Evaluation of a deep learning magnetic resonance imaging reconstruction method for synthetic computed tomography generation in prostate radiotherapy.

Background and Purpose: In magnetic resonance imaging (MRI) only radiotherapy computed tomography (CT) is excluded. The method relies entirely on synthetic CT images generated from MRI. This study evaluates the compatibility of a commercial synthetic CT (sCT) with an accelerated commercial deep learning reconstruction (DLR) in MRI-only prostate radiotherapy. Materials and Methods: For a group of 24 patients (cohort 1) the effects of DLR were studied in isolation. MRI data were reconstructed conventionally and with DLR from identical k-space data, and sCTs were generated for both reconstructions. The sCT quality, Hounsfield Unit (HU) and dosimetric impact were investigated. In another group of 15 patients (cohort 2) effects on sCT generation using accelerated MRI acquisition (40 % time reduction) reconstructed with DLR were investigated. Results: sCT images from both cohorts, generated from DLR MRI data, were of clinically expected image quality. The mean dose differences for targets and organs at risks in cohort 1 were <0.06 Gy, corresponding to a 0.1 % prescribed dose difference. Similar dose differences were observed in cohort 2. Gamma pass rates for cohort 1 were 100 % for criteria 3 %/3mm, 2 %/2mm and 1 %/1mm for all dose levels. Mean error and mean absolute error inside the body, between sCTs, averaged over all cohort 1 subjects, were -1.1 ± 0.6 [-2.4 0.2] and 2.9 ± 0.4 [2.3 3.9] HU, respectively. Conclusions: DLR was suitable for sCT generation with clinically negligible differences in HU and calculated dose compared to the conventional MRI reconstruction method. For sCT generation DLR enables scan time reduction, without compromised sCT quality.

Deep learning-based classification of organs at risk and delineation guideline in pelvic cancer radiation therapy.

Deep learning (DL) models for radiation therapy (RT) image segmentation require accurately annotated training data. Multiple organ delineation guidelines exist; however, information on the used guideline is not provided with the delineation. Extraction of training data with coherent guidelines can therefore be challenging. We present a supervised classification method for pelvis structure delineations where bowel cavity, femoral heads, bladder, and rectum data, with two guidelines, were classified. The impact on DL-based segmentation quality using mixed guideline training data was also demonstrated. Bowel cavity was manually delineated on CT images for anal cancer patients (n = 170) according to guidelines Devisetty and RTOG. The DL segmentation quality from using training data with coherent or mixed guidelines was investigated. A supervised 3D squeeze-and-excite SENet-154 model was trained to classify two bowel cavity delineation guidelines. In addition, a pelvis CT dataset with manual delineations from prostate cancer patients (n = 1854) was used where data with an alternative guideline for femoral heads, rectum, and bladder were generated using commercial software. The model was evaluated on internal (n = 200) and external test data (n = 99). By using mixed, compared to coherent, delineation guideline training data mean DICE score decreased 3% units, mean Hausdorff distance (95%) increased 5 mm and mean surface distance (MSD) increased 1 mm. The classification of bowel cavity test data achieved 99.8% unweighted classification accuracy, 99.9% macro average precision, 97.2% macro average recall, and 98.5% macro average F1. Corresponding metrics for the pelvis internal test data were all 99% or above and for the external pelvis test data they were 96.3%, 96.6%, 93.3%, and 94.6%. Impaired segmentation performance was observed for training data with mixed guidelines. The DL delineation classification models achieved excellent results on internal and external test data. This can facilitate automated guideline-specific data extraction while avoiding the need for consistent and correct structure labels.

Geometric impact and dose estimation of on-patient placement of a lightweight receiver coil in a clinical magnetic resonance imaging-only radiotherapy workflow for prostate cancer.

Background and Purpose: For pelvic magnetic resonance imaging (MRI)-only radiotherapy the use of receiver coil bridges (CB) is recommended to avoid deformation of the patient. Development in coil technology has enabled lightweight, flexible coils. In this work we evaluate the effects of a lightweight coil in a pelvic MRI-only radiotherapy workflow. Materials and Methods: Twenty-one patients, referred to prostate MRI-only radiotherapy, were included. Images were acquired with and without CB. Anatomical deformation from the on-patient coil placement was measured in the anterior-posterior (AP) and left-right (LR) direction. The change in signal-to-noise ratio (SNR) was measured in phantom and in vivo.The clinical treatment plan, created on the image with CB, was transferred and recalculated on the image without the CB. Dose metrics for the targets (planning- and clinical target volume) and organs at risks (OAR) were analyzed. Results: There was a statistically significant increase in SNR in-vivo (median 21 %, p = 0.002) when removing the CB. Anatomical differences after removing the CB in patients were -1.5 mm in AP (median change) and + 2.5 mm in LR direction. Dosimetric differences for the target structures were clinically negligible, but statistically significant. The difference in target mean doses were 0.2 % (both p = 0.004) of the prescribed dose. No dosimetric differences were observed for the OAR, except for the penile bulb. Conclusions: We concluded that anatomical change and dosimetric differences, originating from scanning without CB were minor. The CB can thereby be removed from the workflow, enabling easier patient positioning and increased SNR when using lightweight coils.

MRI-only radiotherapy from an economic perspective: Can new techniques in prostate cancer treatment be cost saving?

Background and Purpose: The aim of this study was to analyze a magnetic resonance imaging (MRI)-only radiotherapy workflow from an economic perspective in terms of reduced time, costs and systematic uncertainties. Material/Methods: A documented Swedish clinical implementation of MRI-only radiotherapy was used as template for cost assessments compared to a combined computed tomography (CT)/MRI workflow. The costs were taken from official regional price lists from 2021. MRI-only specific quality assurance (QA) was assumed necessary in an initial phase. Treatment plans for target volumes with margins of 5-10 mm were created for ten prostate cancer patients prescribed 78 Gy in 39 fractions. The risk of Grade ≥ 2 rectal toxicity or rectal bleeding was calculated using the QUANTEC recommended NTCP model and costs estimated based on subsequent diagnostic examinations. Results: The exclusion of the CT-examination and faster target delineation were the main contributors to cost reductions. Additional QA procedures limited the initial cost reduction to 14 EUR/patient. Long-term MRI-only reduced the costs by 209 EUR/patient. Reducing margins resulted in Grade ≥ 2 rectal toxicity or rectal bleeding probability of 9.7 % for 7 mm margin and 6.0 % for 5 mm margin. This margin reduction resulted in an additional cost reduction of 46 EUR/patient. Conclusion: An MRI-only workflow implementation is associated with reduced costs when the workflow tasks are more time efficient and side effects are reduced as a result of margin reduction. The short-term economic benefits are limited due to extra costs of QA procedures. The economic benefits of MRI-only will make impact first when the workflow is well established, and margin reduction has been included.

Tensor-valued diffusion magnetic resonance imaging in a radiotherapy setting.

Background and purpose: Diagnostic information about cell density variations and microscopic tissue anisotropy can be gained from tensor-valued diffusion magnetic resonance imaging (MRI). These properties of tissue microstructure have the potential to become novel imaging biomarkers for radiotherapy response. However, tensor-valued diffusion encoding is more demanding than conventional encoding, and its compatibility with MR scanners that are dedicated to radiotherapy has not been established. Thus, our aim was to investigate the feasibility of tensor-valued diffusion MRI with radiotherapy dedicated MR equipment. Material and methods: A tensor-valued diffusion protocol was implemented, and five healthy volunteers were scanned with different resolutions using conventional head coil and radiotherapy coil setup with fixation masks. Signal-to-noise-ratio (SNR) was evaluated to assess the risk of signal bias due to rectified noise floor. We also evaluated the repeatability and reproducibility of the microstructure parameters. One patient with brain metastasis was scanned to investigate the image quality and the transferability of the setup to diseased tissue. Results: A resolution of 3 × 3 × 3 mm3 provided images with SNR > 3 for 93 % of the voxels using radiotherapy coil setup. The parameter maps and repeatability characteristics were comparable to those observed with a conventional head coil. The patient evaluation demonstrated successful parameter analysis also in tumor tissue, with SNR > 3 for 93 % of the voxels. Conclusion: We demonstrate that tensor-valued diffusion MRI is compatible with radiotherapy fixation masks and coil setup for investigations of microstructure parameters. The reported reproducibility may be used to plan future investigations of imaging biomarkers in brain cancer radiotherapy.

Deep learning-based classification and structure name standardization for organ at risk and target delineations in prostate cancer radiotherapy.

Radiotherapy (RT) datasets can suffer from variations in annotation of organ at risk (OAR) and target structures. Annotation standards exist, but their description for prostate targets is limited. This restricts the use of such data for supervised machine learning purposes as it requires properly annotated data. The aim of this work was to develop a modality independent deep learning (DL) model for automatic classification and annotation of prostate RT DICOM structures. Delineated prostate organs at risk (OAR), support- and target structures (gross tumor volume [GTV]/clinical target volume [CTV]/planning target volume [PTV]), along with or without separate vesicles and/or lymph nodes, were extracted as binary masks from 1854 patients. An image modality independent 2D InceptionResNetV2 classification network was trained with varying amounts of training data using four image input channels. Channel 1-3 consisted of orthogonal 2D projections from each individual binary structure. The fourth channel contained a summation of the other available binary structure masks. Structure classification performance was assessed in independent CT (n = 200 pat) and magnetic resonance imaging (MRI) (n = 40 pat) test datasets and an external CT (n = 99 pat) dataset from another clinic. A weighted classification accuracy of 99.4% was achieved during training. The unweighted classification accuracy and the weighted average F1 score among different structures in the CT test dataset were 98.8% and 98.4% and 98.6% and 98.5% for the MRI test dataset, respectively. The external CT dataset yielded the corresponding results 98.4% and 98.7% when analyzed for trained structures only, and results from the full dataset yielded 79.6% and 75.2%. Most misclassifications in the external CT dataset occurred due to multiple CTVs and PTVs being fused together, which was not included in the training data. Our proposed DL-based method for automated renaming and standardization of prostate radiotherapy annotations shows great potential. Clinic specific contouring standards however need to be represented in the training data for successful use. Source code is available at https://github.com/jamtheim/DicomRTStructRenamerPublic.

Investigation of the clinical inter-observer bias in prostate fiducial marker image registration between CT and MR images.

BACKGROUND AND PURPOSE: Inter-modality image registration between computed tomography (CT) and magnetic resonance (MR) images is associated with systematic uncertainties and the magnitude of these uncertainties is not well documented. The purpose of this study was to investigate the potential uncertainty of gold fiducial marker (GFM) registration for localized prostate cancer and to estimate the inter-observer bias in a clinical setting. METHODS: Four experienced observers registered CT and MR images for 42 prostate cancer patients. Manual GFM identification was followed by a landmark-based registration. The absolute difference between observers in GFM identification and the displacement of the clinical target volume (CTV) was investigated. The CTV center of mass (CoM) vector displacements, DICE-index and Hausdorff distances for the observer registrations were compared against a clinical baseline registration. The time allocated for the manual registrations was compared. RESULTS: Absolute difference in GFM identification between observers ranged from 0.0 to 3.0 mm. The maximum CTV CoM displacement from the clinical baseline was 3.1 mm. Displacements larger than or equal to 1 mm, 2 mm and 3 mm were 46%, 18% and 4%, respectively. No statistically significant difference was detected between observers in terms of CTV displacement. Median DICE-index and Hausdorff distance for the CTV, with their respective ranges were 0.94 [0.70-1.00] and 2.5 mm [0.7-8.7]. CONCLUSIONS: Registration of CT and MR images using GFMs for localized prostate cancer patients was subject to inter-observer bias on an individual patient level. A CTV displacement as large as 3 mm occurred for individual patients. These results show that GFM registration in a clinical setting is associated with uncertainties, which motivates the removal of inter-modality registrations in the radiotherapy workflow and a transition to an MRI-only workflow for localized prostate cancer.

Volumetric modulated arc therapy dose prediction and deliverable treatment plan generation for prostate cancer patients using a densely connected deep learning model.

BACKGROUND AND PURPOSE: Radiation therapy treatment planning is a manual, time-consuming task that might be accelerated using machine learning algorithms. In this study, we aimed to evaluate if a triplet-based deep learning model can predict volumetric modulated arc therapy (VMAT) dose distributions for prostate cancer patients. MATERIALS AND METHODS: A modified U-Net was trained on triplets, a combination of three consecutive image slices and corresponding segmentations, from 160 patients, and compared to a baseline U-Net. Dose predictions from 17 test patients were transformed into deliverable treatment plans using a novel planning workflow. RESULTS: The model achieved a mean absolute dose error of 1.3%, 1.9%, 1.0% and ≤ 2.6% for clinical target volume (CTV) CTV_D100%, planning target volume (PTV) PTV_D98%, PTV_D95% and organs at risk (OAR) respectively, when compared to the clinical ground truth (GT) dose distributions. All predicted distributions were successfully transformed into deliverable treatment plans and tested on a phantom, resulting in a passing rate of 100% (global gamma, 3%, 2 mm, 15% cutoff). The dose difference between deliverable treatment plans and GT dose distributions was within 4.4%. The difference between the baseline model and our improved model was statistically significant (p < 0.05) for CVT_D100%, PTV_D98% and PTV_D95%. CONCLUSION: Triplet-based training improved VMAT dose distribution predictions when compared to 2D. Dose predictions were successfully transformed into deliverable treatment plans using our proposed treatment planning procedure. Our method may automate parts of the workflow for external beam prostate radiation therapy and improve the overall treatment speed and plan quality.

Clinical validation of a commercially available deep learning software for synthetic CT generation for brain.

BACKGROUND: Most studies on synthetic computed tomography (sCT) generation for brain rely on in-house developed methods. They often focus on performance rather than clinical feasibility. Therefore, the aim of this work was to validate sCT images generated using a commercially available software, based on a convolutional neural network (CNN) algorithm, to enable MRI-only treatment planning for the brain in a clinical setting. METHODS: This prospective study included 20 patients with brain malignancies of which 14 had areas of resected skull bone due to surgery. A Dixon magnetic resonance (MR) acquisition sequence for sCT generation was added to the clinical brain MR-protocol. The corresponding sCT images were provided by the software MRI Planner (Spectronic Medical AB, Sweden). sCT images were rigidly registered and resampled to CT for each patient. Treatment plans were optimized on CT and recalculated on sCT images for evaluation of dosimetric and geometric endpoints. Further analysis was also performed for the post-surgical cases. Clinical robustness in patient setup verification was assessed by rigidly registering cone beam CT (CBCT) to sCT and CT images, respectively. RESULTS: All sCT images were successfully generated. Areas of bone resection due to surgery were accurately depicted. Mean absolute error of the sCT images within the body contour for all patients was 62.2 ± 4.1 HU. Average absorbed dose differences were below 0.2% for parameters evaluated for both targets and organs at risk. Mean pass rate of global gamma (1%/1 mm) for all patients was 100.0 ± 0.0% within PTV and 99.1 ± 0.6% for the full dose distribution. No clinically relevant deviations were found in the CBCT-sCT vs CBCT-CT image registrations. In addition, mean values of voxel-wise patient specific geometric distortion in the Dixon images for sCT generation were below 0.1 mm for soft tissue, and below 0.2 mm for air and bone. CONCLUSIONS: This work successfully validated a commercially available CNN-based software for sCT generation. Results were comparable for sCT and CT images in both dosimetric and geometric evaluation, for both patients with and without anatomical anomalies. Thus, MRI Planner is feasible to use for radiotherapy treatment planning of brain tumours.

Prospective Clinical Feasibility Study for MRI-Only Brain Radiotherapy.

OBJECTIVES: MRI-only radiotherapy (RT) provides a workflow to decrease the geometric uncertainty introduced by the image registration process between MRI and CT data and to streamline the RT planning. Despite the recent availability of validated synthetic CT (sCT) methods for the head region, there are no clinical implementations reported for brain tumors. Based on a preceding validation study of sCT, this study aims to investigate MRI-only brain RT through a prospective clinical feasibility study with endpoints for dosimetry and patient setup. MATERIAL AND METHODS: Twenty-one glioma patients were included. MRI Dixon images were used to generate sCT images using a CE-marked deep learning-based software. RT treatment plans were generated based on MRI delineated anatomical structures and sCT for absorbed dose calculations. CT scans were acquired but strictly used for sCT quality assurance (QA). Prospective QA was performed prior to MRI-only treatment approval, comparing sCT and CT image characteristics and calculated dose distributions. Additional retrospective analysis of patient positioning and dose distribution gamma evaluation was performed. RESULTS: Twenty out of 21 patients were treated using the MRI-only workflow. A single patient was excluded due to an MRI artifact caused by a hemostatic substance injected near the target during surgery preceding radiotherapy. All other patients fulfilled the acceptance criteria. Dose deviations in target were within ±1% for all patients in the prospective analysis. Retrospective analysis yielded gamma pass rates (2%, 2 mm) above 99%. Patient positioning using CBCT images was within ± 1 mm for registrations with sCT compared to CT. CONCLUSION: We report a successful clinical study of MRI-only brain radiotherapy, conducted using both prospective and retrospective analysis. Synthetic CT images generated using the CE-marked deep learning-based software were clinically robust based on endpoints for dosimetry and patient positioning.

MR-PROTECT: Clinical feasibility of a prostate MRI-only radiotherapy treatment workflow and investigation of acceptance criteria.

BACKGROUND: Retrospective studies on MRI-only radiotherapy have been presented. Widespread clinical implementations of MRI-only workflows are however limited by the absence of guidelines. The MR-PROTECT trial presents an MRI-only radiotherapy workflow for prostate cancer using a new single sequence strategy. The workflow incorporated the commercial synthetic CT (sCT) generation software MriPlanner™ (Spectronic Medical, Helsingborg, Sweden). Feasibility of the workflow and limits for acceptance criteria were investigated for the suggested workflow with the aim to facilitate future clinical implementations. METHODS: An MRI-only workflow including imaging, post imaging tasks, treatment plan creation, quality assurance and treatment delivery was created with questionnaires. All tasks were performed in a single MR-sequence geometry, eliminating image registrations. Prospective CT-quality assurance (QA) was performed prior treatment comparing the PTV mean dose between sCT and CT dose-distributions. Retrospective analysis of the MRI-only gold fiducial marker (GFM) identification, DVH- analysis, gamma evaluation and patient set-up verification using GFMs and cone beam CT were performed. RESULTS: An MRI-only treatment was delivered to 39 out of 40 patients. The excluded patient was too large for the predefined imaging field-of-view. All tasks could successfully be performed for the treated patients. There was a maximum deviation of 1.2% in PTV mean dose was seen in the prospective CT-QA. Retrospective analysis showed a maximum deviation below 2% in the DVH-analysis after correction for rectal gas and gamma pass-rates above 98%. MRI-only patient set-up deviation was below 2 mm for all but one investigated case and a maximum of 2.2 mm deviation in the GFM-identification compared to CT. CONCLUSIONS: The MR-PROTECT trial shows the feasibility of an MRI-only prostate radiotherapy workflow. A major advantage with the presented workflow is the incorporation of a sCT-generation method with multi-vendor capability. The presented single sequence approach are easily adapted by other clinics and the general implementation procedure can be replicated. The dose deviation and the gamma pass-rate acceptance criteria earlier suggested was achievable, and these limits can thereby be confirmed. GFM-identification acceptance criteria are depending on the choice of identification method and slice thickness. Patient positioning strategies needs further investigations to establish acceptance criteria.