Search

Background and purpose: Magnetic resonance imaging (MRI)-to-computed tomography (CT) synthesis is essential in MRI-only radiotherapy workflows, particularly through deep learning techniques known for their accuracy. However, current supervised methods are limited to specific center’s learnings and depend on registration precision. The aim of this study was to evaluate the accuracy of unsupervised and supervised approaches in the context of prostate MRI-to-CT generation for radiotherapy dose calculation. Methods: CT/MRI image pairs from 99 prostate cancer patients across three different centers were used. A comparison between supervised and unsupervised conditional Generative Adversarial Networks (cGAN) was conducted. Unsupervised training incorporates a style transfer method with. Content and Style Representation for Enhanced Perceptual synthesis (CREPs) loss. For dose evaluation, the photon prescription dose was 60 Gy delivered in volumetric modulated arc therapy (VMAT). Imaging endpoint for sCT evaluation was Mean Absolute Error (MAE). Dosimetric endpoints included absolute dose differences and gamma analysis between CT and sCT dose calculations. Results: The unsupervised paired network exhibited the highest accuracy for the body with a MAE at 33.6 HU, the highest MAE was 45.5 HU obtained with unsupervised unpaired learning. All architectures provided clinically acceptable results for dose calculation with gamma pass rates above 94 % (1 % 1 mm 10 %). Conclusions: This study shows that multicenter data can produce accurate sCTs via unsupervised learning, eliminating CT-MRI registration. The sCTs not only matched HU values but also enabled precise dose calculations, suggesting their potential for wider use in MRI-only radiotherapy workflows.

Abstract Background Translating research, achieving impact, and assessing impact are important aspirations for all research collaboratives but can prove challenging. The Hunter Cancer Research Alliance (HCRA) was funded from 2014 to 2021 to enhance capacity and productivity in cancer research in a regional centre in Australia. This study aimed to assess the impact and benefit of the HCRA to help inform future research investments of this type. Method The Framework to Assess the Impact from Translational health research (FAIT) was selected as the preferred methodology. FAIT incorporates three validated methodologies for assessing impact: 1) Modified Payback; 2) Economic Analysis; and 3) Narrative overview and case studies. All three FAIT methods are underpinned by a Program Logic Model. Data were collected from HCRA and the University of Newcastle administrative records, directly from HCRA members, and website searches. Results In addition to advancing knowledge and providing capacity building support to members via grants, fellowships, scholarships, training, events and targeted translation support, key impacts of HCRA-member research teams included: (i) the establishment of a regional biobank that has distributed over 13,600 samples and became largely self-sustaining; (ii) conservatively leveraging $43.8 M (s.a.$20.5 M - $160.5 M) in funding and support from the initial $9.7 M investment; (iii) contributing to clinical practice guidelines and securing a patent for identification of stem cells for endometrial cell regeneration; (iv) shifting the treatment paradigm for all tumour types that rely on nerve cell innervation, (v) development and implementation of the world’s first real-time patient treatment verification system (Watchdog); (vi) inventing the effective ‘EAT’ psychological intervention to improve nutrition and outcomes in people experiencing radiotherapy for head and neck cancer; (vi) developing effective interventions to reduce smoking rates among priority groups, currently being rolled out to disadvantaged populations in NSW; and (vii) establishing a Consumer Advisory Panel and Consumer Engagement Committee to increase consumer involvement in research. Conclusion Using FAIT methodology, we have demonstrated the significant impact and downstream benefits that can be achieved by the provision of infrastructure-type funding to regional and rural research collaboratives to help address inequities in research activity and health outcomes and demonstrates a positive return on investment.

Purpose: Online adaptive radiation therapy (OART) uses daily imaging to identify changes in the patient's anatomy and generate a new treatment plan adapted to these changes for each fraction. The aim of this study was to determine the intrafraction motion and planning target volume (PTV) margins required for an OART workflow on the Varian Ethos system. Methods and Materials: Sixty-five fractions from 13 previously treated OART patients were analyzed for this retrospective study. The prostate and seminal vesicles were contoured by a radiation oncologist on 2 cone beam computed tomography scans (CBCT) for each fraction, the initial CBCT at the start of the treatment session, and the verification CBCT immediately before beam-on. In part 1 of the study, PTVs of different sizes were defined on the initial CBCT, and the geometric overlap with the clinical target volume (CTV) on the verification CBCT was used to determine the optimal OART margin. This was performed with and without a patient realignment shift by registering the verification CBCT to the initial CBCT. In part 2 of the study, the margins determined in part 1 were used for simulated Ethos OART treatments on all 65 fractions. The resultant coverage to the CTV on the verification CBCT, was compared with an image guided radiation therapy (IGRT) workflow with 7-mm margins. Results: Part 1 of the study found, if a verification CBCT and shift is performed, a 4-mm margin on the prostate and 5 mm on the seminal vesicles resulted in 95% of the CTV covered by the PTV in >90% of fractions, and 98% of the CTV covered by the PTV in >80% of fractions. Part 2 of the study found when these margins were used in an Ethos OART workflow, they resulted in CTV coverage that was superior to an IGRT workflow with 7-mm margins. Conclusions: A 4mm prostate margin and 5-mm seminal vesicles margin in an OART workflow with verification imaging are adequate to ensure coverage on the Varian Ethos system. Larger margins may be required if using an OART workflow without verification imaging.

Crowd anomaly detection is crucial in enhancing surveillance and crowd management. This paper proposes an efficient approach that combines spatial and temporal visual descriptors, sparse feature tracking, and neural networks for efficient crowd anomaly detection. The proposed approach utilises diverse local feature extraction methods, including SIFT, FAST, and AKAZE, with a sparse feature tracking technique to ensure accurate and consistent tracking. Delaunay triangulation is employed to represent the spatial distribution of features in an efficient way. Visual descriptors are categorised into individual behaviour descriptors and interactive descriptors to capture the temporal and spatial characteristics of crowd dynamics and behaviour, respectively. Neural networks are then utilised to classify these descriptors and pinpoint anomalies, making use of their strong learning capabilities. A significant component of our study is the assessment of how dimensionality reduction methods, particularly autoencoders and PCA, affect the feature set’s performance. This assessment aims to balance computational efficiency and detection accuracy. Tests conducted on benchmark crowd datasets highlight the effectiveness of our method in identifying anomalies. Our approach offers a nuanced understanding of crowd movement and patterns by emphasising both individual and collective characteristics. The visual and local descriptors facilitate high-level analysis by closely relating to semantic information and crowd behaviour. The analysis observed shows that this approach offers an efficient framework for crowd anomaly detection, contributing to improved crowd management and public safety. The proposed model achieves accuracy of 99.5 %, 96.1%, 99.0% and 88.5% in the UMN scenes 1, 2, and 3 and violence in crowds datasets, respectively.

Abstract The large coding potential of vaccinia virus (VV) vectors is a defining feature. However, limited regulatory switches are available to control viral replication as well as timing and dosing of transgene expression in order to facilitate safe and efficacious payload delivery. Herein, we adapt drug-controlled gene switches to enable control of virally encoded transgene expression, including systems controlled by the FDA-approved rapamycin and doxycycline. Using ribosome profiling to characterize viral promoter strength, we rationally design fusions of the operator element of different drug-inducible systems with VV promoters to produce synthetic promoters yielding robust inducible expression with undetectable baseline levels. We also generate chimeric synthetic promoters facilitating additional regulatory layers for VV-encoded synthetic transgene networks. The switches are applied to enable inducible expression of fusogenic proteins, dose-controlled delivery of toxic cytokines, and chemical regulation of VV replication. This toolbox enables the precise modulation of transgene circuitry in VV-vectored oncolytic virus design.

Abstract Introduction In this study, we aimed to investigate the feasibility of gadoxetate low‐temporal resolution (LTR) DCE‐MRI for voxel‐based hepatic extraction fraction (HEF) quantification for liver sparing radiotherapy using a deconvolution analysis (DA) method. Methods The accuracy and consistency of the deconvolution implementation in estimating liver function was first assessed using simulation data. Then, the method was applied to DCE‐MRI data collected retrospectively from 64 patients (25 normal liver function and 39 cirrhotic patients) to generate HEF maps. The normal liver function patient data were used to measure the variability of liver function quantification. Next, a correlation between HEF and ALBI score (a new model for assessing the severity of liver dysfunction) was assessed using Pearson's correlation. Differences in HEF between Child‐Pugh score classifications were assessed for significance using the Kruskal–Wallis test for all patient groups and Mann–Whitney U‐test for inter‐groups. A statistical significance was considered at a P‐value

IntroductionFor radiotherapy based solely on magnetic resonance imaging (MRI), generating synthetic computed tomography scans (sCT) from MRI is essential for dose calculation. The use of deep learning (DL) methods to generate sCT from MRI has shown encouraging results if the MRI images used for training the deep learning network and the MRI images for sCT generation come from the same MRI device. The objective of this study was to create and evaluate a generic DL model capable of generating sCTs from various MRI devices for prostate radiotherapyMaterials and methodsIn total, 90 patients from three centers (30 CT-MR prostate pairs/center) underwent treatment using volumetric modulated arc therapy for prostate cancer (PCa) (60 Gy in 20 fractions). T2 MRI images were acquired in addition to computed tomography (CT) images for treatment planning. The DL model was a 2D supervised conditional generative adversarial network (Pix2Pix). Patient images underwent preprocessing steps, including nonrigid registration. Seven different supervised models were trained, incorporating patients from one, two, or three centers. Each model was trained on 24 CT-MR prostate pairs. A generic model was trained using patients from all three centers. To compare sCT and CT, the mean absolute error in Hounsfield units was calculated for the entire pelvis, prostate, bladder, rectum, and bones. For dose analysis, mean dose differences of D99% for CTV, V95% for PTV, Dmax for rectum and bladder, and 3D gamma analysis (local, 1%/1 mm) were calculated from CT and sCT. Furthermore, Wilcoxon tests were performed to compare the image and dose results obtained with the generic model to those with the other trained models.ResultsConsidering the image results for the entire pelvis, when the data used for the test comes from the same center as the data used for training, the results were not significantly different from the generic model. Absolute dose differences were less than 1 Gy for the CTV D99% for every trained model and center. The gamma analysis results showed nonsignificant differences between the generic and monocentric models.ConclusionThe accuracy of sCT, in terms of image and dose, is equivalent to whether MRI images are generated using the generic model or the monocentric model. The generic model, using only eight MRI-CT pairs per center, offers robust sCT generation, facilitating PCa MRI-only radiotherapy for routine clinical use.

Background and Purpose: Addressing the need for accurate dose calculation in MRI-only radiotherapy, the generation of synthetic Computed Tomography (sCT) from MRI has emerged. Deep learning (DL) techniques, have shown promising results in achieving high sCT accuracies. However, existing sCT synthesis methods are often center-specific, posing a challenge to their generalizability. To overcome this limitation, recent studies have proposed approaches, such as multicenter training . Material and methods: The purpose of this work was to propose a multicenter sCT synthesis by DL, using a 2D cycle-GAN on 128 prostate cancer patients, from four different centers. Four cases were compared: monocenter cases, monocenter training and test on another center, multicenter trainings and a test on a center not included in the training and multicenter trainings with an included center in the test. Trainings were performed using 20 patients. sCT accuracy evaluation was performed using Mean Absolute Error, Mean Error and Peak-Signal-to-Noise-Ratio. Dose accuracy was assessed with gamma index and Dose Volume Histogram comparison. Results: Qualitative, quantitative and dose results show that the accuracy of sCTs for monocenter trainings and multicenter trainings using a seen center in the test did not differ significantly. However, when the test involved an unseen center, the sCT quality was inferior. Conclusions: The aim of this work was to propose generalizable multicenter training for MR-to-CT synthesis. It was shown that only a few data from one center included in the training cohort allows sCT accuracy equivalent to a monocenter study.

Background and purpose: Real-time treatment monitoring with the electronic portal imaging device (EPID) can conceptually provide a more accurate assessment of the quality of deep inspiration breath-hold (DIBH) and patient movement during tangential breast radiotherapy (RT). A system was developed to measure two geometrical parameters, the lung depth (LD) and the irradiated width (named here skin distance, SD), along three user-selected lines in MV EPID images of breast tangents. The purpose of this study was to test the system during tangential breast RT with DIBH. Materials and methods: Measurements of LDs and SDs were carried out in real time. DIBH was guided with a commercial system using a marker block. Results from 17 patients were assessed. Mean midline LDs, , per tangent were compared to the planned mLDs; differences between the largest and smallest observed () per tangent were calculated. Results: For 56% (162/288) of the tangents tested, were outside the tolerance window. All but one patient had at least one fraction showing this behaviour. The largest difference found between an and its planned mLD was −16.9 mm. The accuracy of patient positioning and the quality of marker-block-based DIBH guidance contributed to the differences. Fractions with patient position verification using a single EPID image taken before treatment showed a lower rate (34%), suggesting reassessment of setup procedures. Conclusions: Real-time treatment monitoring of the internal anatomy during DIBH delivery of tangential breast RT is feasible and useful. The new system requires no additional radiation for the patient.

Background and purpose: Magnetic resonance imaging (MRI)-only treatment planning is gaining in popularity in radiation oncology, with various methods available to generate a synthetic computed tomography (sCT) for this purpose. The aim of this study was to validate a sCT generation software for MRI-only radiotherapy planning of male and female pelvic cancers. The secondary aim of this study was to improve dose agreement by applying a derived relative electron and mass density (RED) curve to the sCT. Method and materials: Computed tomography (CT) and MRI scans of forty patients with pelvic neoplasms were used in the study. Treatment plans were copied from the CT scan to the sCT scan for dose comparison. Dose difference at reference point, 3D gamma comparison and dose volume histogram analysis was used to validate the dose impact of the sCT. The RED values were optimised to improve dose agreement by using a linear plot. Results: The average percentage dose difference at isocentre was 1.2% and the mean 3D gamma comparison with a criteria of 1%/1 mm was 84.0% ± 9.7%. The results indicate an inherent systematic difference in the dosimetry of the sCT plans, deriving from the tissue densities. With the adapted REDmod table, the average percentage dose difference was reduced to −0.1% and the mean 3D gamma analysis improved to 92.9% ± 5.7% at 1%/1 mm. Conclusions: CT generation software is a viable solution for MRI-only radiotherapy planning. The option makes it relatively easy for departments to implement a MRI-only planning workflow for cancers of male and female pelvic anatomy.

Abstract Introduction The prostatic urethra is an organ at risk for prostate radiotherapy with genitourinary toxicities a common side effect. Many external beam radiation therapy protocols call for urethral sparing, and with modulated radiotherapy techniques, the radiation dose distribution can be controlled so that maximum doses do not fall within the prostatic urethral volume. Whilst traditional diagnostic MRI sequences provide excellent delineation of the prostate, uncertainty often remains as to the true path of the urethra within the gland. This study aims to assess if a high‐resolution isotropic 3D T2 MRI series can reduce inter‐observer variability in urethral delineation for radiotherapy planning. Methods Five independent observers contoured the prostatic urethra for ten patients on three data sets; a 2 mm axial CT, a diagnostic 3 mm axial T2 TSE MRI and a 0.9 mm isotropic 3D T2 SPACE MRI. The observers were blinded from each other’s contours. A Dice Similarity Coefficient (DSC) score was calculated using the intersection and union of the five observer contours vs an expert reference contour for each data set. Results The mean DSC of the observer vs reference contours was 0.47 for CT, 0.62 for T2 TSE and 0.78 for T2 SPACE (P

Radiotherapy planning for lung cancer typically requires both 3D and 4D Computed Tomography (CT) to account for respiratory related movement. 4D Magnetic Resonance Imaging (MRI) with self-navigation offers a potential alternative with greater reliability in patients with irregular breathing patterns and improved soft tissue contrast. In this study 4D-CT and a 4D-MRI Radial Volumetric Interpolated Breath-hold Examination (VIBE) sequence was evaluated with a 4D phantom and 13 patient respiratory patterns, simulating tumour motion. Quantification of motion related tumour displacement in 4D-MRI and 4D-CT found no statistically significant difference in mean motion range. The results demonstrated the potential viability of 4D-MRI for lung cancer treatment planning.

External beam radiotherapy with photon beams is a highly accurate treatment modality, but requires extensive quality assurance programs to confirm that radiation therapy will be or was administered appropriately. In vivo dosimetry (IVD) is an essential element of modern radiation therapy because it provides the ability to catch treatment delivery errors, assist in treatment adaptation, and record the actual dose delivered to the patient. However, for various reasons, its clinical implementation has been slow and limited. The purpose of this report is to stimulate the wider use of IVD for external beam radiotherapy, and in particular of systems using electronic portal imaging devices (EPIDs). After documenting the current IVD methods, this report provides detailed software, hardware and system requirements for in vivo EPID dosimetry systems in order to help in bridging the current vendor-user gap. The report also outlines directions for further development and research. In vivo EPID dosimetry vendors, in collaboration with users across multiple institutions, are requested to improve the understanding and reduce the uncertainties of the system and to help in the determination of optimal action limits for error detection. Finally, the report recommends that automation of all aspects of IVD is needed to help facilitate clinical adoption, including automation of image acquisition, analysis, result interpretation, and reporting/documentation. With the guidance of this report, it is hoped that widespread clinical use of IVD will be significantly accelerated.

Purpose/Objective(s): The additional personnel and imaging procedures required for Adaptive Radiation Therapy (ART) pose a challenge for a broad implementation. We hypothesize that a change in transit fluence during the treatment course is correlated with the change of quality of life and thus can be used as a replanning trigger. Materials/Methods: Twenty-one head and neck cancer (HNC) patients filled out an MD Anderson Dysphagia Inventory (MDADI) questionnaire, before-and-after the radiotherapy treatment course. The transit fluence was measured by the Watchdog (WD) in-vivo portal dosimetry system. The patients were monitored with daily WD and weekly CBCTs. The region of interest (ROI) of each patient was defined as the outer contour of the patient between approximate spine levels C1 to C4, essentially the neck and mandible inside the beam’s eye view. The nth day integrated transit fluence change, Δϕ n , and the volume change, ΔV ROI , of the ROI of each patient was calculated from the corresponding WD and CBCT measurements. The correlation between MDADI scores and age, gender, planning mean dose to salivary glands , weight change ΔW, ΔV ROI , and Δϕ n , were analyzed using the ranked-Pearson correlation. Results: No statistically significant correlation was found for age, gender and ΔW. was found to have clinically important correlation with functional MDADI (ρ = −0.39, P = 0.081). ΔV ROI was found to have statistically significant correlation of 0.44, 0.47 and 0.44 with global, physical and functional MDADI ( P -value < 0.05). Δϕ n was found to have statistically significant ranked-correlation (−0.46, −0.46 and −0.45) with physical, functional and total MDADI ( P -value < 0.05). Conclusion: A transit fluence based decision support metric (DSM) is statistically correlated with the dysphagia risk. It can not only be used as an early signal in assisting clinicians in the ART patient selection for replanning, but also lowers the resource barrier of ART implementation.

Evidence for reduced expression of cyclin G associated kinase (GAK) in glomeruli of patients with chronic kidney disease was observed in the Nephroseq human database, and GAK was found to be associated with the decline in kidney function. To examine the role of GAK, a protein that functions to uncoat clathrin during endocytosis, we generated podocyte-specific Gak-knockout mice (Gak-KO), which developed progressive proteinuria and kidney failure with global glomerulosclerosis. We isolated glomeruli from the mice carrying the mutation to perform messenger RNA profiling and unearthed evidence for dysregulated podocyte calpain protease activity as an important contributor to progressive podocyte damage. Treatment with calpain inhibitor III specifically inhibited calpain-1/-2 activities, mitigated the degree of proteinuria and glomerulosclerosis, and led to a striking increase in survival in the Gak-KO mice. Podocyte-specific deletion of Capns1, essential for calpain-1 and calpain-2 activities, also improved proteinuria and glomerulosclerosis in Gak-KO mice. Increased podocyte calpain activity–mediated proteolysis of IκBα resulted in increased NF-κB p65–induced expression of growth arrest and DNA-damage-inducible 45 beta in the Gak-KO mice. Our results suggest that loss of podocyte-associated Gak induces glomerular injury secondary to calcium dysregulation and aberrant calpain activation, which when inhibited, can provide a protective role.

IntroductionStereotactic body radiotherapy (SBRT) is a non-invasive alternative to surgery for the treatment of non-metastatic prostate cancer (PC). The objectives of the Novel Integration of New prostate radiation schedules with adJuvant Androgen deprivation (NINJA) clinical trial are to compare two emerging SBRT regimens for efficacy with technical substudies focussing on MRI only planning and the use of knowledge-based planning (KBP) to assess radiotherapy plan quality.Methods and analysisEligible patients must have biopsy-proven unfavourable intermediate or favourable high-risk PC, have an Eastern Collaborative Oncology Group (ECOG) performance status 0-1 and provide written informed consent. All patients will receive 6 months in total of androgen deprivation therapy. Patients will be randomised to one of two SBRT regimens. The first will be 40 Gy in five fractions given on alternating days (SBRT monotherapy). The second will be 20 Gy in two fractions given 1 week apart followed 2 weeks later by 36 Gy in 12 fractions given five times per week (virtual high-dose rate boost (HDRB)). The primary efficacy outcome will be biochemical clinical control at 5 years. Secondary endpoints for the initial portion of NINJA look at the transition of centres towards MRI only planning and the impact of KBP on real-time (RT) plan assessment. The first 150 men will demonstrate accrual feasibility as well as addressing the KBP and MRI planning aims, prior to proceeding with total accrual to 472 patients as a phase III randomised controlled trial.Ethics and disseminationNINJA is a multicentre cooperative clinical trial comparing two SBRT regimens for men with PC. It builds on promising results from several single-armed studies, and explores radiation dose escalation in the Virtual HDRB arm. The initial component includes novel technical elements, and will form an important platform set for a definitive phase III study.Trial registration numberANZCTN 12615000223538.

Purpose: This project investigates the feasibility of implementation of MRI-only prostate planning in a prospective multi-center study.Method and Materials: A two-phase implementation model was utilized where centers performed retrospective analysis of MRI-only plans for five patients followed by prospective MRI-only planning for subsequent patients. Feasibility was assessed if at least 23/25 patients recruited to phase 2 received MRI-only treatment workflow. Whole-pelvic MRI scans (T2 weighted, isotropic 1.6 mm voxel 3D sequence) were converted to pseudo-CT using an established atlas-based method. Dose plans were generated using MRI contoured anatomy with pseudo-CT for dose calculation. A conventional CT scan was acquired subsequent to MRI-only plan approval for quality assurance purposes (QA-CT). 3D Gamma evaluation was performed between pseudo-CT calculated plan dose and recalculation on QA-CT. Criteria was 2%, 2 mm criteria with 20% low dose threshold. Gold fiducial marker positions for image guidance were compared between pseudo-CT and QA-CT scan prior to treatment.Results: All 25 patients recruited to phase 2 were treated using the MRI-only workflow. Isocenter dose differences between pseudo-CT and QA-CT were −0.04 ± 0.93% (mean ± SD). 3D Gamma dose comparison pass-rates were 99.7% ± 0.5% with mean gamma 0.22 ± 0.07. Results were similar for the two centers using two different scanners. All gamma comparisons exceeded the 90% pass-rate tolerance with a minimum gamma pass-rate of 98.0%. In all cases the gold fiducial markers were correctly identified on MRI and the distances of all seeds to centroid were within the tolerance of 1.0 mm of the distances on QA-CT (0.07 ± 0.41 mm), with a root-mean-square difference of 0.42 mm.Conclusion: The results support the hypothesis that an MRI-only prostate workflow can be implemented safely and accurately with appropriate quality assurance methods.

Objectives: To report feasibility, early toxicity, and PSA kinetics following gantry-based, stereotactic radiotherapy (SBRT) boost within a prospective, phase 2, multicenter study (PROMETHEUS: ACTRN12615000223538).Methods: Patients were treated with gantry-based SBRT, 19–20 Gy in two fractions delivered 1 week apart, followed by conventionally fractionated IMRT (46 Gy in 23 fractions). The study mandated MRI fusion for RT planning, rectal displacement, and intrafraction image guidance. Toxicity was prospectively graded using the Common Terminology Criteria for Adverse Events version 4.0 (CTCAE v4).Results: Between March 2014 and July 2018, 135 patients (76% intermediate, 24% high-risk) with a median age of 70 years (range 53–81) were treated across five centers. Short course (≤6 months) androgen deprivation therapy (ADT) was used in 36% and long course in 18%. Rectal displacement method was SpaceOAR in 59% and Rectafix in 41%. Forty-two and ninety-three patients were treated at the 19 Gy and 20 Gy dose levels, respectively. Median follow-up was 24 months. Acute grade 2 gastrointestinal (GI) and urinary toxicity occurred in 4.4 and 26.6% with no acute grade 3 toxicity. At 6, 12, 18, 24, and 36 months post-treatment the prevalence of late grade ≥2 gastrointestinal toxicity was 1.6, 3.7, 2.2, 0, and 0%, respectively, and the prevalence of late grade ≥2 urinary toxicity was 0.8, 11, 12, 7.1, and 6.3%, respectively. Three patients experienced grade 3 late toxicity at 12 to 18 months which subsequently resolved to grade 2 or less. For patients not receiving ADT the median PSA value pre-treatment was 7.6 ug/L (1.1–20) and at 12, 24, and 36 months post-treatment was 0.86, 0.36, and 0.20 ug/L.Conclusions: Delivery of a gantry-based SBRT boost is feasible in a multicenter setting, is well-tolerated with low rates of early toxicity and is associated with promising PSA responses. A second transient peak in urinary toxicity was observed at 18 months which subsequently resolved. Follow-up is ongoing to document late toxicity, long-term patient reported outcomes, and tumor control with this approach.

In this paper, we develop an optimised state-of-the-art 2D U-Net model by studying the effects of the individual deep learning model components in performing prostate segmentation. We found that for upsampling, the combination of interpolation and convolution is better than the use of transposed convolution. For combining feature maps in each convolution block, it is only beneficial if a skip connection with concatenation is used. With respect to pooling, average pooling is better than strided-convolution, max, RMS or L2 pooling. Introducing a batch normalisation layer before the activation layer gives further performance improvement. The optimisation is based on a private dataset as it has a fixed 2D resolution and voxel size for every image which mitigates the need of a resizing operation in the data preparation process. Non-enhancing data preprocessing was applied and five-fold cross-validation was used to evaluate the fully automatic segmentation approach. We show it outperforms the traditional methods that were previously applied on the private dataset, as well as outperforming other comparable state-of-the-art 2D models on the public dataset PROMISE12.