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.

Pelvic U-Net: multi-label semantic segmentation of pelvic organs at risk for radiation therapy anal cancer patients using a deeply supervised shuffle attention convolutional neural network.

BACKGROUND: Delineation of organs at risk (OAR) for anal cancer radiation therapy treatment planning is a manual and time-consuming process. Deep learning-based methods can accelerate and partially automate this task. The aim of this study was to develop and evaluate a deep learning model for automated and improved segmentations of OAR in the pelvic region. METHODS: A 3D, deeply supervised U-Net architecture with shuffle attention, referred to as Pelvic U-Net, was trained on 143 computed tomography (CT) volumes, to segment OAR in the pelvic region, such as total bone marrow, rectum, bladder, and bowel structures. Model predictions were evaluated on an independent test dataset (n = 15) using the Dice similarity coefficient (DSC), the 95th percentile of the Hausdorff distance (HD95), and the mean surface distance (MSD). In addition, three experienced radiation oncologists rated model predictions on a scale between 1-4 (excellent, good, acceptable, not acceptable). Model performance was also evaluated with respect to segmentation time, by comparing complete manual delineation time against model prediction time without and with manual correction of the predictions. Furthermore, dosimetric implications to treatment plans were evaluated using different dose-volume histogram (DVH) indices. RESULTS: Without any manual corrections, mean DSC values of 97%, 87% and 94% were found for total bone marrow, rectum, and bladder. Mean DSC values for bowel cavity, all bowel, small bowel, and large bowel were 95%, 91%, 87% and 81%, respectively. Total bone marrow, bladder, and bowel cavity segmentations derived from our model were rated excellent (89%, 93%, 42%), good (9%, 5%, 42%), or acceptable (2%, 2%, 16%) on average. For almost all the evaluated DVH indices, no significant difference between model predictions and manual delineations was found. Delineation time per patient could be reduced from 40 to 12 min, including manual corrections of model predictions, and to 4 min without corrections. CONCLUSIONS: Our Pelvic U-Net led to credible and clinically applicable OAR segmentations and showed improved performance compared to previous studies. Even though manual adjustments were needed for some predicted structures, segmentation time could be reduced by 70% on average. This allows for an accelerated radiation therapy treatment planning workflow for anal cancer patients.

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.

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.

Correction for Ion Recombination in a Built-in Monitor Chamber of a Clinical Linear Accelerator at Ultra-High Dose Rates.

In the novel and promising radiotherapy technique known as FLASH, ultra-high dose-rate electron beams are used. As a step towards clinical trials, dosimetric advances will be required for accurate dose delivery of FLASH. The purpose of this study was to determine whether a built-in transmission chamber of a clinical linear accelerator can be used as a real-time dosimeter to monitor the delivery of ultra-high-dose-rate electron beams. This was done by modeling the drop-in ion-collection efficiency of the chamber with increasing dose-per-pulse values, so that the ion recombination effect could be considered. The raw transmission chamber signal was extracted from the linear accelerator and its response was measured using radiochromic film at different dose rates/dose-per-pulse values, at a source-to-surface distance of 100 cm. An increase of the polarizing voltage, applied over the transmission chamber, by a factor of 2 and 3, improved the ion-collection efficiency, with corresponding increased efficiency at the highest dose-per-pulse values by a factor 1.4 and 2.2, respectively. The drop-in ion-collection efficiency with increasing dose-per-pulse was accurately modeled using a logistic function fitted to the transmission chamber data. The performance of the model was compared to that of the general theoretical Boag models of ion recombination in ionization chambers. The logistic model was subsequently used to correct for ion recombination at dose rates ranging from conventional to ultra-high, making the transmission chamber useful as a real-time monitor for the dose delivery of FLASH electron beams in a clinical setup.

The FLASH effect depends on oxygen concentration.

OBJECTIVE: Recent in vivo results have shown prominent tissue sparing effect of radiotherapy with ultra-high dose rates (FLASH) compared to conventional dose rates (CONV). Oxygen depletion has been proposed as the underlying mechanism, but in vitro data to support this have been lacking. The aim of the current study was to compare FLASH to CONV irradiation under different oxygen concentrations in vitro. METHODS: Prostate cancer cells were irradiated at different oxygen concentrations (relative partial pressure ranging between 1.6 and 20%) with a 10 MeV electron beam at a dose rate of either 600 Gy/s (FLASH) or 14 Gy/min (CONV), using a modified clinical linear accelerator. We evaluated the surviving fraction of cells using clonogenic assays after irradiation with doses ranging from 0 to 25 Gy. RESULTS: Under normoxic conditions, no differences between FLASH and CONV irradiation were found. For hypoxic cells (1.6%), the radiation response was similar up to a dose of about 5-10 Gy, above which increased survival was shown for FLASH compared to CONV irradiation. The increased survival was shown to be significant at 18 Gy, and the effect was shown to depend on oxygen concentration. CONCLUSION: The in vitro FLASH effect depends on oxygen concentration. Further studies to characterize and optimize the use of FLASH in order to widen the therapeutic window are indicated. ADVANCES IN KNOWLEDGE: This paper shows in vitro evidence for the role of oxygen concentration underlying the difference between FLASH and CONV irradiation.

Modifying a clinical linear accelerator for delivery of ultra-high dose rate irradiation.

OBJECTIVES: The purpose of this study was to modify a clinical linear accelerator, making it capable of electron beam ultra-high dose rate (FLASH) irradiation. Modifications had to be quick, reversible, and without interfering with clinical treatments. METHODS: Performed modifications: (1) reduced distance with three setup positions, (2) adjusted/optimized gun current, modulator charge rate and beam steering values for a high dose rate, (3) delivery was controlled with a microcontroller on an electron pulse level, and (4) moving the primary and/or secondary scattering foils from the beam path. RESULTS: The variation in dose for a five-pulse delivery was measured to be 1% (using a diode, 4% using film) during 10 minutes after a warm-up procedure, later increasing to 7% (11% using film). A FLASH irradiation dose rate was reached at the cross-hair foil, MLC, and wedge position, with ≥30, ≥80, and ≥300 Gy/s, respectively. Moving the scattering foils resulted in an increased output of ≥120, ≥250, and ≥1000 Gy/s, at the three positions. The beam flatness was 5% at the cross-hair position for a 20 × 20 and a 10 × 10 cm2 area, with and without both scattering foils in the beam. The beam flatness was 10% at the wedge position for a 6 and 2.5 cm diametric area, with and without the scattering foils in the beam path. CONCLUSIONS: A clinical accelerator was modified to produce ultra-high dose rates, high enough for FLASH irradiation. Future work aims to fine-tune the dose delivery, using the on-board transmission chamber signal and adjusting the dose-per-pulse.

Development of a novel radiotherapy motion phantom using a stepper motor driver circuit and evaluation using optical surface scanning.

Recent developments in radiotherapy have focused on the management of patient motion during treatment. Studies have shown that significant gains in treatment quality can be made by 'gating' certain treatments, simultaneously keeping target coverage, and increasing separation to nearby organs at risk (OAR). Motion phantoms can be used to simulate patient breathing motion and provide the means to perform quality control (QC) and quality assurance (QA) of gating functionality as well as to assess the dosimetric impact of motion on individual patient treatments. The aim of this study was to design and build a motion phantom that accurately reproduces the breathing motion of patients to enable end-to-end gating system quality control of various gating systems as well as patient specific quality assurance. A motion phantom based on a stepper motor driver circuit was designed. The phantom can be programmed with both real patient data from an external gating system and with custom signals. The phantom was programmed and evaluated with patient data and with a square wave signal to be tracked with a Sentinel™ (C-Rad, Uppsala, Sweden) motion monitoring system. Results were compared to the original curves with respect to amplitude and phase. The comparison of patient curve data showed a mean error value of -0.09 mm with a standard deviation of 0.24 mm and a mean absolute error of 0.29 mm. The square wave signals could be reproduced with a mean error value of -0.03 mm, a standard deviation of 0.04 mm and a mean absolute error of 0.13 mm. Breathing curve data acquired from an optical scanning system can be reproduced accurately with the help of the in-house built motion phantom. The phantom can also be programmed to follow user designed curve data. This offers the potential for QC of gating systems and various dosimetric quality control applications.