease.

Patch-based generative adversarial neural network models for head and neck MR-only planning.

PURPOSE: To evaluate pix2pix and CycleGAN and to assess the effects of multiple combination strategies on accuracy for patch-based synthetic computed tomography (sCT) generation for magnetic resonance (MR)-only treatment planning in head and neck (HN) cancer patients. MATERIALS AND METHODS: Twenty-three deformably registered pairs of CT and mDixon FFE MR datasets from HN cancer patients treated at our institution were retrospectively analyzed to evaluate patch-based sCT accuracy via the pix2pix and CycleGAN models. To test effects of overlapping sCT patches on estimations, we (a) trained the models for three orthogonal views to observe the effects of spatial context, (b) we increased effective set size by using per-epoch data augmentation, and (c) we evaluated the performance of three different approaches for combining overlapping Hounsfield unit (HU) estimations for varied patch overlap parameters. Twelve of twenty-three cases corresponded to a curated dataset previously used for atlas-based sCT generation and were used for training with leave-two-out cross-validation. Eight cases were used for independent testing and included previously unseen image features such as fused vertebrae, a small protruding bone, and tumors large enough to deform normal body contours. We analyzed the impact of MR image preprocessing including histogram standardization and intensity clipping on sCT generation accuracy. Effects of mDixon contrast (in-phase vs water) differences were tested with three additional cases. The sCT generation accuracy was evaluated using mean absolute error (MAE) and mean error (ME) in HU between the plan CT and sCT images. Dosimetric accuracy was evaluated for all clinically relevant structures in the independent testing set and digitally reconstructed radiographs (DRRs) were evaluated with respect to the plan CT images. RESULTS: The cross-validated MAEs for the whole-HN region using pix2pix and CycleGAN were 66.9 ± 7.3 vs 82.3 ± 6.4 HU, respectively. On the independent testing set with additional artifacts and previously unseen image features, whole-HN region MAEs were 94.0 ± 10.6 and 102.9 ± 14.7 HU for pix2pix and CycleGAN, respectively. For patients with different tissue contrast (water mDixon MR images), the MAEs increased to 122.1 ± 6.3 and 132.8 ± 5.5 HU for pix2pix and CycleGAN, respectively. Our results suggest that combining overlapping sCT estimations at each voxel reduced both MAE and ME compared to single-view non-overlapping patch results. Absolute percent mean/max dose errors were 2% or less for the PTV and all clinically relevant structures in our independent testing set, including structures with image artifacts. Quantitative DRR comparison between planning CTs and sCTs showed agreement of bony region positions to <1 mm. CONCLUSIONS: The dosimetric and MAE based accuracy, along with the similarity between DRRs from sCTs, indicate that pix2pix and CycleGAN are promising methods for MR-only treatment planning for HN cancer. Our methods investigated for overlapping patch-based HU estimations also indicate that combining transformation estimations of overlapping patches is a potential method to reduce generation errors while also providing a tool to potentially estimate the MR to CT aleatoric model transformation uncertainty. However, because of small patient sample sizes, further studies are required.

Dual-input tracer kinetic modeling of dynamic contrast-enhanced MRI in thoracic malignancies.

Pulmonary perfusion with dynamic contrast-enhanced (DCE-) MRI is typically assessed using a single-input tracer kinetic model. Preliminary studies based on perfusion CT are indicating that dual-input perfusion modeling of lung tumors may be clinically valuable as lung tumors have a dual blood supply from the pulmonary and aortic system. This study aimed to investigate the feasibility of fitting dual-input tracer kinetic models to DCE-MRI datasets of thoracic malignancies, including malignant pleural mesothelioma (MPM) and nonsmall cell lung cancer (NSCLC), by comparing them to single-input (pulmonary or systemic arterial input) tracer kinetic models for the voxel-level analysis within the tumor with respect to goodness-of-fit statistics. Fifteen patients (five MPM, ten NSCLC) underwent DCE-MRI prior to radiotherapy. DCE-MRI data were analyzed using five different single- or dual-input tracer kinetic models: Tofts-Kety (TK), extended TK (ETK), two compartment exchange (2CX), adiabatic approximation to the tissue homogeneity (AATH) and distributed parameter (DP) models. The pulmonary blood flow (BF), blood volume (BV), mean transit time (MTT), permeability-surface area product (PS), fractional interstitial volume (vI ), and volume transfer constant (KTrans ) were calculated for both single- and dual-input models. The pulmonary arterial flow fraction (γ), pulmonary arterial blood flow (BFPA ) and systemic arterial blood flow (BFA ) were additionally calculated for only dual-input models. The competing models were ranked and their Akaike weights were calculated for each voxel according to corrected Akaike information criterion (cAIC). The optimal model was chosen based on the lowest cAIC value. In both types of tumors, all five dual-input models yielded lower cAIC values than their corresponding single-input models. The 2CX model was the best-fitted model and most optimal in describing tracer kinetic behavior to assess microvascular properties in both MPM and NSCLC. The dual-input 2CX-model-derived BFA was the most significant parameter in differentiating adenocarcinoma from squamous cell carcinoma histology for NSCLC patients.

Dynamic multiatlas selection-based consensus segmentation of head and neck structures from CT images.

PURPOSE: Manual delineation of head and neck (H&N) organ-at-risk (OAR) structures for radiation therapy planning is time consuming and highly variable. Therefore, we developed a dynamic multiatlas selection-based approach for fast and reproducible segmentation. METHODS: Our approach dynamically selects and weights the appropriate number of atlases for weighted label fusion and generates segmentations and consensus maps indicating voxel-wise agreement between different atlases. Atlases were selected for a target as those exceeding an alignment weight called dynamic atlas attention index. Alignment weights were computed at the image level and called global weighted voting (GWV) or at the structure level and called structure weighted voting (SWV) by using a normalized metric computed as the sum of squared distances of computed tomography (CT)-radiodensity and modality-independent neighborhood descriptors (extracting edge information). Performance comparisons were performed using 77 H&N CT images from an internal Memorial Sloan-Kettering Cancer Center dataset (N = 45) and an external dataset (N = 32) using Dice similarity coefficient (DSC), Hausdorff distance (HD), 95th percentile of HD, median of maximum surface distance, and volume ratio error against expert delineation. Pairwise DSC accuracy comparisons of proposed (GWV, SWV) vs single best atlas (BA) or majority voting (MV) methods were performed using Wilcoxon rank-sum tests. RESULTS: Both SWV and GWV methods produced significantly better segmentation accuracy than BA (P < 0.001) and MV (P < 0.001) for all OARs within both datasets. SWV generated the most accurate segmentations with DSC of: 0.88 for oral cavity, 0.85 for mandible, 0.84 for cord, 0.76 for brainstem and parotids, 0.71 for larynx, and 0.60 for submandibular glands. SWV's accuracy exceeded GWV's for submandibular glands (DSC = 0.60 vs 0.52, P = 0.019). CONCLUSIONS: The contributed SWV and GWV methods generated more accurate automated segmentations than the other two multiatlas-based segmentation techniques. The consensus maps could be combined with segmentations to visualize voxel-wise consensus between atlases within OARs during manual review.

Clinical evaluation of 4D MRI in the delineation of gross and internal tumor volumes in comparison with 4DCT.

PURPOSE: To evaluate clinical utility of respiratory-correlated (RC) four-dimensional magnetic resonance imaging (4DMRI) for lung tumor delineation and motion assessment, in comparison with the current clinical standard of 4D computed tomography (4DCT). METHODS AND MATERIALS: A prospective T2-weighted (T2w) RC-4DMRI technique was applied to acquire coronal 4DMRI images for 14 lung cancer patients (16 lesions) during free breathing (FB) under an IRB-approved protocol, together with a breath-hold (BH) T1w 3DMRI and axial 4DMRI. Clinical simulation CT and 4DCT were acquired within 2 h. An internal navigator was applied to trigger amplitude-binned 4DMRI acquisition whereas a bellows or real-time position management (RPM) was used in the 4DCT reconstruction. Six radiation oncologists manually delineated the gross and internal tumor volumes (GTV and ITV) in 399 3D images using programmed clinical workflows under a tumor delineation guideline. The ITV was the union of GTVs within the breathing cycle without margin. Average GTV and motion range were assessed and ITV variation between 4DMRI and 4DCT was evaluated using the Dice similarity index, mean distance agreement (MDA), and volume difference. RESULTS: The mean tumor volume is similar between 4DCT (GTV4DCT  = 1.0, as the reference) and T2w-4DMRI (GTVT2wMR  = 0.97), but smaller in T1w MRI (GTVT1wMR  = 0.76), suggesting possible peripheral edema around the tumor. Average GTV variation within the breathing cycle (22%) in 4DMRI is slightly greater than 4DCT (17%). GTV motion variation (-4 to 12 mm) and ITV variation (∆VITV =-25 to 95%) between 4DCT and 4DMRI are large, confirmed by relatively low ITV similarity (Dice = 0.72 ± 0.11) and large MDA = 2.9 ± 1.5 mm. CONCLUSION: Average GTVs are similar between T2w-4DMRI and 4DCT, but smaller by 25% in T1w BH MRI. Physician training and breathing coaching may be necessary to reduce ITV variability between 4DMRI and 4DCT. Four-dimensional magnetic resonance imaging is a promising and viable technique for clinical lung tumor delineation and motion assessment.

Impact of daily soft-tissue image guidance to prostate on pelvic lymph node (PLN) irradiation for prostate patients receiving SBRT.

PURPOSE: To determine the impact of using fiducial match for daily image-guidance on pelvic lymph node (PLN) coverage for prostate cancer patients receiving stereotactic body radiation therapy (SBRT). METHODS: Thirty patients underwent SBRT treatment to the prostate and PLN from 2014 to 2016. Each patient received either 800cGy × 5 or 500cGy × 5 to the prostate and 500cGy × 5 to the PLN. A 5 mm clinical target volume (CTV)-to-planning target volume (PTV) margin around the PLN was used for planning. Two registrations with planning computed tomography (PCT) for each of the daily cone beam CTs (CBCTs) were performed: a rigid registration to fiducials and to the bony anatomy. The average translational difference between fiducial and bony match as well as percentage of fractions with differences > 5mm were calculated. Changes in bladder and rectal volume as well as center-of-mass (COM) position from simulation parameters, and their correlation with translational difference were also evaluated. The dosimetric impact of the translational differences was calculated by shifting the plan isocenter. RESULTS: The average translational difference between fiducial and bony match was 0.06 ± 0.82, 2.1 ± 4.1, -2.8 ± 4.3, and 5.5 ± 4.2 mm for lateral, vertical, longitudinal, and vector directions. The average change in bladder and rectal volume from simulation was -67.2 ± 163.04 cc (-12 ± 52%) and -1.6 ± 18.75 (-2 ± 30%) cc. The average change in COM of bladder from the simulation position was 0.34 ± 2.49, 4.4 ± 8.1, and -3.9 ± 7.5 mm along the LR, AP, and SI directions. The corresponding COM change for the rectum was 0.17 ± 1.9, 1.34 ± 3.5, and -0.6 ± 5.2 mm. CONCLUSIONS: The 5 mm margin covered ~75% of fractions receiving PLN irradiation with SBRT, daily CBCT and fiducial-guided setup. The dosimetric impact on PLN coverage was significant in 19% of fractions or 25% of patients. A larger translational shift was due to variation in rectal volume and changes in COM position of the bladder and rectum. A consistent bladder positioning and/or rectum filling compared with presimulation volume were essential for adequate coverage of PLN in a hypofractionated treatment regime.