Contrast-agent-based perfusion MRI code repository and testing framework: ISMRM Open Science Initiative for Perfusion Imaging (OSIPI).

PURPOSE: Software has a substantial impact on quantitative perfusion MRI values. The lack of generally accepted implementations, code sharing and transparent testing reduces reproducibility, hindering the use of perfusion MRI in clinical trials. To address these issues, the ISMRM Open Science Initiative for Perfusion Imaging (OSIPI) aimed to establish a community-led, centralized repository for sharing open-source code for processing contrast-based perfusion imaging, incorporating an open-source testing framework. METHODS: A repository was established on the OSIPI GitHub website. Python was chosen as the target software language. Calls for code contributions were made to OSIPI members, the ISMRM Perfusion Study Group, and publicly via OSIPI websites. An automated unit-testing framework was implemented to evaluate the output of code contributions, including visual representation of the results. RESULTS: The repository hosts 86 implementations of perfusion processing steps contributed by 12 individuals or teams. These cover all core aspects of DCE- and DSC-MRI processing, including multiple implementations of the same functionality. Tests were developed for 52 implementations, covering five analysis steps. For T1 mapping, signal-to-concentration conversion and population AIF functions, different implementations resulted in near-identical output values. For the five pharmacokinetic models tested (Tofts, extended Tofts-Kety, Patlak, two-compartment exchange, and two-compartment uptake), differences in output parameters were observed between contributions. CONCLUSIONS: The OSIPI DCE-DSC code repository represents a novel community-led model for code sharing and testing. The repository facilitates the re-use of existing code and the benchmarking of new code, promoting enhanced reproducibility in quantitative perfusion imaging.

Geometric distortions in clinical MRI sequences for radiotherapy: insights gained from a multicenter investigation.

BACKGROUND: As magnetic resonance imaging (MRI) becomes increasingly integrated into radiotherapy (RT) for enhanced treatment planning and adaptation, the inherent geometric distortion in acquired MR images pose a potential challenge to treatment accuracy. This study aimed to evaluate the geometric distortion levels in the clinical MRI protocols used across Danish RT centers and discuss influence of specific sequence parameters. Based on the variety in geometric performance across centers, we assess if harmonization of MRI sequences is a relevant measure. MATERIALS AND METHODS: Nine centers participated with 12 MRI scanners and MRI-Linacs (MRL). Using a travelling phantom approach, a reference MRI sequence was used to assess variation in baseline distortion level between scanners. The phantom was also scanned with local clinical MRI sequences for brain, head/neck (H/N), abdomen, and pelvis. The influence of echo time, receiver bandwidth, image weighting, and 2D/3D acquisition was investigated. RESULTS: We found a large variation in geometric accuracy across 93 clinical sequences examined, exceeding the baseline variation found between MRI scanners (σ = 0.22 mm), except for abdominal sequences where the variation was lower. Brain and abdominal sequences showed lowest distortion levels ([0.22, 2.26] mm), and a large variation in performance was found for H/N and pelvic sequences ([0.19, 4.07] mm). Post hoc analyses revealed that distortion levels decreased with increasing bandwidth and a less clear increase in distortion levels with increasing echo time. 3D MRI sequences had lower distortion levels than 2D (median of 1.10 and 2.10 mm, respectively), and in DWI sequences, the echo-planar imaging read-out resulted in highest distortion levels. CONCLUSION: There is a large variation in the geometric distortion levels of clinical MRI sequences across Danish RT centers, and between anatomical sites. The large variation observed makes harmonization of MRI sequences across institutions and adoption of practices from well-performing anatomical sites, a relevant measure within RT.

Dose comparison of robustly optimized intensity modulated proton therapy (IMPT) vs IMRT and VMAT photon plans for testicular seminoma.

BACKGROUND: Patients with stage II seminoma have traditionally been treated with photons to the retroperitoneal and iliac space, which leads to a substantial dose bath to abdominal and pelvic organs at risk (OAR). As these patients are young and with excellent prognosis, reducing dose to OAR and thereby the risk of secondary cancer is of utmost importance. We compared IMPT to opposing IMRT fields and VMAT, assessing dose to OAR and both overall and organ-specific secondary cancer risk. MATERIAL AND METHODS: A comparative treatment planning study was conducted on planning CT-scans from ten patients with stage II seminoma, treated with photons to a 'dog-leg' field with doses ranging from 20 to 25 Gy and a 10 Gy sequential boost to the metastatic lymph node(s). Photon plans were either 3-4 field IMRT (Eclipse) or 1-2 arc VMAT (Pinnacle). Proton plans used robust (5 mm; 3.5%) IMPT (Eclipse), multi field optimization with 3 posterior fields supplemented by 2 anterior fields at the level of the iliac vessels. Thirty plans were generated. Mean doses to OARs were compared for IMRT vs IMPT and VMAT vs IMPT. The risk of secondary cancer was calculated according to the model described by Schneider, using excess absolute risk (EAR, per 10,000 persons per year) for body outline, stomach, duodenum, pancreas, bowel, bladder and spinal cord. RESULTS: Mean doses to all OARs were significantly lower with IMPT except similar kidney (IMRT) and spinal cord (VMAT) doses. The relative EAR for body outline was 0.59 for IMPT/IMRT (p < .05) and 0.33 for IMPT/VMAT (p < .05). Organ specific secondary cancer risk was also lower for IMPT except for pancreas and duodenum. CONCLUSION: Proton therapy reduced radiation dose to OAR compared to both IMRT and VMAT plans, and potentially reduce the risk of secondary cancer both overall and for most OAR.

Coverage probability planning for simultaneously integrated boosts of inguinal lymph nodes in vulvar cancer.

BACKGROUND AND PURPOSE: Strategies for minimizing irradiation of organs at risk (OARs) from pathological inguinal lymph node (PILN) boosting are needed to minimize the risk of morbidity. Coverage probability (CovP) is a conformal planning strategy for simultaneously integrated boost (SIB). Our aim was to investigate if SIB of PILN using CovP can be delivered safely in vulvar cancer. MATERIALS AND METHODS: Ten consecutive patients treated with definitive radiotherapy (RT) including SIB of PILN and with daily cone beam CT (CBCT) were included. Dose prescription was 51.2/32 fx to the elective target and 64 Gy/32 fx to the gross disease at the vulva and to positive lymph nodes (LN). PILN were contoured on both planning CT and MRI (GTV-N) and combined to form ITV-N. Each PILN GTV-N was contoured on every third CBCT, in total 11 CBCT for each patient. OARs were subcutaneous tissue (SC), inguinal vessels, skin rim, bowel, and body contour. Three plans were created for every patient: A) Standard CT-based planning; PTV-N based on GTV-NCT with a 10 mm isotropic margin. B) CT and MRI-based planning with smaller margins: PTV-N based on ITV-N with a 5 mm isotropic margin. C) CovP. The total delivered dose to GTV-Ns was estimated by accumulating dose across all fractions based on GTV-Ns contoured on CBCT. RESULTS: Thirty-five PILNs were boosted. There was no significant difference in accumulated GTV-N D98% between the three plans. CovP delivered a higher mean dose to the GTV-N D50% and D2% (p < 0.001). The planned mean doses to the OARs were reduced when applying CovP. CONCLUSIONS: SIB of PILN in vulvar cancer based on CovP and a 5 mm PTV margin does not compromise target coverage during RT and reduces the dose to normal tissues in the groin.

A systematic review on the use of quantitative imaging to detect cancer therapy adverse effects in normal-appearing brain tissue.

Cancer therapy for both central nervous system (CNS) and non-CNS tumors has been previously associated with transient and long-term cognitive deterioration, commonly referred to as 'chemo fog'. This therapy-related damage to otherwise normal-appearing brain tissue is reported using post-mortem neuropathological analysis. Although the literature on monitoring therapy effects on structural magnetic resonance imaging (MRI) is well established, such macroscopic structural changes appear relatively late and irreversible. Early quantitative MRI biomarkers of therapy-induced damage would potentially permit taking these treatment side effects into account, paving the way towards a more personalized treatment planning.This systematic review (PROSPERO number 224196) provides an overview of quantitative tomographic imaging methods, potentially identifying the adverse side effects of cancer therapy in normal-appearing brain tissue. Seventy studies were obtained from the MEDLINE and Web of Science databases. Studies reporting changes in normal-appearing brain tissue using MRI, PET, or SPECT quantitative biomarkers, related to radio-, chemo-, immuno-, or hormone therapy for any kind of solid, cystic, or liquid tumor were included. The main findings of the reviewed studies were summarized, providing also the risk of bias of each study assessed using a modified QUADAS-2 tool. For each imaging method, this review provides the methodological background, and the benefits and shortcomings of each method from the imaging perspective. Finally, a set of recommendations is proposed to support future research.

A national study on the inter-observer variability in the delineation of organs at risk in the brain.

BACKGROUND: The Danish Neuro Oncology Group (DNOG) has established national consensus guidelines for the delineation of organs at risk (OAR) structures based on published literature. This study was conducted to finalise these guidelines and evaluate the inter-observer variability of the delineated OAR structures by expert observers. MATERIAL AND METHODS: The DNOG delineation guidelines were formed by participants from all Danish centres that treat brain tumours with radiotherapy. In a two-day workshop, guidelines were discussed and finalised based on a pilot study. Following this, the ten participants contoured the following OARs on T1-weighted gadolinium enhanced MRI from 13 patients with brain tumours: optic tracts, optic nerves, chiasm, spinal cord, brainstem, pituitary gland and hippocampus. The metrics used for comparison were the Dice similarity coefficient (Dice), mean surface distance (MSD) and others. RESULTS: A total of 968 contours were delineated across the 13 patients. On average eight (range six to nine) individual contour sets were made per patient. Good agreement was found across all structures with a median MSD below 1 mm for most structures, with the chiasm performing the best with a median MSD of 0.45 mm. The Dice was as expected highly volume dependent, the brainstem (the largest structure) had the highest Dice value with a median of 0.89 whereas smaller volumes such as the chiasm had a Dice of 0.71. CONCLUSION: Except for the caudal definition of the spinal cord, the variances observed in the contours of OARs in the brain were generally low and consistent. Surface mapping revealed sub-regions of higher variance for some organs. The data set is being prepared as a validation data set for auto-segmentation algorithms for use within the Danish Comprehensive Cancer Centre - Radiotherapy and potential collaborators.

Dosimetric impact of edema on inguinal lymph node boost in locally advanced vulvar cancer.

We aimed to evaluate the extent of groin edema and its dosimetric effect in boosted inguinal lymph nodes (LN) for vulvar cancer patients. The level of edema was determined in 10 patients treated with radical radiotherapy. A dosimetric evaluation of six LNs in the patient with the maximum level of edema was performed. The accumulated dose across CBCT fractions was acceptable for all six LNs (>94% of prescribed dose) even with the development of up to 13 mm of edema. The major contributor to fractional dose degradation was geographical displacement of the nodes. We suggest evaluation of edema on daily CBCT.

Are complex DCE-MRI models supported by clinical data?

PURPOSE: To ascertain whether complex dynamic contrast enhanced (DCE) MRI tracer kinetic models are supported by data acquired in the clinic and to determine the consequences of limited contrast-to-noise. METHODS: Generically representative in silico and clinical (cervical cancer) DCE-MRI data were examined. Bayesian model selection evaluated support for four compartmental DCE-MRI models: the Tofts model (TM), Extended Tofts model, Compartmental Tissue Uptake model (CTUM), and Two-Compartment Exchange model. RESULTS: Complex DCE-MRI models were more sensitive to noise than simpler models with respect to both model selection and parameter estimation. Indeed, as contrast-to-noise decreased, complex DCE models became less probable and simpler models more probable. The less complex TM and CTUM were the optimal models for the DCE-MRI data acquired in the clinic. [In cervical tumors, Ktrans, Fp, and PS increased after radiotherapy (P = 0.004, 0.002, and 0.014, respectively)]. CONCLUSION: Caution is advised when considering application of complex DCE-MRI kinetic models to data acquired in the clinic. It follows that data-driven model selection is an important prerequisite to DCE-MRI analysis. Model selection is particularly important when high-order, multiparametric models are under consideration. (Parameters obtained from kinetic modeling of cervical cancer clinical DCE-MRI data showed significant changes at an early stage of radiotherapy.) Magn Reson Med 77:1329-1339, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

Comparison of linear and nonlinear implementation of the compartmental tissue uptake model for dynamic contrast-enhanced MRI.

PURPOSE: Fitting tracer kinetic models using linear methods is much faster than using their nonlinear counterparts, although this comes often at the expense of reduced accuracy and precision. The aim of this study was to derive and compare the performance of the linear compartmental tissue uptake (CTU) model with its nonlinear version with respect to their percentage error and precision. THEORY AND METHODS: The linear and nonlinear CTU models were initially compared using simulations with varying noise and temporal sampling. Subsequently, the clinical applicability of the linear model was demonstrated on 14 patients with locally advanced cervical cancer examined with dynamic contrast-enhanced magnetic resonance imaging. RESULTS: Simulations revealed equal percentage error and precision when noise was within clinical achievable ranges (contrast-to-noise ratio >10). The linear method was significantly faster than the nonlinear method, with a minimum speedup of around 230 across all tested sampling rates. Clinical analysis revealed that parameters estimated using the linear and nonlinear CTU model were highly correlated (ρ ≥ 0.95). CONCLUSION: The linear CTU model is computationally more efficient and more stable against temporal downsampling, whereas the nonlinear method is more robust to variations in noise. The two methods may be used interchangeably within clinical achievable ranges of temporal sampling and noise. Magn Reson Med 77:2414-2423, 2017. © 2016 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Diffusion tensor magnetic resonance imaging driven growth modeling for radiotherapy target definition in glioblastoma.

BACKGROUND: The clinical target volume (CTV) in radiotherapy is routinely based on gadolinium contrast enhanced T1 weighted (T1w + Gd) and T2 weighted fluid attenuated inversion recovery (T2w FLAIR) magnetic resonance imaging (MRI) sequences which have been shown to over- or underestimate the microscopic tumor cell spread. Gliomas favor spread along the white matter fiber tracts. Tumor growth models incorporating the MRI diffusion tensors (DTI) allow to account more consistently for the glioma growth. The aim of the study was to investigate the potential of a DTI driven growth model to improve target definition in glioblastoma (GBM). MATERIAL AND METHODS: Eleven GBM patients were scanned using T1w, T2w FLAIR, T1w + Gd and DTI. The brain was segmented into white matter, gray matter and cerebrospinal fluid. The Fisher-Kolmogorov growth model was used assuming uniform proliferation and a difference in white and gray matter diffusion of a ratio of 10. The tensor directionality was tested using an anisotropy weighting parameter set to zero (γ0) and twenty (γ20). The volumetric comparison was performed using Hausdorff distance, Dice similarity coefficient (DSC) and surface area. RESULTS: The median of the standard CTV (CTVstandard) was 180 cm3. The median surface area of CTVstandard was 211 cm2. The median surface area of respective CTVγ0 and CTVγ20 significantly increased to 338 and 376 cm2, respectively. The Hausdorff distance was greater than zero and significantly increased for both CTVγ0 and CTVγ20 with respective median of 18.7 and 25.2 mm. The DSC for both CTVγ0 and CTVγ20 were significantly below one with respective median of 0.74 and 0.72, which means that 74 and 72% of CTVstandard were included in CTVγ0 and CTVγ20, respectively. CONCLUSIONS: DTI driven growth models result in CTVs with a significantly increased surface area, a significantly increased Hausdorff distance and decreased overlap between the standard and model derived volume.

Diffusion-weighted magnetic resonance imaging during radiotherapy of locally advanced cervical cancer--treatment response assessment using different segmentation methods.

BACKGROUND: Diffusion-weighted magnetic resonance imaging (DW-MRI) and the derived apparent diffusion coefficient (ADC) value has potential for monitoring tumor response to radiotherapy (RT). Method used for segmentation of volumes with reduced diffusion will influence both volume size and observed distribution of ADC values. This study evaluates: 1) different segmentation methods; and 2) how they affect assessment of tumor ADC value during RT. MATERIAL AND METHODS: Eleven patients with locally advanced cervical cancer underwent MRI three times during their RT: prior to start of RT (PRERT), two weeks into external beam RT (WK2RT) and one week prior to brachytherapy (PREBT). Volumes on DW-MRI were segmented using three semi-automatic segmentation methods: "cluster analysis", "relative signal intensity (SD4)" and "region growing". Segmented volumes were compared to the gross tumor volume (GTV) identified on T2-weighted MR images using the Jaccard similarity index (JSI). ADC values from segmented volumes were compared and changes of ADC values during therapy were evaluated. RESULTS: Significant difference between the four volumes (GTV, DWIcluster, DWISD4 and DWIregion) was found (p < 0.01), and the volumes changed significantly during treatment (p < 0.01). There was a significant difference in JSI among segmentation methods at time of PRERT (p < 0.016) with region growing having the lowest JSIGTV (mean± sd: 0.35 ± 0.1), followed by the SD4 method (mean± sd: 0.50 ± 0.1) and clustering (mean± sd: 0.52 ± 0.3). There was no significant difference in mean ADC value compared at same treatment time. Mean tumor ADC value increased significantly (p < 0.01) for all methods across treatment time. CONCLUSION: Among the three semi-automatic segmentations of hyper-intense intensities on DW-MR images implemented, cluster analysis and relative signal thresholding had the greatest similarity to the clinical tumor volume. Evaluation of mean ADC value did not depend on segmentation method.

Correction of diffusion-weighted magnetic resonance imaging for brachytherapy of locally advanced cervical cancer.

BACKGROUND: Geometrical distortion is a major obstacle for the use of echo planar diffusion-weighted magnetic resonance imaging (DW-MRI) in planning of radiotherapy. This study compares geometrical distortion correction methods of DW-MRI at time of brachytherapy (BT) in locally advanced cervical cancer patients. MATERIAL AND METHODS: In total 21 examinations comprising DW-MRI, dual gradient echo (GRE) for B0 field map calculation and T2-weighted (T2W) fat-saturated MRI of eight patients with locally advanced cervical cancer were acquired during BT with a plastic tandem and ring applicator in situ. The ability of B0 field map correction (B0M) and deformable image registration (DIR) to correct DW-MRI geometric image distortion was compared to the non-corrected DW-MRI including evaluation of apparent diffusion coefficient (ADC) for the gross tumor volume (GTV). RESULTS: Geometrical distortion correction decreased tandem displacement from 3.3 ± 0.9 mm (non-corrected) to 2.9 ± 1.0 mm (B0M) and 1.9 ± 0.6 mm (DIR), increased mean normalized cross-correlation from 0.69 ± 0.1 (non- corrected) to 0.70 ± 0.10 (B0M) and 0.77 ± 0.1 (DIR), and increased the Jaccard similarity coefficient from 0.72 ± 0.1 (non-corrected) to 0.73 ± 0.06 (B0M) and 0.77 ± 0.1 (DIR). For all parameters only DIR corrections were significant (p < 0.05). ADC of the GTV did not change significantly with either correction method. CONCLUSION: DIR significantly improved geometrical accuracy of DW-MRI, with remaining residual uncertainties of less than 2 mm, while no significant improvement was seen using B0 field map correction.

Tracer kinetic model selection for dynamic contrast-enhanced magnetic resonance imaging of locally advanced cervical cancer.

BACKGROUND: Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) offers a unique capability to probe tumour microvasculature. Different analysis of the acquired data will possibly lead to different conclusions. Therefore, the objective of this study was to investigate under which conditions the Tofts (TM), extended Tofts (ETM), compartmental tissue uptake model (C-TU) and 2-compartment exchange model (2CXM) were the optimal tracer kinetic models (TKMs) for the analysis of DCE-MRI in patients with cervical cancer. MATERIAL AND METHODS: Ten patients with locally advanced cervical cancer (FIGO: IIA/IIB/IIIB/IVA - 1/5/3/1) underwent DCE-MRI prior to radiotherapy. From the two-parameter TM it was possible to extract the forward volume transfer constant (K(trans)) and the extracellular-extravascular volume fraction (ve). From the three-parameter ETM, additionally the plasma volume fraction (vp) could be extracted. From the three-parameter C-TU it was possible to extract information about the blood flow (Fp), permeability-surface area product (PS) and vp. Finally, the four-parameter 2CXM extended the C-TU to include ve. For each voxel, corrected Akaike information criterion (AICc) values were calculated, taking into account both the goodness-of-fit and the number of model parameters. The optimal model was defined as the model with the lowest AICc. RESULTS: All four TKMs were the optimal model in different contiguous regions of the cervical tumours. For the 24 999 analysed voxels, the TM was optimal in 17.0%, the ETM was optimal in 2.2%, the C-TU in 23.4% and the 2CXM was optimal in 57.3%. Throughout the tumour, a high correlation was found between K(trans)(TM) and Fp(2CXM), ρ = 0.91. CONCLUSION: The 2CXM was most often optimal in describing the contrast agent enhancement of pre-treatment cervical cancers, although this model broke down in a subset of the tumour voxels where overfitting resulted in non-physiological parameter estimates. Due to the possible overfitting of the 2CXM, the C-TU was found more robust and when 2CXM was excluded from comparison the C-TU was the preferred model.

A method for selection of beam angles robust to intra-fractional motion in proton therapy of lung cancer.

BACKGROUND: Proton therapy offers the potential for sparing the normal tissue surrounding the target. However, due to well-defined proton ranges around the Bragg peak, dose deposition is more sensitive to changes in the water equivalent path length (WEPL) than with photons. In this study, we assess WEPL variations caused by breathing-induced motion for all possible beam angles in a series of lung cancer patients. By studying the association between measures for WEPL variation and breathing-induced target dose degradation we aimed to develop and explore a tool to identify beam angles that are robust to patient-specific patterns of intra-fractional motion. MATERIAL AND METHODS: Using four-dimensional computed tomography (4DCT) images of three lung cancer patients we evaluated the impact of the WEPL changes on target dose coverage for a series of coplanar single-beam plans. The plans were optimised for the internal target volume (ITV) at the maximum intensity projection (MIP) CT for every 3° gantry interval. The plans were transferred to the ten 4DCT phases and the average reduction in ITV V95 over the ten phases, relative to the original MIP CT calculation, was quantified. The target dose reduction was associated with the mean difference between the WEPL and the phase-averaged WEPL computed for all beam rays across all possible gantry-couch angle combinations. RESULTS: The gantry-couch angle maps showed areas of both high and low WEPL variation, with overall quite similar patterns yet with individual differences reflecting differences in tumour position and breathing-induced motion. The coplanar plans showed a strong association between WEPL changes and ITV V95 reduction, with a correlation coefficient ranging between 0.92 and 0.98 for the three patients (p < 0.01). CONCLUSION: We have presented a 4DCT-based method to quantify WEPL changes during the breathing cycle. The method identified proton field gantry-couch angle combinations that were either sensitive or robust to WEPL changes. WEPL variations along the beam path were associated with target under-dosage.

The effect of editing clinical contours on deep-learning segmentation accuracy of the gross tumor volume in glioblastoma.

Background and purpose: Deep-learning (DL) models for segmentation of the gross tumor volume (GTV) in radiotherapy are generally based on clinical delineations which suffer from inter-observer variability. The aim of this study was to compare performance of a DL-model based on clinical glioblastoma GTVs to a model based on a single-observer edited version of the same GTVs. Materials and methods: The dataset included imaging data (Computed Tomography (CT), T1, contrast-T1 (T1C), and fluid-attenuated-inversion-recovery (FLAIR)) of 259 glioblastoma patients treated with post-operative radiotherapy between 2012 and 2019 at a single institute. The clinical GTVs were edited using all imaging data. The dataset was split into 207 cases for training/validation and 52 for testing.GTV segmentation models (nnUNet) were trained on clinical and edited GTVs separately and compared using Surface Dice with 1 mm tolerance (sDSC1mm). We also evaluated model performance with respect to extent of resection (EOR), and different imaging combinations (T1C/T1/FLAIR/CT, T1C/FLAIR/CT, T1C/FLAIR, T1C/CT, T1C/T1, T1C). A Wilcoxon test was used for significance testing. Results: The median (range) sDSC1mm of the clinical-GTV-model and edited-GTV-model both evaluated with the edited contours, was 0.76 (0.43-0.94) vs. 0.92 (0.60-0.98) respectively (p < 0.001). sDSC1mm was not significantly different between patients with a biopsy, partial, and complete resection. T1C as single input performed as good as use of imaging combinations. Conclusions: High segmentation accuracy was obtained by the DL-models. Editing of the clinical GTVs significantly increased DL performance with a relevant effect size. DL performance was robust for EOR and highly accurate using only T1C.

RadDeploy: A framework for integrating in-house developed software and artificial intelligence models seamlessly into radiotherapy workflows.

The use of and research in automation and artificial intelligence (AI) in radiotherapy is moving with incredible pace. Many innovations do, however, not make it into the clinic. One technical reason for this may be the lack of a platform to deploy such software into clinical practice. We suggest RadDeploy as a framework for integrating containerized software in clinical workflows outside of treatment planning systems. RadDeploy supports multiple DICOM as input for model containers and can run model containers asynchronously across GPUs and computers. This technical note summarizes the inner workings of RadDeploy and demonstrates three use-cases with varying complexity.