Planning CTs Research Articles

CT- and MR-based image-based data mining are consistent in the brain

PurposeImage-based data mining (IBDM) is a voxel-based analysis technique to investigate dose–response. Most often, IBDM uses radiotherapy planning CTs because of their broad accessibility, however, it was unknown whether CT provided sufficient soft tissue contrast for brain IBDM. This study evaluates whether MR-based IBDM improves upon CT-based IBDM for studies of children with brain tumours. MethodsWe compared IBDM pipelines using either CT- or MRI-based spatial normalisation in 128 children (ages 3.3–19.7 years) who received photon radiotherapy for primary brain tumours at a single institution. We quantified spatial-normalisation accuracy using contour comparison measures (centre-of-mass separation, average contour distance-to-agreement (DTavg), and Hausdorff distance) at multiple anatomic loci. We performed an end-to-end test of CT- and MRI-IBDM using modified clinical dose distributions and simulated effect labels to detect associations in pre-defined anatomic loci. Accuracy was assessed via sensitivity and specificity. ResultsSpatial normalisation accuracy was comparable for both modalities, with a significant but small improvement for MRI compared to CT in all structures except the brainstem. The median (range) difference between the DTavg for the two pipelines was 0.37 (0.00–2.91) mm. The end-to-end test revealed no significant difference in sensitivity of the IBDM-identified regions for the two pipelines. Specificity slightly improved for MR-IBDM at the 99% significance level. ConclusionCT-based IBDM was comparable to MR-based IBDM, despite a small advantage in spatial normalisation accuracy with MRI. The use of CT-IBDM over MR-IBDM is useful for multi-institutional retrospective IBDM studies, where the availability of standardised MRI data can be limited.

Technical note: Evaluation of deep learning based synthetic CTs clinical readiness for dose and NTCP driven head and neck adaptive proton therapy.

Adaptive proton therapy workflows rely on accurate imaging throughout the treatment course. Our centre currently utilizes weekly repeat CTs (rCTs) for treatment monitoring and plan adaptations. However, deep learning-based methods have recently shown to successfully correct CBCT images, which suffer from severe imaging artifacts, and generate high quality synthetic CT (sCT) images which enable CBCT-based proton dose calculations. To compare daily CBCT-based sCT images to planning CTs (pCT) and rCTs of head and neck (HN) cancer patients to investigate the dosimetric accuracy of CBCT-based sCTs in a scenario mimicking actual clinical practice. Data of 56 HN cancer patients, previously treated with proton therapy was used to generate 1.962 sCT images, using a previously developed and trained deep convolutional neural network. Clinical IMPT treatment plans were recalculated on the pCT, weekly rCTs and daily sCTs. The dosimetric accuracy of sCTs was compared to same day rCTs and the initial planning CT. As a reference, rCTs were also compared to pCTs. The dose difference between sCTs and rCTs/pCT was quantified by calculating the D98 difference for target volumes and Dmean difference for organs-at-risk. To investigate the clinical relevancy of possible dose differences, NTCP values were calculated for dysphagia and xerostomia. For target volumes, only minor dose differences were found for sCT versus rCT and sCT versus pCT, with dose differences mostly within ±1.5%. Larger dose differences were observed in OARs, where a general shift towards positive differences was found, with the largest difference in the left parotid gland. Delta NTCP values for grade 2 dysphagia and xerostomia were within ±2.5% for 90% of the sCTs. Target doses showed high similarity between rCTs and sCTs. Further investigations are required to identify the origin of the dose differences at OAR levels and its relevance in clinical decision making.

Association of Left Anterior Descending Coronary Artery Calcium Progression and Radiation Dose with Major Adverse Cardiac Events in Breast Cancer

Coronary artery calcium (CAC) is associated with increased risk of major adverse cardiac events (MACE). Accelerated CAC progression has been observed in patients with breast cancer after radiotherapy (RT) and there is a relationship between left anterior descending (LAD) coronary artery RT dose and the risk of coronary events. However, there is lack of consensus on LAD dose constraints for breast RT and limited data on the extent and impact of CAC progression. Our objective was to evaluate the association of LAD dose exposure and CAC progression with the risk of MACE in patients with breast cancer following RT. Retrospective analysis of 181 patients with breast cancer treated with RT between 2008 and 2019. CAC was manually measured on RT planning and follow-up CTs (with at least one-year interval) using the Agatston method. Coronary arteries were segmented using a deep learning-based automated algorithm and dosimetric parameters collected. MACE cumulative incidence was estimated, and Fine and Gray regressions performed, accounting for non-cardiac death as a competing risk. The median follow-up following RT was 70 months (interquartile range [IQR], 53-86). The median age was 63 years (IQR, 53-72), 43% had hypertension, 40% hyperlipidemia, 8% coronary heart disease (CHD). Most had pathologic stage I-II disease (76%). RT was targeted to breast/chest wall only in 60% and included regional nodes in 40% (internal mammary chain in 4%). The most common dose/fractionation was 48-50.4 Gy/25-28 fractions (67%) and 42.6-42.7 Gy/16 fractions (30%). At the time of RT, 68 (38%) had at least moderate CAC burden (CAC >100; statin-therapy indicated), but only 29 (43%) were on statin therapy. At a median interval of 44 months (IQR, 26-63), 55% (n = 84) had CAC progression, with a median increase of 52%/year (IQR, 18-193). The median time to MACE was 68 months (IQR, 53-85), with a 5-year cumulative incidence of 7.3% (15 MACE overall). Accounting for age and CHD, there was an increased risk of MACE with LAD CAC progression (subdistribution hazard ratio [SHR] 1.02/10 CAC points; 95% confidence interval [CI] 1.01 = 1.03; p = .007) and the volume of LAD receiving 15 Gy (LAD V15 Gy; SHR 1.03/%; 95% CI, 1.01-1.06; p = .004). There was no association between mean heart dose, chemotherapy, or Her2 therapy exposure and MACE (p>.05). LAD CAC progression and LAD V15 Gy dose exposure were associated with an increased risk of MACE following RT. Accelerated CAC progression was commonly observed, however most patients were under-optimized for cardiovascular (CV) risk, with less than half of statin-eligible patients with at least moderate CAC burden on statin therapy. Together, these data support more aggressive cardiac risk mitigation approaches, including guidelines-based CV risk factor modification and optimized sparing of LAD radiation dose.

Analysis of the Cancer Imaging Archive for Patterns of Locoregional Failure in Relation to the Mucous Membranes among Patients with Head and Neck Squamous Cell Carcinoma: Potential Organ at Risk that can be Spared?

Radiation-induced oral mucositis (RIOM), a painful inflammation and ulceration of the mucous membranes (MM), occurs as a common and severe side effect of radiation treatment (RT) for head and neck cancers. It worsens significant quality of life leading to treatment interruption or cancellation. In RIOM, radiation injury of the MM basal epithelial cells may be one of the critical pathophysiological mechanisms. However, it is unclear if advanced radiation treatments could spare MM, without compromise on tumor control. The purpose of this study is to determine whether the MM that do not overlap with the prescription target volumes are a common site of locoregional recurrence in head and neck cancers. A de-identified data collection of 215 head and neck cancer patients stored within The Cancer Imaging Archive was downloaded for analysis. Imaging series both before and after RT were downloaded, as well as RT structures and RT dosimetric data. Diagnostic CECT and/or PET/CT and/or MRI scans documenting the initial evidence of recurrence were co-registered with corresponding original planning CTs. Radiologically evident recurrent gross disease (rGTV) was manually contoured, as well as the MM (defined as an approximately 3mm thick wall of mucosa tissue lining the oral cavity and pharynx and excluding any overlap with the radiation prescription targets). Contoured MM structures were cross-referenced with the sites of recurrence by examining the overlap of the MM volumes with the rGTV volumes. A total of 40 patients had documented evidence of recurrence; 38 of which had images available for analysis. Overall, 23 (60.5%) were oropharynx cancer. Of the 38 recurrences, 35 (92%) were locoregional: 23 with only local (60.5%) recurrences, 9 with only regional (23.7%) recurrences and 3 (7.9%) with both local and regional recurrences. One patient (2.6%) had distant recurrence, one patient had both regional and distant recurrences concomitantly, and one patient had local, regional and distant recurrences simultaneously. Overall, 30 patients (78.9%) had recurrences located exclusively within prescription PTV volumes/isodose lines. Four patients (10.5%) had recurrences located outside prescription PTV volumes/isodose lines, four patients (10.5%) had recurrences located both within and outside prescription PTV volumes/isodose lines (i.e., 2 sites of concomitant recurrence). Importantly, 0 patients (0%) had recurrences located in the contoured MM (excluding any overlap with the radiation prescription targets). Based on this retrospective series, the MM (excluding the areas of overlap with radiation prescription targets) are a rare location of locoregional recurrence. Hence, sparing the MM using radiation advanced dosimetry techniques could be a reasonable approach to reducing the incidence/severity of radiation-induced oral mucositis, without compromising tumor control. However, this requires further prospective investigation.

Radiation dose to heart and cardiac substructures and risk of coronary artery disease in early breast cancer patients: A DBCG study based on modern radiation therapy techniques

BackgroundCoronary artery disease (CAD) has been reported as a late effect following radiation therapy (RT) of early breast cancer (BC). This study aims to report individual RT doses to the heart and cardiac substructures in patients treated with CT-based RT and to investigate if a dose–response relationship between RT dose and CAD exists using modern radiation therapy techniques. MethodsPatients registered in the Danish Breast Cancer Group database from 2005 to 2016 with CT-based RT were eligible. Among 15,765 patients, the study included 204 with CAD after irradiation (cases) and 408 matched controls. Individual planning CTs were retrieved, the heart and cardiac substructures were delineated and dose-volume parameters were extracted. ResultsThe median follow-up time was 7.3 years (IQR: 4.6–10.0). Among cases, the median mean heart dose was 1.6 Gy (IQR 0.2–6.1) and 0.8 Gy (0.1–2.9) for left-sided and right-sided patients, respectively (p < 0.001). The highest RT doses were observed in the left ventricle and left anterior descending coronary artery for left-sided RT and in the right atrium and the right coronary artery after right-sided RT. The highest left-minus-right dose-difference was located in the distal part of the left anterior descending coronary artery where also the highest left-versus-right ratio of events was observed. However, no significant difference in the distribution of CAD was observed by laterality. Furthermore, no significant differences in the dose-volume parameters were observed for cases versus controls. ConclusionsCAD tended to occur in the part of the heart with the highest left-minus- right dose difference, however, no significant risk of CAD was observed at 7 years' median follow-up.

Automatic detection and classification of treatment deviations in proton therapy from realistically simulated prompt gamma imaging data.

A clinical study regarding the potential of range verification in proton therapy (PT) by prompt gamma imaging (PGI) is carried out at our institution. Manual interpretation of the detected spot-wise range shift information is time-consuming, highly complex, and therefore not feasible in a broad routine application. Here, we present an approach to automatically detect and classify treatment deviations in realistically simulated PGI data for head-and-neck cancer (HNC) treatments using convolutional neural networks (CNNs) and conventional machine learning (ML) approaches. For 12 HNC patients and 1 anthropomorphic head phantom (n=13), pencil beam scanning (PBS) treatment plans were generated, and 1 field per plan was assumed to be monitored with a PGI slit camera system. In total, 386 scenarios resembling different relevant or non-relevant treatment deviations were simulated on planning and control CTs and manually classified into 7 classes: non-relevant changes (NR) and relevant changes (RE) triggering treatment intervention due to range prediction errors (±RP), setup errors in beam direction (±SE), anatomical changes (AC), or a combination of such errors (CB). PBS spots with reliable PGI information were considered with their nominal Bragg peak position for the generation of two 3D spatial maps of 16× 16×16 voxels containing PGI-determined range shift and proton number information. Three complexity levels of simulated PGI data were investigated: (I) optimal PGI data, (II) realistic PGI data with simulated Poisson noise based on the locally delivered proton number, and (III) realistic PGI data with an additional positioning uncertainty of the slit camera following an experimentally determined distribution. For each complexity level, 3D-CNNs were trained on a data subset (n=9) using patient-wise leave-one-out cross-validation and tested on an independent test cohort (n=4). Both the binary task of detecting RE and the multi-class task of classifying the underlying error source were investigated. Similarly, four different conventional ML classifiers (logistic regression, multilayer perceptron, random forest, and support vector machine) were trained using five previously established handcrafted features extracted from the PGI data and used for performance comparison. On the test data, the CNN ensemble achieved a binary accuracy of 0.95, 0.96, and 0.93 and a multi-class accuracy of 0.83, 0.81, and 0.76 for the complexity levels (I), (II), and (III), respectively. In the case of binary classification, the CNN ensemble detected treatment deviations in the most realistic scenario with a sensitivity of 0.95 and a specificity of 0.88. The best performing ML classifiers showed a similar test performance. This study demonstrates that CNNs can reliably detect relevant changes in realistically simulated PGI data and classify most of the underlying sources of treatment deviations. The CNNs extracted meaningful features from the PGI data with a performance comparable to ML classifiers trained on previously established handcrafted features. These results highlight the potential of a reliable, automatic interpretation of PGI data for treatment verification, which is highly desired for a broad clinical application and a prerequisite for the inclusion of PGI in an automated feedback loop for online adaptive PT.

Multi-Omics Classifier of Tumor Recurrence vs. Radiation-Induced Lung Fibrosis in NSCLC Patients Treated with SBRT

<h3>Purpose/Objective(s)</h3> Differentiation between radiation-induced lung fibrosis (RILF) and tumor local recurrence (LR) remains challenging in SBRT treated NSCLC patients. We aimed to integrate spatially and time-resolved 4D radiomics with dosiomics biomarkers to develop a novel robust multi-omics classifier of LR vs RILF. <h3>Materials/Methods</h3> 210 NSCLC patients (101/48% T1-2N0M0, 109/52% T1-3N1/xM0/x) treated with SBRT (median dose 60 Gy/8 fractions) between 2009 and 2019 were identified. Image-based quantitative features were extracted from the SBRT planning and follow-up (FU) CTs at the time point of diagnosis of LR/RILF or a matched time point for CTs without LR/RILF (radiomics) and from dose distributions (dosiomics). The region of interest was the planning target volume plus a 10mm margin (PTV_+10mm). Time-dependent alterations of the radiomics features in FU CTs were integrated (delta-radiomics). To identify relevant features associated with RILF or LR, resampling (iterations = 1000) of feature selection methods were applied. An ensemble classifier comprising random forests, neural networks, and logistic regression was then trained. The area under the ROC curve (AUC) with a 70%/30% training/testing split and a 5-fold cross-validated AUC on the entire cohort were calculated for performance assessment at CT FU. <h3>Results</h3> Out of the 210, 36 patients (17%) were reported with LR and 44 (21%) with RILF. The median FU time was 19 months (range, 5-87). 20 LR and 23 RILF FU CTs, both with median time after SBRT of 15 months, were identified. 53 patients with matching clinical characteristics without LR/RILF (none), at matched FU time points (median time after SBRT: 17 months), were additionally selected, resulting in a cohort of 96 patients (20/21% LR, 23/24% RILF, 53/55% none). Discrimination of RILF versus LC using 4D radiomics features was achieved with a testing and 5-fold AUC of 0.82 [95%CI 0.79 0.86] and 0.85 [0.83 0.88]. The addition of a dosiomics feature improved performance to 0.85 [0.82 0.87] and 0.88 [0.85 0.91]. The PTV_+10mm-based classifier includes 4 textural features (2 delta-radiomics, 1 radiomics from the FU CT and 1 dosiomics). A non-significant correlation was found between the significant features and tumor volume before RT (Spearman's r <0.1, p<0.05). <h3>Conclusion</h3> Our data indicate that integrative omics by combining radiotherapy volume and dose constraints with spatially and time-resolved radiomics may provide a novel mean for better discrimination of tumor recurrence vs RILF after SBRT.

A Prospective Observational Study of Clinical Acceptability of Deep Learning Model for the Automated Segmentation of Organs at Risk for Head and Neck Radiotherapy Treatment Planning

<h3>Purpose/Objective(s)</h3> Accurate organ-at-risk (OAR) contouring is essential to improve plan quality. Head and neck (H&N) cancer has many OARs and contouring requires substantial time and effort. We developed a deep learning model (DLM) for autosegmentation of H&N OARs and performed a prospective study to assess time savings using DLM to generate contours for review compared to those created by medical dosimetrist's assistants (MDAs) with experience in H&N OAR contouring as part of routine practice. A prospective study was run to assess if the use of the DLM would reduce contouring time by more than 30%. <h3>Materials/Methods</h3> Using a custom 3D U-net architecture operating on 32 × 512 × 512-voxel subvolumes, the DLM was trained to segment 42 OARs for H&N cancer using retrospective planning CTs with OAR contours generated by H&N expert radiation oncologists (ROs) using sufficient protected time to represent the institutional "gold standard" (312 training, 51 model tuning, 82 testing), with 20 held out for prospective validation through an IRB-approved study with H&N specialized MDAs and ROs as consenting participants. The prospective study involved two arms for each hold out CT: the standard arm involved MDAs creating initial contours followed by revision by an RO (Manual arm) and initial contours for the same CT created by the DLM and revised by a different H&N RO (AI arm), the ROs were blinded to the origin of the initial contours and were randomly assigned an equal number of cases from each arm and asked to revise the contours to meet clinical standards for planning. The study was powered to detect greater than 30% reduction in contouring time (<b>α</b> = 0.025). Final RO-edited contours from each arm were compared using Dice Similarity Coefficient (DSC) to the gold standard (GS). Participant surveys were obtained after each session and the completion of the study. <h3>Results</h3> The mean total time for the manual arm was 3.4 hours (95% CI, 2.9-3.9 hours), while for the AI arm it was 0.7 hours (95% CI, 0.6-1.6 hours), a reduction of 76% (95% CI, 65%-88%) which was significantly greater than 30% (p<0.0001). 36 OARs (86%) had mean DSC greater than 0.8 on the AI arm compared to 20 (45%) for the Manual arm. The mean DSC for select OARs are given in table below. Survey of ROs showed satisfaction with DLM contours and interest in using for future cases. <h3>Conclusion</h3> The DLM demonstrated a significant reduction in the time needed for contouring the OARs for H&N cancer while improving agreement with the institutional gold standard. Integration of the DLM into the clinic for prospective evaluation is underway.

Automatic Cardiac Structure Contouring for Small Datasets with Cascaded Deep Learning Models

Cardiac structure contouring is a time consuming and tedious manual activity used for radiotherapeutic dose toxicity planning. We developed an automatic cardiac structure segmentation pipeline for use in low-dose non-contrast planning CT based on deep learning algorithms for small datasets. Fifty CT scans were retrospectively selected and the whole heart, ventricles and atria were contoured. A two stage deep learning pipeline was trained on 41 non contrast planning CTs, tuned with 3 CT scans and validated on 6 CT scans. In the first stage, An InceptionResNetV2 network was used to identify the slices that contained cardiac structures. The second stage consisted of three deep learning models trained on the images containing cardiac structures to segment the structures. The three deep learning models predicted the segmentations/contours on axial, coronal and sagittal images and are combined to create the final prediction. The final accuracy of the pipeline was quantified on 6 volumes by calculating the Dice similarity coefficient (DC), 95% Hausdorff distance (95% HD) and volume ratios between predicted and ground truth volumes. Median DC and 95% HD of 0.96, 0.88, 0.92, 0.80 and 0.82, and 1.86, 2.98, 2.02, 6.16 and 6.46 were achieved for the whole heart, right and left ventricle, and right and left atria respectively. The median differences in volume were -4, -1, + 5, -16 and -20% for the whole heart, right and left ventricle, and right and left atria respectively. The automatic contouring pipeline achieves good results for whole heart and ventricles. Robust automatic contouring with deep learning methods seems viable for local centers with small datasets.

Deep learning-based thoracic CBCT correction with histogram matching

Kilovoltage cone-beam computed tomography (CBCT)-based image-guided radiation therapy (IGRT) is used for daily delivery of radiation therapy, especially for stereotactic body radiation therapy (SBRT), which imposes particularly high demands for setup accuracy. The clinical applications of CBCTs are constrained, however, by poor soft tissue contrast, image artifacts, and instability of Hounsfield unit (HU) values. Here, we propose a new deep learning-based method to generate synthetic CTs (sCT) from thoracic CBCTs. A deep-learning model which integrates histogram matching (HM) into a cycle-consistent adversarial network (Cycle-GAN) framework, called HM-Cycle-GAN, was trained to learn mapping between thoracic CBCTs and paired planning CTs. Perceptual supervision was adopted to minimize blurring of tissue interfaces. An informative maximizing loss was calculated by feeding CBCT into the HM-Cycle-GAN to evaluate the image histogram matching between the planning CTs and the sCTs. The proposed algorithm was evaluated using data from 20 SBRT patients who each received 5 fractions and therefore 5 thoracic CBCTs. To reduce the effect of anatomy mismatch, original CBCT images were pre-processed via deformable image registrations with the planning CT before being used in model training and result assessment. We used planning CTs as ground truth for the derived sCTs from the correspondent co-registered CBCTs. The mean absolute error (MAE), peak signal-to-noise ratio (PSNR), and normalized cross-correlation (NCC) indices were adapted as evaluation metrics of the proposed algorithm. Assessments were done using Cycle-GAN as the benchmark. The average MAE, PSNR, and NCC of the sCTs generated by our method were 66.2 HU, 30.3 dB, and 0.95, respectively, over all CBCT fractions. Superior image quality and reduced noise and artifact severity were seen using the proposed method compared to the results from the standard Cycle-GAN method. Our method could therefore improve the accuracy of IGRT and corrected CBCTs could help improve online adaptive RT by offering better contouring accuracy and dose calculation.

A Dosimetric Analysis of Locoregional Failure Using Deformable Image Registration in Hypopharyngeal Cancer After Sequential-Boost Intensity-Modulated Radiotherapy

<h3>Purpose/Objective(s)</h3> Sequential boost (SQB) is an IMRT approach that is considered comparable to simultaneous-integrated boost. Dosimetric analysis in SQB using two planning CTs requires advanced dose-accumulation techniques, such as a deformable image registration (DIR) workflow. The purpose of this study was to assess patterns of failure and identify dosimetric features for hypopharyngeal cancer patients treated with SQB-IMRT. <h3>Materials/Methods</h3> Between 2013 and 2019, 102 hypopharyngeal cancer patients were treated with definitive SQB-IMRT. In short, both the PTV-high risk, including the primary tumor and metastatic lymph nodes, and the PTV-low risk, including elective nodal regions, received 46 Gy at 2 Gy/fraction using the first planning CT. The PTV-high risk sequentially received an additional 24 Gy using a second planning CT. The doses for the first and second plans were accumulated and registered onto the second planning CT using the DIR workflow. All locoregional recurrences were contoured on the follow-up CT and registered with the second planning CT. Recurrences were classified as in-field (more than 95% of the recurrence volume (Vrec) received 95% of the prescribed dose), marginal (20-95%), or out-of-field (less than 20%). The toxicity was also evaluated using CTCAE ver4.0. <h3>Results</h3> The median age was 66 years. Stage I, II, III and IV diseases were diagnosed in 3, 23, 20 and 56 patients, respectively. Eighty-three patients (81%) received concurrent chemoradiotherapy, and 19 patients (19%) received radiation therapy alone. The median total dose was 70 Gy. The median follow-up period was 22 months for all patients and 24 months for survivors. The 2-year overall survival rate and locoregional control rate were 79% and 57%, respectively. Thirty-four (33%) locoregional recurrences were observed. All locoregional failures were defined as in-field, and there was no marginal or out-of-field recurrence. The median mean dose and of the Vrec were 72.5 Gy (range, 71.6-74.1). In the patients who experienced locoregional failure, the median mean dose to the elective nodal region was assessed to be 60.1 (range, 55.4-68.4) Gy, and no recurrence occurred in the area. Late toxicities 6 months after RT were evaluated in 34 patients, and 24%, 3%, 25% and 2% of the patients experienced grade 2 and 3 dysphagia, grade 2 dry mouth and G2 dysgeusia, respectively. There were no grade 4/5 late toxicities. <h3>Conclusion</h3> In the detailed dosimetric analysis using DIR, all locoregional relapses after SQB-IMRT were in-field recurrences in the high-dose region. There was no recurrence in the prophylactic dose area, in which the accumulated dose reached as high as 60 Gy. More aggressive treatment for gross tumors and lower prophylactic doses may contribute to improved treatment outcomes and reduced side effects.

First Application of Predictive Model to Assist Adaptive Proton Therapy

<h3>Purpose/Objective(s)</h3> To assess a model to predict anatomical changes in head and neck cancer patients and explore the applicability of the model as a tool for adaptive replanning in intensity-modulated proton therapy (IMPT). <h3>Materials/Methods</h3> 20 radiotherapy patients with nasopharyngeal cancer were included in this retrospective study. Each patient had a planning CT and weekly CTs during radiotherapy. To build the anatomical model, we deform the weekly CTs to planning CTs of our training population (n = 19) and obtain the average anatomical change per week. To predict a deformation for the remaining patient, the average deformation of the training population is applied to the patient's planning CT. The model is updated based on the patient's progression during treatment. K-fold cross validation (n = 5) was used to obtain a sample of 5 patients. For those 5 patients, we compare the accumulated dose of two adaptive IMPT strategies. 1) Predictive replan (PR): Replans were optimized on predicted images of weeks 3 and 5 and applied to week 3/4 and week 5/6 respectively. 2) Best-case replan (BcR): Adaptive replan on weekly CT is triggered when target coverage D95% criteria are not met and parotid gland mean dose deviation > 3 Gy (RBE). All plans (original plan, PR, BcR) were robustly optimized (+/-3 mm setup and +/-3.5% range uncertainty). Dosimetric goals: CTV D95(%) > 95% prescription dose in robust evaluation; maximum dose to spinal cord and brainstem < 45 Gy (RBE) and < 55 Gy (RBE), respectively. To ensure fairness of comparison, we kept dosimetric parameters between BcR and PR consistent. <h3>Results</h3> Accumulated dose differences from two adaptive strategies (PR-BcR) of our 5 testing patients are shown (tab 1). We observe a maximum dose difference of 2.2% in the CTV D₉₅, -1.35 Gy (RBE) in the brainstem D_max, and -1.24 Gy (RBE) in the parotid D_mean. The parentheses indicate the dose difference of the replans only (no accumulation). <h3>Conclusion</h3> A predictive model was applied to the replanning process for adaptive proton therapy and compared to a best-case replan strategy. Our suggested PR produces clinically acceptable plans, comparable to the best-case replan strategy, suggesting that the predictive models can be used for adaptive replanning. Application of predictive replanning provides the possibility to prepare adaptive plans in advance, streamlining the clinical workflow.

A novel TPS toolkit to assess correlation between transit fluence dosimetry and DVH metrics for adaptive head and neck radiotherapy.

Inter-fractional anatomical variations in head and neck (H&N) cancer patients can lead to clinically significant dosimetric changes. Adaptive re-planning should thus commence to negate any potential over-dosage to organs-at-risk (OAR), as well as potential under-dosage to target lesions. The aim of this study is to explore the correlation between transit fluence, as measured at an electronic portal imaging device (EPID), and dose volume histogram (DVH) metrics to target and OAR structures in a simulated environment. Planning data of eight patients that have previously undergone adaptive radiotherapy for H&N cancer using volumetric modulated arc therapy (VMAT) at the Royal Adelaide Hospital were selected for this study. Through delivering the original treatment plan to both the planning and rescan CTs of these eight patients, predicted electronic portal images (EPIs) and DVH metrics corresponding to each data set were extracted using a novel RayStation script. A weighted projection mask was developed for target and OAR structures through considering the intra-angle overlap between fluence and structure contours projected onto the EPIs. The correlation between change in transit fluence and planning target volume (PTV) D98 and spinal cord D0.03cc with and without the weighting mask applied was investigated. PTV D98 was strongly correlated with mean fluence percentage difference both with and without the weighting mask applied (RMask = 0.69, RNo Mask = 0.79, N = 14, p < 0.05), where spinal cord D0.03cc exhibited a weak correlation (RMask = 0.35, RNo Mask = 0.53, N = 7, p > 0.05) however this result was not statistically significant. The simulation toolkit developed in this work provided a useful means to investigate the relationship between change in transit fluence and change in key dosimetric parameters for H&N cancer patients.

Surrogate-free machine learning-based organ dose reconstruction for pediatric abdominal radiotherapy

To study radiotherapy-related adverse effects, detailed dose information (3D distribution) is needed for accurate dose-effect modeling. For childhood cancer survivors who underwent radiotherapy in the pre-CT era, only 2D radiographs were acquired, thus 3D dose distributions must be reconstructed from limited information. State-of-the-art methods achieve this by using 3D surrogate anatomies. These can however lack personalization and lead to coarse reconstructions. We present and validate a surrogate-free dose reconstruction method based on Machine Learning (ML). Abdominal planning CTs (n = 142) of recently-treated childhood cancer patients were gathered, their organs at risk were segmented, and 300 artificial Wilms’ tumor plans were sampled automatically. Each artificial plan was automatically emulated on the 142 CTs, resulting in 42,600 3D dose distributions from which dose-volume metrics were derived. Anatomical features were extracted from digitally reconstructed radiographs simulated from the CTs to resemble historical radiographs. Further, patient and radiotherapy plan features typically available from historical treatment records were collected. An evolutionary ML algorithm was then used to link features to dose-volume metrics. Besides 5-fold cross validation, a further evaluation was done on an independent dataset of five CTs each associated with two clinical plans. Cross-validation resulted in mean absolute errors ≤ 0.6 Gy for organs completely inside or outside the field. For organs positioned at the edge of the field, mean absolute errors ≤ 1.7 Gy for , ≤ 2.9 Gy for , and ≤ 13% for and , were obtained, without systematic bias. Similar results were found for the independent dataset. To conclude, we proposed a novel organ dose reconstruction method that uses ML models to predict dose-volume metric values given patient and plan features. Our approach is not only accurate, but also efficient, as the setup of a surrogate is no longer needed.

A novel framework for spatial normalization of dose distributions in voxel-based analyses of brain irradiation outcomes

PurposeTo devise a novel Spatial Normalization framework for Voxel-based analysis (VBA) in brain radiotherapy. VBAs rely on accurate spatial normalization of different patients’ planning CTs on a common coordinate system (CCS). The cerebral anatomy, well characterized by MRI, shows instead poor contrast in CT, resulting in potential inaccuracies in VBAs based on CT alone. MethodsWe analyzed 50 meningioma patients treated with proton-therapy, undergoing planning CT and T1-weighted (T1w) MRI. The spatial normalization pipeline based on MR and CT images consisted in: intra-patient registration of CT to T1w, inter-patient registration of T1w to MNI space chosen as CCS, doses propagation to MNI.The registration quality was compared with that obtained by Statistical Parametric Mapping software (SPM), used as benchmark. To evaluate the accuracy of dose normalization, the dose organ overlap (DOO) score was computed on gray matter, white matter and cerebrospinal fluid before and after normalization. In addition, the trends in the DOOs distribution were investigated by means of cluster analysis. ResultsThe registration quality was higher for the proposed method compared to SPM (p < 0.001). The DOO scores showed a significant improvement after normalization (p < 0.001). The cluster analysis highlighted 2 clusters, with one of them including the majority of data and exhibiting acceptable DOOs. ConclusionsOur study presents a robust tool for spatial normalization, specifically tailored for brain dose VBAs. Furthermore, the cluster analysis provides a formal criterion for patient exclusion in case of non-acceptable normalization results. The implemented framework lays the groundwork for future reliable VBAs in brain irradiation studies.

Patch-based generative adversarial neural network models for head and neck MR-only planning.

To evaluate pix2pix and CycleGAN and to assess the effects of multiple combination strategies on accuracy for patch-based synthetic computed tomography (sCT) generation for magnetic resonance (MR)-only treatment planning in head and neck (HN) cancer patients. Twenty-three deformably registered pairs of CT and mDixon FFE MR datasets from HN cancer patients treated at our institution were retrospectively analyzed to evaluate patch-based sCT accuracy via the pix2pix and CycleGAN models. To test effects of overlapping sCT patches on estimations, we (a) trained the models for three orthogonal views to observe the effects of spatial context, (b) we increased effective set size by using per-epoch data augmentation, and (c) we evaluated the performance of three different approaches for combining overlapping Hounsfield unit (HU) estimations for varied patch overlap parameters. Twelve of twenty-three cases corresponded to a curated dataset previously used for atlas-based sCT generation and were used for training with leave-two-out cross-validation. Eight cases were used for independent testing and included previously unseen image features such as fused vertebrae, a small protruding bone, and tumors large enough to deform normal body contours. We analyzed the impact of MR image preprocessing including histogram standardization and intensity clipping on sCT generation accuracy. Effects of mDixon contrast (in-phase vs water) differences were tested with three additional cases. The sCT generation accuracy was evaluated using mean absolute error (MAE) and mean error (ME) in HU between the plan CT and sCT images. Dosimetric accuracy was evaluated for all clinically relevant structures in the independent testing set and digitally reconstructed radiographs (DRRs) were evaluated with respect to the plan CT images. The cross-validated MAEs for the whole-HN region using pix2pix and CycleGAN were 66.9±7.3 vs 82.3±6.4HU, respectively. On the independent testing set with additional artifacts and previously unseen image features, whole-HN region MAEs were 94.0±10.6 and 102.9±14.7HU for pix2pix and CycleGAN, respectively. For patients with different tissue contrast (water mDixon MR images), the MAEs increased to 122.1±6.3 and 132.8±5.5HU for pix2pix and CycleGAN, respectively. Our results suggest that combining overlapping sCT estimations at each voxel reduced both MAE and ME compared to single-view non-overlapping patch results. Absolute percent mean/max dose errors were 2% or less for the PTV and all clinically relevant structures in our independent testing set, including structures with image artifacts. Quantitative DRR comparison between planning CTs and sCTs showed agreement of bony region positions to <1mm. The dosimetric and MAE based accuracy, along with the similarity between DRRs from sCTs, indicate that pix2pix and CycleGAN are promising methods for MR-only treatment planning for HN cancer. Our methods investigated for overlapping patch-based HU estimations also indicate that combining transformation estimations of overlapping patches is a potential method to reduce generation errors while also providing a tool to potentially estimate the MR to CT aleatoric model transformation uncertainty. However, because of small patient sample sizes, further studies are required.

The GIRAFE phase II trial on MVCT-based “volumes of the day” and “dose of the day” addresses when and how to implement adaptive radiotherapy for locally advanced head and neck cancer

During exclusive curative radiotherapy for head and neck tumors, the patient’s organs at risk (OAR) and target volumes frequently change size and shape, leading to a risk of higher toxicity and lower control than expected on planned dosimetry. Adaptive radiotherapy is often necessary but 1) tools are needed to define the optimal time for replanning, and 2) the subsequent workflow is time-consuming.We designed a prospective study to evaluate 1) the validity of automatically deformed contours on the daily MVCT, in order to safely use the “dose-of the day” tool to check daily if replanning is necessary; 2) the automatically deformed contours on the replanning CT and the time gained in the replanning workflow.Forty-eight patients with T3-T4 and/or involved node >2 cm head and neck squamous cell carcinomas, planned for curative radiotherapy without surgery, will be enrolled. They will undergo treatment with helical IMRT including daily repositioning MVCTs. The contours proposed will be compared weekly on intermediate planning CTs (iCTs) on weeks 3, 4, 5 and 6. On these iCTs both manual recontouring and automated deformable registration of the initial contours will be compared with the contours automatically defined on the MVCT. The primary objective is to evaluate the Dice similarity coefficient (DSC) of the volumes of each parotid gland. The secondary objectives will evaluate, for target volumes and all OARs: the DSC, the mean distance to agreement, and the average surface-to-surface distance. Time between the automatic and the manual recontouring workflows will be compared.

CBCT-guided evolutive library for cervical adaptive IMRT.

In the context of adaptive radiation therapy (ART) for locally advanced cervical carcinoma (LACC), this study proposed an original cone-beam computed tomography (CBCT)-guided "Evolutive library" and evaluated it against four other known radiotherapy (RT) strategies. For 20 patients who underwent intensity-modulated radiation therapy (IMRT) for LACC, three planning CTs [with empty (EB), intermediate (IB), and full (FB) bladder volumes], a CT scan at 20 Gy and bi-weekly CBCTs for 5 weeks were performed. Five RT strategies were simulated for each patient: "Standard RT" was based on one IB planning CT; "internal target volume (ITV)-based RT" was an ITV built from the three planning CTs; "RT with one mid-treatment replanning (MidTtReplan)" corresponded to the standard RT with a replanning at 20 Gy; "Pretreatment library ART" using a planning library based on the three planning CTs; and the "Evolutive library ART", which was the "Pretreatment library ART" strategy enriched by including some CBCT anatomies into the library when the daily clinical target volume (CTV) shape differed from the ones in the library. Two planning target volume (PTV) margins of 7 and 10 mm were evaluated. All the strategies were geometrically compared in terms of the percentage of coverage by the PTV, for the CTV and the organs at risk (OAR) delineated on the CBCT. Inadequate coverage of the CTV and OARs by the PTV was also assessed using deformable image registration. The cumulated dose distributions of each strategy were likewise estimated and compared for one patient. The "Evolutive library ART" strategy involved a number of added CBCTs: 0 for 55%; 1 for 30%; 2 for 5%; and 3 for 10% of patients. Compared with the other four, this strategy provided the highest CTV geometric coverage by the PTV, with a mean (min-max) coverage of 98.5% (96.4-100) for 10 mm margins and 96.2% (93.0-99.7) for 7 mm margins (P < 0.05). Moreover, this strategy significantly decreased the geometric coverage of the bowel. CTV undercoverage by PTV occurred in the anterior and superior uterine regions for all strategies. The dosimetric analysis at 7 mm similarly demonstrated that the "Evolutive library ART" increased the V42.75Gy of the CTV by 27%, 20%, 13%, and 28% compared with "Standard RT", "ITV-based RT", "MidTtReplan", and "Pretreatment library ART", respectively. The dose to the bowel was also decreased by the "Evolutive library ART" compared with that by the other strategies. The "Evolutive library ART" is a personalized ART strategy that comprises a pretreatment planning library of three CT scans, enriched for half of the patients by one to three per-treatment CBCTs. This original strategy increased both the CTV coverage and bowel sparing compared with all the other tested strategies and enables us to consider a PTV margin reduction.

Image-guided radiotherapy for muscle invasive bladder cancer with intravesical lipiodol injection. A new option for bladder sparing treatment

To implement lipiodol as a fiducial marker of the tumor bed for image-guided radiotherapy with simultaneous integrated boost technique as part of radiochemotherapy for muscle invasive bladder tumors. Since April 2016, radiochemotherapy was performed in 3 male patients with muscle invasive, transitional cell bladder carcinoma. Prior to radiochemotherapy, tumor bed resection was performed for each patient, at the same time 10 ml of lipiodol solution was injected submucosally into the resection site, thus marking the tumor bed for escalated dose irradiation. During radiochemotherapy 51 Gy (1.7 Gy/die) to the pelvis, 57 Gy (1.9 Gy/die) to the whole bladder, and 63 Gy (2.1 Gy/die) to the lipiodol-labeled tumor bed was delivered with simultaneous integrated boost technique. The accuracy of the irradiation was controlled by daily kilovoltage CT. Early radiogenic urogenital and gastrointestinal side effects were recorded according to Radiation Therapy Oncology Group side-effects grading recommendation. Substantial perioperative side effect or toxicity were not observed during and after the injection of lipiodol. The prescribed dose was successfully delivered in all patients. Radiotherapy duration was 6 weeks. The lipiodol-labeled tumor bed was clearly visible on daily kilovoltage cone beam CT. In one patient grade II cystitis and proctitis was observed, another patient experienced only grade I cystitis. These complaints improved with symptomatic medication. In the third patient no significant side effect occurred. The injection of lipiodol into the bladder wall is a safe technique, without any perioperative toxicity or complication. The tumor bed demarcated by lipiodol was visible both on treatment planning and kilovoltage CTs. The total treatment time was shortened by 4 days. The treatment was well tolerated, early side effects were moderate, or slight. Orv Hetil. 2017; 158(51): 2041-2047.

Prospective analysis of in vivo landmark point-based MRI geometric distortion in head and neck cancer patients scanned in immobilized radiation treatment position: Results of a prospective quality assurance protocol

PurposeUncertainties related to geometric distortion are a major obstacle for effectively utilizing MRI in radiation oncology. We aim to quantify the geometric distortion in patient images by comparing their in-treatment position MRIs with the corresponding planning CTs, using CT as the non-distorted gold standard. MethodsTwenty-one head and neck cancer patients were imaged with MRI as part of a prospective Institutional Review Board approved study. MR images were acquired with a T2 SE sequence (0.5×0.5×2.5mm voxel size) in the same immobilization position as in the CTs. MRI to CT rigid registration was then done and geometric distortion comparison was assessed by measuring the corresponding anatomical landmarks on both the MRI and the CT images. Several landmark measurements were obtained including; skin to skin (STS), bone to bone, and soft tissue to soft tissue at specific levels in horizontal and vertical planes of both scans. Inter-observer variability was assessed and interclass correlation (ICC) was calculated. ResultsA total of 430 landmark measurements were obtained. The median distortion for all landmarks in all scans was 1.06mm (IQR 0.6–1.98). For each patient 48% of the measurements were done in the right-left direction and 52% were done in the anteroposterior direction. The measured geometric distortion was not statistically different in the right-left direction compared to the anteroposterior direction (1.5±1.6 vs. 1.6±1.7mm, respectively, p=0.4). The magnitude of distortion was higher in the STS peripheral landmarks compared to the more central landmarks (2.0±1.9 vs. 1.2±1.3mm, p<0.0001). The mean distortion measured by observer one was not significantly different compared to observer 2, 3, and 4 (1.05, 1.23, 1.06 and 1.05mm, respectively, p=0.4) with ICC=0.84. ConclusionMRI geometric distortions were quantified in radiotherapy planning applications with a clinically insignificant error of less than 2mm compared to the gold standard CT.