Development of a national deep learning-based auto-segmentation model for the heart on clinical delineations from the DBCG RT nation cohort.

BACKGROUND: This study aimed at investigating the feasibility of developing a deep learning-based auto-segmentation model for the heart trained on clinical delineations. MATERIAL AND METHODS: This study included two different datasets. The first dataset contained clinical heart delineations from the DBCG RT Nation study (1,561 patients). The second dataset was smaller (114 patients), but with corrected heart delineations. Before training the model on the clinical delineations an outlier-detection was performed, to remove cases with gross deviations from the delineation guideline. No outlier detection was performed for the dataset with corrected heart delineations. Both models were trained with a 3D full resolution nnUNet. The models were evaluated with the dice similarity coefficient (DSC), 95% Hausdorff distance (HD95) and Mean Surface Distance (MSD). The difference between the models were tested with the Mann-Whitney U-test. The balance of dataset quantity versus quality was investigated, by stepwise reducing the cohort size for the model trained on clinical delineations. RESULTS: During the outlier-detection 137 patients were excluded from the clinical cohort due to non-compliance with delineation guidelines. The model trained on the curated clinical cohort performed with a median DSC of 0.96 (IQR 0.94-0.96), median HD95 of 4.00 mm (IQR 3.00 mm-6.00 mm) and a median MSD of 1.49 mm (IQR 1.12 mm-2.02 mm). The model trained on the dedicated and corrected cohort performed with a median DSC of 0.95 (IQR 0.93-0.96), median HD95 of 5.65 mm (IQR 3.37 mm-8.62 mm) and median MSD of 1.63 mm (IQR 1.35 mm-2.11 mm). The difference between the two models were found non-significant for all metrics (p > 0.05). Reduction of cohort size showed no significant difference for all metrics (p > 0.05). However, with the smallest cohort size, a few outlier structures were found. CONCLUSIONS: This study demonstrated a deep learning-based auto-segmentation model trained on curated clinical delineations which performs on par with a model trained on dedicated delineations, making it easier to develop multi-institutional auto-segmentation models.

Comparing different CT, PET and MRI multi-modality image combinations for deep learning-based head and neck tumor segmentation.

BACKGROUND: Manual delineation of gross tumor volume (GTV) is essential for radiotherapy treatment planning, but it is time-consuming and suffers inter-observer variability (IOV). In clinics, CT, PET, and MRI are used to inform delineation accuracy due to their different complementary characteristics. This study aimed to investigate deep learning to assist GTV delineation in head and neck squamous cell carcinoma (HNSCC) by comparing various modality combinations. MATERIALS AND METHODS: This retrospective study had 153 patients with multiple sites of HNSCC including their planning CT, PET, and MRI (T1-weighted and T2-weighted). Clinical delineations of gross tumor volume (GTV-T) and involved lymph nodes (GTV-N) were collected as the ground truth. The dataset was randomly divided into 92 patients for training, 31 for validation, and 30 for testing. We applied a residual 3 D UNet as the deep learning architecture. We independently trained the UNet with four different modality combinations (CT-PET-MRI, CT-MRI, CT-PET, and PET-MRI). Additionally, analogical to post-processing, an average fusion of three bi-modality combinations (CT-PET, CT-MRI, and PET-MRI) was produced as an ensemble. Segmentation accuracy was evaluated on the test set, using Dice similarity coefficient (Dice), Hausdorff Distance 95 percentile (HD95), and Mean Surface Distance (MSD). RESULTS: All imaging combinations including PET provided similar average scores in range of Dice: 0.72-0.74, HD95: 8.8-9.5 mm, MSD: 2.6-2.8 mm. Only CT-MRI had a lower score with Dice: 0.58, HD95: 12.9 mm, MSD: 3.7 mm. The average of three bi-modality combinations reached Dice: 0.74, HD95: 7.9 mm, MSD: 2.4 mm. CONCLUSION: Multimodal deep learning-based auto segmentation of HNSCC GTV was demonstrated and inclusion of the PET image was shown to be crucial. Training on combined MRI, PET, and CT data provided limited improvements over CT-PET and PET-MRI. However, when combining three bimodal trained networks into an ensemble, promising improvements were shown.

Fully automated radiotherapy treatment planning: A scan to plan challenge.

BACKGROUND AND PURPOSE: Over the past decade, tools for automation of various sub-tasks in radiotherapy planning have been introduced, such as auto-contouring and auto-planning. The purpose of this study was to benchmark what degree of automation is possible. MATERIALS AND METHODS: A challenge to perform automated treatment planning for prostate and prostate bed radiotherapy was set up. Participants were provided with simulation CTs and a treatment prescription and were asked to use automated tools to produce a deliverable radiotherapy treatment plan with as little human intervention as possible. Plans were scored for their adherence to the protocol when assessed using consensus expert contours. RESULTS: Thirteen entries were received. The top submission adhered to 81.8% of the minimum objectives across all cases using the consensus contour, meeting all objectives in one of the ten cases. The same system met 89.5% of objectives when assessed with their own auto-contours, meeting all objectives in four of the ten cases. The majority of systems used in the challenge had regulatory clearance (Auto-contouring: 82.5%, Auto-planning: 77%). Despite the 'hard' rule that participants should not check or edit contours or plans, 69% reported looking at their results before submission. CONCLUSIONS: Automation of the full planning workflow from simulation CT to deliverable treatment plan is possible for prostate and prostate bed radiotherapy. While many generated plans were found to require none or minor adjustment to be regarded as clinically acceptable, the result indicated there is still a lack of trust in such systems preventing full automation.

Potential of E-Learning Interventions and Artificial Intelligence-Assisted Contouring Skills in Radiotherapy: The ELAISA Study.

PURPOSE: Most research on artificial intelligence-based auto-contouring as template (AI-assisted contouring) for organs-at-risk (OARs) stem from high-income countries. The effect and safety are, however, likely to depend on local factors. This study aimed to investigate the effects of AI-assisted contouring and teaching on contouring time and contour quality among radiation oncologists (ROs) working in low- and middle-income countries (LMICs). MATERIALS AND METHODS: Ninety-seven ROs were randomly assigned to either manual or AI-assisted contouring of eight OARs for two head-and-neck cancer cases with an in-between teaching session on contouring guidelines. Thereby, the effect of teaching (yes/no) and AI-assisted contouring (yes/no) was quantified. Second, ROs completed short-term and long-term follow-up cases all using AI assistance. Contour quality was quantified with Dice Similarity Coefficient (DSC) between ROs' contours and expert consensus contours. Groups were compared using absolute differences in medians with 95% CIs. RESULTS: AI-assisted contouring without previous teaching increased absolute DSC for optic nerve (by 0.05 [0.01; 0.10]), oral cavity (0.10 [0.06; 0.13]), parotid (0.07 [0.05; 0.12]), spinal cord (0.04 [0.01; 0.06]), and mandible (0.02 [0.01; 0.03]). Contouring time decreased for brain stem (-1.41 [-2.44; -0.25]), mandible (-6.60 [-8.09; -3.35]), optic nerve (-0.19 [-0.47; -0.02]), parotid (-1.80 [-2.66; -0.32]), and thyroid (-1.03 [-2.18; -0.05]). Without AI-assisted contouring, teaching increased DSC for oral cavity (0.05 [0.01; 0.09]) and thyroid (0.04 [0.02; 0.07]), and contouring time increased for mandible (2.36 [-0.51; 5.14]), oral cavity (1.42 [-0.08; 4.14]), and thyroid (1.60 [-0.04; 2.22]). CONCLUSION: The study suggested that AI-assisted contouring is safe and beneficial to ROs working in LMICs. Prospective clinical trials on AI-assisted contouring should, however, be conducted upon clinical implementation to confirm the effects.

Evaluation of methods for selecting the midventilation bin in 4DCT scans of lung cancer patients.

BACKGROUND: In lung cancer radiotherapy, planning on the midventilation (MidV) bin of a four-dimensional (4D) CT scan can reduce the systematic errors introduced by respiratory tumour motion compared to conventional CT. In this study four different methods for MidV bin selection are evaluated. MATERIAL AND METHODS: The study is based on 4DCT scans of 19 patients with a total of 23 peripheral lung tumours having peak-to-peak displacement ≥ 5 mm in at least one of the left-right (LR), anterior-posterior (AP) or cranio-caudal (CC) directions. For each tumour, the MidV bin was selected based on: 1) visual evaluation of tumour displacement; 2) rigid registration of tumour position; 3) diaphragm displacement in the CC direction; and 4) carina displacement in the CC direction. Determination of the MidV bin based on the displacement of the manually delineated gross tumour volume (GTV) was used as a reference method. The accuracy of each method was evaluated by the distance between GTV position in the selected MidV bin and the time-weighted mean position of GTV throughout the bins (i.e. the geometric MidV error). RESULTS: Median (range) geometric MidV error was 1.4 (0.4-5.4) mm, 1.4 (0.4-5.4) mm, 1.9 (0.5-6.9) mm, 2.0 (0.5-12.3) mm and 1.1 (0.4-5.4) mm for the visual, rigid registration, diaphragm, carina, and reference method. Median (range) absolute difference between geometric MidV error for the evaluated methods and the reference method was 0.0 (0.0-1.2) mm, 0.0 (0.0-1.7) mm, 0.7 (0.0-3.9) mm and 1.0 (0.0-6.9) mm for the visual, rigid registration, diaphragm and carina method. CONCLUSION: The visual and semi-automatic rigid registration methods were equivalent in accuracy for selecting the MidV bin of a 4DCT scan. The methods based on diaphragm and carina displacement cannot be recommended without modifications.

Towards interactive deep-learning for tumour segmentation in head and neck cancer radiotherapy.

Background and purpose: With deep-learning, gross tumour volume (GTV) auto-segmentation has substantially been improved, but still substantial manual corrections are needed. With interactive deep-learning (iDL), manual corrections can be used to update a deep-learning tool while delineating, minimising the input to achieve acceptable segmentations. We present an iDL tool for GTV segmentation that took annotated slices as input and simulated its performance on a head and neck cancer (HNC) dataset. Materials and methods: Multimodal image data of 204 HNC patients with clinical tumour and lymph node GTV delineations were used. A baseline convolutional neural network (CNN) was trained (n = 107 training, n = 22 validation) and tested (n = 24). Subsequently, user input was simulated on initial test set by replacing one or more of predicted slices with ground truth delineation, followed by re-training the CNN. The objective was to optimise re-training parameters and simulate slice selection scenarios while limiting annotations to maximally-five slices. The remaining 51 patients were used as an independent test set, where Dice similarity coefficient (DSC), mean surface distance (MSD), and 95% Hausdorff distance (HD95%) were assessed at baseline and after every update. Results: Median segmentation accuracy at baseline was DSC = 0.65, MSD = 4.3 mm, HD95% = 17.5 mm. Updating CNN using three slices equally sampled from the craniocaudal axis of the GTV in the first round, followed by two rounds of annotating one extra slice, gave the best results. The accuracy improved to DSC = 0.82, MSD = 1.6 mm, HD95% = 4.8 mm. Every CNN update took 30 s. Conclusions: The presented iDL tool achieved substantial segmentation improvement with only five annotated slices.

A simple single-cycle interactive strategy to improve deep learning-based segmentation of organs-at-risk in head-and-neck cancer.

Background and purpose: Interactive segmentation seeks to incorporate human knowledge into segmentation models and thereby reducing the total amount of editing of auto-segmentations. By performing only interactions which provide new information, segmentation performance may increase cost-effectively. The aim of this study was to develop, evaluate and test feasibility of a deep learning-based single-cycle interactive segmentation model with the input being computer tomography (CT) and a small amount of information rich contours. Methods and Materials: A single-cycle interactive segmentation model, which took CT and the most cranial and caudal contour slices for each of 16 organs-at-risk for head-and-neck cancer as input, was developed. A CT-only model served as control. The models were evaluated with Dice similarity coefficient, Hausdorff Distance 95th percentile and average symmetric surface distance. A subset of 8 organs-at-risk were selected for a feasibility test. In this, a designated radiation oncologist used both single-cycle interactive segmentation and atlas-based auto-contouring for three cases. Contouring time and added path length were recorded. Results: The medians of Dice coefficients increased with single-cycle interactive segmentation in the range of 0.004 (Brain)-0.90 (EyeBack_merged) when compared to CT-only. In the feasibility test, contouring time and added path length were reduced for all three cases as compared to editing atlas-based auto-segmentations. Conclusion: Single-cycle interactive segmentation improved segmentation metrics when compared to the CT-only model and was clinically feasible from a technical and usability point of view. The study suggests that it may be cost-effective to add a small amount of contouring input to deep learning-based segmentation models.

Patient anatomy-specific trade-offs between sub-clinical disease coverage and normal tissue dose reduction in head-and-neck cancer.

PURPOSE: Risk of subclinical disease decreases with increasing distance from the GTV in head- and-neck squamous cell carcinoma (HNSCC). Depending on individual patient anatomy, OAR sparing could be improved by reducing target coverage in regions with low risk of subclinical spread. Using automated multi-criteria optimization, we investigate patient-specific optimal trade-offs between target periphery coverage and OAR sparing. METHODS: VMAT plans for 39 HNSCC patients were retrospectively created following our clinical three-target-level protocol: high-risk (PTV1), intermediate-risk (PTV2, 5 mm expansion from PTV1), and elective (PTV3). A baseline plan fulfilling clinical constraints (D 99 % ≥95 % for all PTVs) was compared to three plans with reduced PTV2 coverage (goals: PTV2 D 99 % ≥90 % or 85 %, or no PTV2) at the outer edge of PTV2. Plans were compared on PTV D 99 %, OAR D mean, and NTCP (xerostomia/dysphagia). RESULTS: Trade-offs between PTV2 coverage and OAR doses varied considerably between patients. For plans with PTV2 D 99 % -goal 90 %, median PTV2 D 99 % was 91.5 % resulting in xerostomia (≥grade 4) and dysphagia (≥grade 2) NTCP decrease of median [maximum] 1.9 % [5.3 %] and 1.1 % [4.1 %], respectively, compared to nominal PTV2 D 99 % -goal 95 %. For PTV2 D 99 % -goal 85 % median PTV D 99 % was 87 % with NTCP improvements of 4.6 % [9.9 %] and 1.5 % [5.4 %]. For no-margin plans, PTV2 D 99 % decreased to 83.3 % with NTCP reductions of 5.1 % [10.2 %] and 1.4 % [6.1 %]. CONCLUSION: Clinically relevant, patient-specific reductions in OARs and NTCP were observed at limited cost in target under-coverage at the outermost PTV edge. Given the observed inter-patient variations, individual evaluation is warranted to determine whether trade- offs would benefit a specific patient.

End-to-end framework for automated collection of large multicentre radiotherapy datasets demonstrated in a Danish Breast Cancer Group cohort.

Large Digital Imaging and Communications in Medicine (DICOM) datasets are key to support research and the development of machine learning technology in radiotherapy (RT). However, the tools for multi-centre data collection, curation and standardisation are not readily available. Automated batch DICOM export solutions were demonstrated for a multicentre setup. A Python solution, Collaborative DICOM analysis for RT (CORDIAL-RT) was developed for curation, standardisation, and analysis of the collected data. The setup was demonstrated in the DBCG RT-Nation study, where 86% (n = 7748) of treatments in the inclusion period were collected and quality assured, supporting the applicability of the end-to-end framework.

A systematically compiled set of quantitative metrics to describe spatial characteristics of radiotherapy dose distributions and aid in treatment planning.

PURPOSE: Many quantitative metrics have been proposed in literature for characterization of spatial dose properties. The aim of this study is to work towards much-needed consensus in the radiotherapy community on which of these metrics to use. We do this by comparing characteristics of the metrics and providing a systematically selected set of metrics to comprehensively quantify properties of the spatial dose distribution. METHODS: We searched the literature for metrics to quantitatively evaluate dose conformity, homogeneity, gradient (overall and directional), and distribution and location of over- and under-dosed sub-volumes. For each spatial dose property, we compared the responses of its corresponding metrics to simulated dose variations in a virtual water phantom. Selection criteria were a metric's ability to describe simulated scenarios robustly and to be visualized in an intuitive way. RESULTS: We saw substantial differences in the responses of metrics to the simulated dose variations. Some conformity and homogeneity metrics were unable to quantify certain types of changes (e.g. target under-coverage). Others showed a large dependency on the shape and volume of targets and isodoses. Metric values differed between calculations in a static plan and in simulated full treatment courses including setup errors, especially for metrics quantifying distribution and location of hot and cold spots. We provide an Eclipse plugin script to calculate and visualize selected metrics. CONCLUSION: The selected set of metrics provides complementary and comprehensive quantitative information about the spatial dose distribution. This work serves as a step towards broader consensus on the use of spatial dose metrics.

Intra- and interfraction breathing variations during curative radiotherapy for lung cancer.

BACKGROUND AND PURPOSE: This study aimed at quantifying the breathing variations among lung cancer patients over full courses of fractionated radiotherapy. The intention was to relate these variations to the margins assigned to lung tumours, to account for respiratory motion, in fractionated radiotherapy. MATERIALS AND METHODS: Eleven lung cancer patients were included in the study. The patients' chest wall motions were monitored as a surrogate measure for breathing motion during each fraction of radiotherapy by use of an external optical marker. The exhale level variations were evaluated with respect to exhale points and fraction-baseline, defined for intra- and interfraction variations respectively. The breathing amplitude was evaluated as breathing cycle amplitudes and fraction-max-amplitudes defined for intra- and interfraction breathing, respectively. RESULTS: The breathing variations over a full treatment course, including both intra- and interfraction variations, were 15.2mm (median over the patient population), range 5.5-26.7mm, with the variations in exhale level as the major contributing factor. The median interfraction span in exhale level was 14.8mm, whereas the median fraction-max-amplitude was 6.1mm (median of patient individual SD 1.4). The median intrafraction span in exhale level was 1.6mm, and the median breathing cycle amplitude was 4.0mm (median of patient individual SD 1.4). CONCLUSIONS: The variations in externally measured exhale levels are larger than variations in breathing amplitude. The interfraction variations in exhale level are in general are up to 10 times larger than intrafraction variations. Margins to account for respiratory motion cannot safely be based on one planning session, especially not if relying on measuring external marker motion. Margins for lung tumours should include interfraction variations in breathing.

Artifacts in conventional computed tomography (CT) and free breathing four-dimensional CT induce uncertainty in gross tumor volume determination.

PURPOSE: Artifacts impacting the imaged tumor volume can be seen in conventional three-dimensional CT (3DCT) scans for planning of lung cancer radiotherapy but can be reduced with the use of respiration-correlated imaging, i.e., 4DCT or breathhold CT (BHCT) scans. The aim of this study was to compare delineated gross tumor volume (GTV) sizes in 3DCT, 4DCT, and BHCT scans of patients with lung tumors. METHODS AND MATERIALS: A total of 36 patients with 46 tumors referred for stereotactic radiotherapy of lung tumors were included. All patients underwent positron emission tomography (PET)/CT, 4DCT, and BHCT scans. GTVs in all CT scans of individual patients were delineated during one session by a single physician to minimize systematic delineation uncertainty. The GTV size from the BHCT was considered the closest to true tumor volume and was chosen as the reference. The reference GTV size was compared to GTV sizes in 3DCT, at midventilation (MidV), at end-inspiration (Insp), and at end-expiration (Exp) bins from the 4DCT scan. RESULTS: The median BHCT GTV size was 4.9 cm(3) (0.1-53.3 cm(3)). Median deviation between 3DCT and BHCT GTV size was 0.3 cm(3) (-3.3 to 30.0 cm(3)), between MidV and BHCT size was 0.2 cm(3) (-5.7 to 19.7 cm(3)), between Insp and BHCT size was 0.3 cm(3) (-4.7 to 24.8 cm(3)), and between Exp and BHCT size was 0.3 cm(3) (-4.8 to 25.5 cm(3)). The 3DCT, MidV, Insp, and Exp median GTV sizes were all significantly larger than the BHCT median GTV size. CONCLUSIONS: In the present study, the choice of CT method significantly influenced the delineated GTV size, on average, leading to an increase in GTV size compared to the reference BHCT. The uncertainty caused by artifacts is estimated to be in the same magnitude as delineation uncertainty and should be considered in the design of margins for radiotherapy.

Deviations in delineated GTV caused by artefacts in 4DCT.

BACKGROUND AND PURPOSE: Four-dimensional computed tomography (4DCT) is used for breathing-adapted radiotherapy planning. Irregular breathing, large tumour motion or interpolation of images can cause artefacts in the 4DCT. This study evaluates the impact of artefacts on gross tumour volume (GTV) size. MATERIAL AND METHODS: In 19 4DCT scans of patients with peripheral lung tumours, GTV was delineated in all bins. Variations in GTV size between bins in each 4DCT scan were analysed and correlated to tumour motion and variations in breathing signal amplitude and breathing signal period. End-expiration GTV size (GTVexp) was considered as reference for GTV size. Intra-session delineation error was estimated by re-delineation of GTV in eight of the 4DCT scans. RESULTS: In 16 of the 4DCT scans the maximum deviations from GTVexp were larger than could be explained by delineation error. The deviations were largest in the bins adjacent to the end-inspiration bin. The coefficient of variation of GTV size was significantly correlated to tumour motion in the cranio-caudal direction, but no significant correlation was found to breathing signal variations. CONCLUSION: We found considerable variations in GTV size throughout the 4DCT scans. Awareness of the error introduced by artefacts is important especially if radiotherapy planning is based on a single 4DCT bin.