Search

Objective: To analyse how often innovations in healthcare are evaluated regarding output, especially in radiotherapy. Output was defined as either survival, toxicity, safety, service, efficiency or cost-effectiveness., Methods: A systematic literature review was conducted, using three search strategies: (1) innovations in general healthcare; (2) radiotherapy-specific innovations, i.e. organizational innovations and general implementation of innovations; (3) innovations per tumour group/radiotherapy technique. Scientific levels were classified according to the system used in European Society for Medical Oncology guidelines. Finally, we calculated the percentage of implemented innovations in Dutch radiotherapy centres for which we found evidence regarding output in the literature review., Results: Only 94/1072 unique articles matched the inclusion criteria. Significant results on patient outcome, service or safety were reported in 65% of papers, which rose to 76% if confined to radiotherapy reviews. A significant technological improvement was identified in 26%, cost-effectiveness in 10% and costs/efficiency in 36% of the papers. The scientific level of organizational innovations was lower than that of clinical papers. Dutch radiotherapy treatment innovations were adequately evaluated on outcome data before implementation in clinical routine in a minimum of 64-92% of cases., Conclusion: Only few studies report on output when considering innovations in general, but radiotherapy reviews give a reasonably good insight into innovation output effects, with a higher level of evidence. In Dutch radiotherapy centres only small improvements are possible regarding evaluation of treatment innovations before implementation. Advances in knowledge: This study is the first of its kind measuring how innovations are evaluated in scientific literature, before implementation in clinical practice.

Objective: To study the implementation of innovation activities in Dutch radiotherapy (RT) centres in a broad sense (product, technological, market and organizational innovations)., Methods: A descriptive cross-sectional study was conducted in 15 Dutch RT centres. A list of innovations implemented from 2011 to 2013 was drawn up for each centre using semi-structured interviews. These innovations were classified into innovation categories according to previously defined innovation indicators. Where applicable, each innovation was rated by each centre on the effort required to implement it and on its expected effects, to get an impression of how far reaching and radical the innovations were and to be able to compare the number of innovations between centres., Results: The participating RT centres in the Netherlands implemented 12 innovations per year on average (range 5-25); this number was not significantly different for academic (n = 13) or non-academic centres (n = 10). Several centres were dealing with the same innovations at the same time. The average required effort and expected output did not differ significantly between product, technological and organizational innovation or between academic and non-academic centres., Conclusion: The number of innovations observed per centre varied across a large range, with a large overlap in terms of the type of innovations that were implemented. Registering innovations using the innovation indicators applied in our study would make it possible to improve collaboration between centres, e.g. with common training modules, to avoid duplication of work. Advances in knowledge: This study is the first of its kind investigating innovation implementation in RT in a broad sense.

PurposeMachine learning classification algorithms (classifiers) for prediction of treatment response are becoming more popular in radiotherapy literature. General Machine learning literature provides evidence in favor of some classifier families (random forest, support vector machine, gradient boosting) in terms of classification performance. The purpose of this study is to compare such classifiers specifically for (chemo)radiotherapy datasets and to estimate their average discriminative performance for radiation treatment outcome prediction.MethodsWe collected 12 datasets (3496 patients) from prior studies on post-(chemo)radiotherapy toxicity, survival, or tumor control with clinical, dosimetric, or blood biomarker features from multiple institutions and for different tumor sites, that is, (non-)small-cell lung cancer, head and neck cancer, and meningioma. Six common classification algorithms with built-in feature selection (decision tree, random forest, neural network, support vector machine, elastic net logistic regression, LogitBoost) were applied on each dataset using the popular open-source R package caret. The R code and documentation for the analysis are available online (https://github.com/timodeist/classifier_selection_code). All classifiers were run on each dataset in a 100-repeated nested fivefold cross-validation with hyperparameter tuning. Performance metrics (AUC, calibration slope and intercept, accuracy, Cohen's kappa, and Brier score) were computed. We ranked classifiers by AUC to determine which classifier is likely to also perform well in future studies. We simulated the benefit for potential investigators to select a certain classifier for a new dataset based on our study (pre-selection based on other datasets) or estimating the best classifier for a dataset (set-specific selection based on information from the new dataset) compared with uninformed classifier selection (random selection).ResultsRandom forest (best in 6/12 datasets) and elastic net logistic regression (best in 4/12 datasets) showed the overall best discrimination, but there was no single best classifier across datasets. Both classifiers had a median AUC rank of 2. Preselection and set-specific selection yielded a significant average AUC improvement of 0.02 and 0.02 over random selection with an average AUC rank improvement of 0.42 and 0.66, respectively.ConclusionRandom forest and elastic net logistic regression yield higher discriminative performance in (chemo)radiotherapy outcome and toxicity prediction than other studied classifiers. Thus, one of these two classifiers should be the first choice for investigators when building classification models or to benchmark one's own modeling results against. Our results also show that an informed preselection of classifiers based on existing datasets can improve discrimination over random selection.

Purpose: Artificial intelligence applications in radiation oncology have been the focus of study in the last decade. The introduction of automated and intelligent solutions for routine clinical tasks, such as treatment planning and quality assurance, has the potential to increase safety and efficiency of radiotherapy. In this work, we present a multi-institutional study across three different institutions internationally on a Bayesian network (BN)-based initial plan review assistive tool that alerts radiotherapy professionals for potential erroneous or suboptimal treatment plans. Methods: Clinical data were collected from the oncology information systems in three institutes in Europe (Maastro clinic - 8753 patients treated between 2012 and 2020) and the United States of America (University of Vermont Medical Center [UVMMC] - 2733 patients, University of Washington [UW] - 6180 patients, treated between 2018 and 2021). We trained the BN model to detect potential errors in radiotherapy treatment plans using different combinations of institutional data and performed single-site and cross-site validation with simulated plans with embedded errors. The simulated errors consisted of three different categories: i) patient setup, ii) treatment planning and iii) prescription. We also compared the strategy of using only diagnostic parameters or all variables as evidence for the BN. We evaluated the model performance utilizing the area under the receiver-operating characteristic curve (AUC). Results: The best network performance was observed when the BN model is trained and validated using the dataset in the same center. In particular, the testing and validation using UVMMC data has achieved an AUC of 0.92 with all parameters used as evidence. In cross-validation studies, we observed that the BN model performed better when it was trained and validated in institutes with similar technology and treatment protocols (for instance, when testing on UVMMC data, the model trained on UW data achieved an AUC of 0.84, compared with an AUC of 0.64 for the model trained on Maastro data). Also, combining training data from larger clinics (UW and Maastro clinic) and using it on smaller clinics (UVMMC) leads to satisfactory performance with an AUC of 0.85. Lastly, we found that in general the BN model performed better when all variables are considered as evidence. Conclusion: We have developed and validated a Bayesian network model to assist initial treatment plan review using multi-institutional data with different technology and clinical practices. The model has shown good performance even when trained on data from clinics with divergent profiles, suggesting that the model is able to adapt to different data distributions. [ABSTRACT FROM AUTHOR]

Simple Summary: The developed decision tree uses patient status, the Charlson comorbidity index, the Eastern Cooperative Oncology Group Performance scale, age, sex, cancer type, health insurance status, radiotherapy aim, and fractionation type to distinguish between compliant and noncompliant patients. The developed tree's ability to identify those patients who are likely to discontinue their radiotherapy treatment is reasonably good, providing caregivers with a rationale for deciding whether to start radiotherapy treatment or look for alternative treatment for these patients. Additionally, the developed decision tree can help to boost treatment compliance by targeting those patients who are likely to discontinue therapy with incentives and techniques to help them adhere to treatment, especially for patients already receiving therapy. This study aims to analyze the relationship between the available variables and treatment compliance in elderly cancer patients treated with radiotherapy and to establish a decision tree model to guide caregivers in their decision-making process. For this purpose, 456 patients over 74 years of age who received radiotherapy between 2005 and 2017 were included in this retrospective analysis. The outcome of interest was radiotherapy compliance, determined by whether patients completed their scheduled radiotherapy treatment (compliance means they completed their treatment and noncompliance means they did not). A bootstrap (B = 400) technique was implemented to select the best tuning parameters to establish the decision tree. The developed decision tree uses patient status, the Charlson comorbidity index, the Eastern Cooperative Oncology Group Performance scale, age, sex, cancer type, health insurance status, radiotherapy aim, and fractionation type (conventional fractionation versus hypofractionation) to distinguish between compliant and noncompliant patients. The decision tree's mean area under the curve and 95% confidence interval was 0.71 (0.66–0.77). Although external validation is needed to determine the decision tree's clinical usefulness, its discriminating ability was moderate and it could serve as an aid for caregivers to select the optimal treatment for elderly cancer patients. [ABSTRACT FROM AUTHOR]

• Introduction of the FAIR data principles in radiotherapy. • Artificial Intelligence applications begin to manifest themselves in radiotherapy. • Examples of FAIR data in radiotherapy. • Stakeholders action items for the implementation of FAIR principles. Given the rapid growth of artificial intelligence (AI) applications in radiotherapy and the related transformations toward the data-driven healthcare domain, this article summarizes the need and usage of the FAIR (Findable, Accessible, Interoperable, Reusable) data principles in radiotherapy. This work introduces the FAIR data concept, presents practical and relevant use cases and the future role of the different parties involved. The goal of this article is to provide guidance and potential applications of FAIR to various radiotherapy stakeholders, focusing on the central role of medical physicists. [ABSTRACT FROM AUTHOR]

Atlas-based automatic segmentation is used in radiotherapy planning to accelerate the delineation of organs at risk (OARs). Atlas selection has been proposed as a way to improve the accuracy and execution time of segmentation, assuming that, the more similar the atlas is to the patient, the better the results will be. This paper presents an analysis of atlas selection methods in the context of radiotherapy treatment planning. For a range of commonly contoured OARs, a thorough comparison of a large class of typical atlas selection methods has been performed. For this evaluation, clinically contoured CT images of the head and neck (${N}={316}$) and thorax (${N}={280}$) were used. The state-of-the-art intensity and deformation similarity-based atlas selection methods were found to compare poorly to perfect atlas selection. Counter-intuitively, atlas selection methods based on a fixed set of representative atlases outperformed atlas selection methods based on the patient image. This study suggests that atlas-based segmentation with currently available selection methods compares poorly to the potential best performance, hampering the clinical utility of atlas-based segmentation. Effective atlas selection remains an open challenge in atlas-based segmentation for radiotherapy planning. [ABSTRACT FROM AUTHOR]

Purpose: To investigate the impact of radiation treatment quality assurance (RTQA) on treatment outcomes in a phase III trial for advanced head and neck cancer. Materials and Methods: A total of 767 patients from NRG/RTOG 0522 were included in this study. The contours of target volume (TV) and organ at risk (OAR), and dose-volume coverage of targets were reviewed and scored (per-protocol, variation-acceptable and deviation-unacceptable) according to the protocol. We performed log-rank tests for RTQA scores with patients' outcomes, including local control (LC), distant control (DC) and overall survival (OS). Cox models with and without RTQA score data were established. To obtain a more reasonable model, per-protocol and variation acceptable were combined into a single acceptable score. Results: The log-rank test showed that all RTQA scores correlated with LC, which was significantly different between the per-protocol and variation-acceptable patients in target and OAR contouring (p -value = 0.004 and 0.043). For dose-volume score, the per-protocol and variation-acceptable patients were significantly different from unacceptable patients in the LC, with a p -value = 0.020 and 0.006, respectively. The DC of patients with variation-acceptable was significantly different than that of the unacceptable patients (p -value = 0.043). There were no correlations between RTQA scores with other outcomes. By incorporating RTQA scores into outcome modeling, the performance of LC model can be improved from 0.62 to 0.63 (c-index). The RTQA scores had no impact on DC and OS. Conclusion: RTQA scores are related to patients' local control rates in head and neck cancer radiotherapy. [ABSTRACT FROM AUTHOR]

Atlas-based segmentation is used in radiotherapy planning to accelerate the delineation of organs at risk (OARs). Atlas selection has been proposed to improve the performance of segmentation, assuming that the more similar the atlas is to the patient, the better the result. It follows that the larger the database of atlases from which to select, the better the results should be. This paper seeks to estimate a clinically achievable expected performance under this assumption. Assuming a perfect atlas selection, an extreme value theory has been applied to estimate the accuracy of single-atlas and multi-atlas segmentation given a large database of atlases. For this purpose, clinical contours of most common OARs on computed tomography of the head and neck (${N}=\textsf {316}$) and thoracic (${N}=\textsf {280}$) cases were used. This paper found that while for most organs, perfect segmentation cannot be reasonably expected, auto-contouring performance of a level corresponding to clinical quality could be consistently expected given a database of 5000 atlases under the assumption of perfect atlas selection. [ABSTRACT FROM AUTHOR]

Purpose: Personalized medicine is expected to yield improved health outcomes. Data mining over massive volumes of patients' clinical data is an appealing, low‐cost and noninvasive approach toward personalization. Machine learning algorithms could be trained over clinical "big data" to build prediction models for personalized therapy. To reach this goal, a scalable "big data" architecture for the medical domain becomes essential, based on data standardization to transform clinical data into FAIR (Findable, Accessible, Interoperable and Reusable) data. Using Ontologies and Semantic Web technologies, we attempt to reach mentioned goal. Methods: We developed an ontology to be used in the field of radiation oncology to map clinical data from relational databases. We combined ontology with semantic Web techniques to publish mapped data and easily query them using SPARQL. Results: The Radiation Oncology Ontology (ROO) contains 1,183 classes and 211 properties between classes to represent clinical data (and their relationships) in the radiation oncology domain following FAIR principles. We combined the ontology with Semantic Web technologies showing how to efficiently and easily integrate and query data from different (relational database) sources without a priori knowledge of their structures. Discussion: When clinical FAIR data sources are combined (linked data) using mentioned technologies, new relationships between entities are created and discovered, representing a dynamic body of knowledge that is continuously accessible and increasing. [ABSTRACT FROM AUTHOR]

With the era of big data, the utilization of machine learning algorithms in radiation oncology is rapidly growing with applications including: treatment response modeling, treatment planning, contouring, organ segmentation, image‐guidance, motion tracking, quality assurance, and more. Despite this interest, practical clinical implementation of machine learning as part of the day‐to‐day clinical operations is still lagging. The aim of this white paper is to further promote progress in this new field of machine learning in radiation oncology by highlighting its untapped advantages and potentials for clinical advancement, while also presenting current challenges and open questions for future research. The targeted audience of this paper includes newcomers as well as practitioners in the field of medical physics/radiation oncology. The paper also provides general recommendations to avoid common pitfalls when applying these powerful data analytic tools to medical physics and radiation oncology problems and suggests some guidelines for transparent and informative reporting of machine learning results. [ABSTRACT FROM AUTHOR]

We present a methodology which can be utilized to select proton or photon radiotherapy in prostate cancer patients. Four state-of-the-art competing treatment modalities were compared (by way of an in silico trial) for a cohort of 25 prostate cancer patients, with and without correction strategies for prostate displacements. Metrics measured from clinical image guidance systems were used. Three correction strategies were investigated; no-correction, extended-no-action-limit, and online-correction. Clinical efficacy was estimated via radiobiological models incorporating robustness (how probable a given treatment plan was delivered) and stability (the consistency between the probable best and worst delivered treatments at the 95% confidence limit). The results obtained at the cohort level enabled the determination of a threshold for likely clinical benefit at the individual level. Depending on the imaging system and correction strategy; 24%, 32% and 44% of patients were identified as suitable candidates for proton therapy. For the constraints of this study: Intensity-modulated proton therapy with online-correction was on average the most effective modality. Irrespective of the imaging system, each treatment modality is similar in terms of robustness, with and without the correction strategies. Conversely, there is substantial variation in stability between the treatment modalities, which is greatly reduced by correction strategies. This study provides a 'proof-of-concept' methodology to enable the prospective identification of individual patients that will most likely (above a certain threshold) benefit from proton therapy. [ABSTRACT FROM AUTHOR]

Objective: To study the efficiency of research implementation in a large radiotherapy institute, in either an internal review board-approved clinical trial or clinical routine. Methods: Scientific publications of the institute were listed. We asked clinicians from tumour expert groups whether the study had been implemented yet in a clinical trial or in clinical practice and which facilitators or barriers were relevant. An independent investigator verified all results. We calculated the implementation rates and the frequency of mentioned facilitators and barriers. Results: Resident researchers had published 234 studies over the past 4 years. Overall, 70/234 (30%) technical or preclinical studies were tested or implemented in a clinical environment in either trials or routine. In total, 45/234 (19%) studies were routinely implemented; in the 61 clinical studies, this percentage was higher: 38% (23/61). The main facilitator was the level of evidence and the main barriers were workload and high complexity. Conclusion: We were able to calculate the implementation ratio of published research into clinical practice and set benchmark figures for other radiotherapy clinics. Level of evidence was an important facilitator, while workload and high complexity of the new procedures were important barriers for implementation. Recent articles suggest that academic entrepreneurship will facilitate this process further. [ABSTRACT FROM AUTHOR]

'Radiogenomics' is the study of genetic variation associated with response to radiotherapy. Radiogenomics aims to uncover the genes and biologic pathways responsible for radiotherapy toxicity that could be targeted with radioprotective agents and; identify genetic markers that can be used in risk prediction models in the clinic. The long-term goal of the field is to develop single nucleotide polymorphism-based risk models that can be used to stratify patients to more precisely tailored radiotherapy protocols. The field has evolved over the last two decades in parallel with advances in genomics, moving from narrowly focused candidate gene studies to large, collaborative genome-wide association studies. Several confirmed genetic variants have been identified and the field is making progress toward clinical translation. [ABSTRACT FROM AUTHOR]

Background and purpose A rapid learning approach has been proposed to extract and apply knowledge from routine care data rather than solely relying on clinical trial evidence. To validate this in practice we deployed a previously developed decision support system (DSS) in a typical, busy clinic for non-small cell lung cancer (NSCLC) patients. Material and methods Gender, age, performance status, lung function, lymph node status, tumor volume and survival were extracted without review from clinical data sources for lung cancer patients. With these data the DSS was tested to predict overall survival. Results 3919 lung cancer patients were identified with 159 eligible for inclusion, due to ineligible histology or stage, non-radical dose, missing tumor volume or survival. The DSS successfully identified a good prognosis group and a medium/poor prognosis group (2 year OS 69% vs. 27/30%, p < 0.001). Stage was less discriminatory (2 year OS 47% for stage I–II vs. 36% for stage IIIA–IIIB, p = 0.12) with most good prognosis patients having higher stage disease. The DSS predicted a large absolute overall survival benefit (∼40%) for a radical dose compared to a non-radical dose in patients with a good prognosis, while no survival benefit of radical radiotherapy was predicted for patients with a poor prognosis. Conclusions A rapid learning environment is possible with the quality of clinical data sufficient to validate a DSS. It uses patient and tumor features to identify prognostic groups in whom therapy can be individualized based on predicted outcomes. Especially the survival benefit of a radical versus non-radical dose predicted by the DSS for various prognostic groups has clinical relevance, but needs to be prospectively validated. [ABSTRACT FROM AUTHOR]

• A CT-derived radiomics model can predict OS in patients with a glioblastoma. • Discrimination based on the combined clinical and radiomics model was comparable to previous MRI-based models. • Qualitatively high-level datasets will support further model development. Glioblastoma (GBM) is the most common malignant primary brain tumour which has, despite extensive treatment, a median overall survival of 15 months. Radiomics is the high-throughput extraction of large amounts of image features from radiographic images, which allows capturing the tumour phenotype in 3D and in a non-invasive way. In this study we assess the prognostic value of CT radiomics for overall survival in patients with a GBM. Clinical data and pre-treatment CT images were obtained from 218 patients diagnosed with a GBM via biopsy who underwent radiotherapy +/− temozolomide between 2004 and 2015 treated at three independent institutes (n = 93, 62 and 63). A clinical prognostic score (CPS), a simple radiomics model consisting of volume based score (VPS), a complex radiomics prognostic score (RPS) and a combined clinical and radiomics (C + R)PS model were developed. The population was divided into three risk groups for each prognostic score and respective Kaplan–Meier curves were generated. Patient characteristics were broadly comparable. Clinically significant differences were observed with regards to radiation dose, tumour volume and performance status between datasets. Image acquisition parameters differed between institutes. The cross-validated c-indices were moderately discriminative and for the CPS ranged from 0.63 to 0.65; the VPS c-indices ranged between 0.52 and 0.61; the RPS c-indices ranged from 0.57 to 0.64 and the combined clinical and radiomics model resulted in c-indices of 0.59–0.71. In this study clinical and CT radiomics features were used to predict OS in GBM. Discrimination between low-, middle- and high-risk patients based on the combined clinical and radiomics model was comparable to previous MRI-based models. [ABSTRACT FROM AUTHOR]

• In radiation oncology, big multisource data and metadata, often presenting interoperability issues, are routinely produced in the clinic. • [R15] "Artificial intelligence can offer powerful methods for analyzing large amounts of data." • To reach the promise of a learning health care system, we need a data science community in radiation oncology. • We define the basis for setting up this community by proposing a list of milestones to support our vision. Big data are no longer an obstacle; now, by using artificial intelligence (AI), previously undiscovered knowledge can be found in massive data collections. The radiation oncology clinic daily produces a large amount of multisource data and metadata during its routine clinical and research activities. These data involve multiple stakeholders and users. Because of a lack of interoperability, most of these data remain unused, and powerful insights that could improve patient care are lost. Changing the paradigm by introducing powerful AI analytics and a common vision for empowering big data in radiation oncology is imperative. However, this can only be achieved by creating a clinical data science community in radiation oncology. In this work, we present why such a community is needed to translate multisource data into clinical decision aids. [ABSTRACT FROM AUTHOR]

Background and purpose The gross tumour volume (GTV) is predictive of clinical outcome and consequently features in many machine-learned models. 4D-planning, however, has prompted substitution of the GTV with the internal gross target volume (iGTV). We present and validate a method to synthesise GTV data from the iGTV, allowing the combination of 3D and 4D planned patient cohorts for modelling. Material and methods Expert delineations in 40 non-small cell lung cancer patients were used to develop linear fit and erosion methods to synthesise the GTV volume and shape. Quality was assessed using Dice Similarity Coefficients (DSC) and closest point measurements; by calculating dosimetric features; and by assessing the quality of random forest models built on patient populations with and without synthetic GTVs. Results Volume estimates were within the magnitudes of inter-observer delineation variability. Shape comparisons produced mean DSCs of 0.8817 and 0.8584 for upper and lower lobe cases, respectively. A model trained on combined true and synthetic data performed significantly better than models trained on GTV alone, or combined GTV and iGTV data. Conclusions Accurate synthesis of GTV size from the iGTV permits the combination of lung cancer patient cohorts, facilitating machine learning applications in thoracic radiotherapy. [ABSTRACT FROM AUTHOR]

Background and purpose Contouring of organs at risk (OARs) is an important but time consuming part of radiotherapy treatment planning. The aim of this study was to investigate whether using institutional created software-generated contouring will save time if used as a starting point for manual OAR contouring for lung cancer patients. Material and methods Twenty CT scans of stage I–III NSCLC patients were used to compare user adjusted contours after an atlas-based and deep learning contour, against manual delineation. The lungs, esophagus, spinal cord, heart and mediastinum were contoured for this study. The time to perform the manual tasks was recorded. Results With a median time of 20 min for manual contouring, the total median time saved was 7.8 min when using atlas-based contouring and 10 min for deep learning contouring. Both atlas based and deep learning adjustment times were significantly lower than manual contouring time for all OARs except for the left lung and esophagus of the atlas based contouring. Conclusions User adjustment of software generated contours is a viable strategy to reduce contouring time of OARs for lung radiotherapy while conforming to local clinical standards. In addition, deep learning contouring shows promising results compared to existing solutions. [ABSTRACT FROM AUTHOR]

Purpose: Tools for survival prediction for non-small cell lung cancer (NSCLC) patients treated with chemoradiation or radiation therapy are of limited quality. In this work, we developed a predictive model of survival at 2 years. The model is based on a large volume of historical patient data and serves as a proof of concept to demonstrate the distributed learning approach.Methods and Materials: Clinical data from 698 lung cancer patients, treated with curative intent with chemoradiation or radiation therapy alone, were collected and stored at 2 different cancer institutes (559 patients at Maastro clinic (Netherlands) and 139 at Michigan university [United States]). The model was further validated on 196 patients originating from The Christie (United Kingdon). A Bayesian network model was adapted for distributed learning (the animation can be viewed at https://www.youtube.com/watch?v=ZDJFOxpwqEA). Two-year posttreatment survival was chosen as the endpoint. The Maastro clinic cohort data are publicly available at https://www.cancerdata.org/publication/developing-and-validating-survival-prediction-model-nsclc-patients-through-distributed, and the developed models can be found at www.predictcancer.org.Results: Variables included in the final model were T and N category, age, performance status, and total tumor dose. The model has an area under the curve (AUC) of 0.66 on the external validation set and an AUC of 0.62 on a 5-fold cross validation. A model based on the T and N category performed with an AUC of 0.47 on the validation set, significantly worse than our model (P<.001). Learning the model in a centralized or distributed fashion yields a minor difference on the probabilities of the conditional probability tables (0.6%); the discriminative performance of the models on the validation set is similar (P=.26).Conclusions: Distributed learning from federated databases allows learning of predictive models on data originating from multiple institutions while avoiding many of the data-sharing barriers. We believe that distributed learning is the future of sharing data in health care. [ABSTRACT FROM AUTHOR]

• Association between EBRT characteristics and SPC risk after EBRT for PCa. • Relative risks (SIRs) to the general population were calculated for the RT groups. • Increased pelvis SPC risks were lower with advanced EBRT versus 3D-CRT. • Non-pelvis SPC risks did not correlate with evaluated EBRT characteristics. Modelling studies suggest that advanced intensity-modulated radiotherapy may increase second primary cancer (SPC) risks, due to increased radiation exposure of tissues located outside the treatment fields. In the current study we investigated the association between SPC risks and characteristics of applied external beam radiotherapy (EBRT) protocols for localized prostate cancer (PCa). We collected EBRT protocol characteristics (2000–2016) from five Dutch RT institutes for the 3D-CRT and advanced EBRT era (N = 7908). From the Netherlands Cancer Registry we obtained patient/tumour characteristics, SPC data, and survival information. Standardized incidence ratios (SIR) were calculated for pelvis and non-pelvis SPC. Nationwide SIRs were calculated as a reference, using calendar period as a proxy to label 3D-CRT/advanced EBRT. From 2000-2006, 3D-CRT with 68–78 Gy in 2 Gy fractions, delivered with 10–23 MV and weekly portal imaging was the most dominant protocol. By the year 2010 all institutes routinely used advanced EBRT (IMRT, VMAT, tomotherapy), mainly delivering 78 Gy in 2 Gy fractions, using various kV/MV imaging protocols. Sixteen percent (N = 1268) developed ≥ 1 SPC. SIRs for pelvis and non-pelvis SPC (all institutes, advanced EBRT vs 3D-CRT) were 1.17 (1.00–1.36) vs 1.39 (1.21–1.59), and 1.01 (0.89–1.07) vs 1.03 (0.94–1.13), respectively. Nationwide non-pelvis SIR was 1.07 (1.01–1.13) vs 1.02 (0.98–1.07). Other RT protocol characteristics did not correlate with SPC endpoints. None of the studied RT characteristics of advanced EBRT was associated with increased out-of-field SPC risks. With constantly evolving EBRT protocols, evaluation of associated SPC risks remains important. [ABSTRACT FROM AUTHOR]

• CT and radiation dose can play synergistically with each other to model. • Deep learning methods have higher predictive power than traditional methods. • The predictive logic of deep learning can be inferred by activation maps. • Simple tuning method can be used in deep learning model to adapt to different data. To develop a deep learning model that combines CT and radiation dose (RD) images to predict the occurrence of radiation pneumonitis (RP) in lung cancer patients who received radical (chemo)radiotherapy. CT, RD images and clinical parameters were obtained from 314 retrospectively-collected patients (training set) and 35 prospectively-collected patients (test-set-1) who were diagnosed with lung cancer and received radical radiotherapy in the dose range of 50 Gy and 70 Gy. Another 194 (60 Gy group, test-set-2) and 158 (74 Gy group, test-set-3) patients from the clinical trial RTOG 0617 were used for external validation. A ResNet architecture was used to develop a prediction model that combines CT and RD features. Thereafter, the CT and RD weights were adjusted by using 40 patients from test-set-2 or 3 to accommodate cohorts with different clinical settings or dose delivery patterns. Visual interpretation was implemented using a gradient-weighted class activation map (grad-CAM) to observe the area of model attention during the prediction process. To improve the usability, ready-to-use online software was developed. The discriminative ability of a baseline trained model had an AUC of 0.83 for test-set-1, 0.55 for test-set-2, and 0.63 for test-set-3. After adjusting CT and RD weights of the model using a subset of the RTOG-0617 subjects, the discriminatory power of test-set-2 and 3 improved to AUC 0.65 and AUC 0.70, respectively. Grad-CAM showed the regions of interest to the model that contribute to the prediction of RP. A novel deep learning approach combining CT and RD images can effectively and accurately predict the occurrence of RP, and this model can be adjusted easily to fit new cohorts. [ABSTRACT FROM AUTHOR]

Purpose One of the major hurdles in enabling personalized medicine is obtaining sufficient patient data to feed into predictive models. Combining data originating from multiple hospitals is difficult because of ethical, legal, political, and administrative barriers associated with data sharing. In order to avoid these issues, a distributed learning approach can be used. Distributed learning is defined as learning from data without the data leaving the hospital. Patients and methods Clinical data from 287 lung cancer patients, treated with curative intent with chemoradiation (CRT) or radiotherapy (RT) alone were collected from and stored in 5 different medical institutes (123 patients at MAASTRO (Netherlands, Dutch), 24 at Jessa (Belgium, Dutch), 34 at Liege (Belgium, Dutch and French), 48 at Aachen (Germany, German) and 58 at Eindhoven (Netherlands, Dutch)). A Bayesian network model is adapted for distributed learning (watch the animation: http://youtu.be/nQpqMIuHyOk ). The model predicts dyspnea, which is a common side effect after radiotherapy treatment of lung cancer. Results We show that it is possible to use the distributed learning approach to train a Bayesian network model on patient data originating from multiple hospitals without these data leaving the individual hospital. The AUC of the model is 0.61 (95%CI, 0.51–0.70) on a 5-fold cross-validation and ranges from 0.59 to 0.71 on external validation sets. Conclusion Distributed learning can allow the learning of predictive models on data originating from multiple hospitals while avoiding many of the data sharing barriers. Furthermore, the distributed learning approach can be used to extract and employ knowledge from routine patient data from multiple hospitals while being compliant to the various national and European privacy laws. [ABSTRACT FROM AUTHOR]

Disconnected cancer research data management and lack of information exchange about planned and ongoing research are complicating the utilisation of internationally collected medical information for improving cancer patient care. Rapidly collecting/pooling data can accelerate translational research in radiation therapy and oncology. The exchange of study data is one of the fundamental principles behind data aggregation and data mining. The possibilities of reproducing the original study results, performing further analyses on existing research data to generate new hypotheses or developing computational models to support medical decisions (e.g. risk/benefit analysis of treatment options) represent just a fraction of the potential benefits of medical data-pooling. Distributed machine learning and knowledge exchange from federated databases can be considered as one beyond other attractive approaches for knowledge generation within “Big Data”. Data interoperability between research institutions should be the major concern behind a wider collaboration. Information captured in electronic patient records (EPRs) and study case report forms (eCRFs), linked together with medical imaging and treatment planning data, are deemed to be fundamental elements for large multi-centre studies in the field of radiation therapy and oncology. To fully utilise the captured medical information, the study data have to be more than just an electronic version of a traditional (un-modifiable) paper CRF. Challenges that have to be addressed are data interoperability, utilisation of standards, data quality and privacy concerns, data ownership, rights to publish, data pooling architecture and storage. This paper discusses a framework for conceptual packages of ideas focused on a strategic development for international research data exchange in the field of radiation therapy and oncology. [ABSTRACT FROM AUTHOR]

Abstract: Purpose: An overview of the Rapid Learning methodology, its results, and the potential impact on radiotherapy. Material and results: Rapid Learning methodology is divided into four phases. In the data phase, diverse data are collected about past patients, treatments used, and outcomes. Innovative information technologies that support semantic interoperability enable distributed learning and data sharing without additional burden on health care professionals and without the need for data to leave the hospital. In the knowledge phase, prediction models are developed for new data and treatment outcomes by applying machine learning methods to data. In the application phase, this knowledge is applied in clinical practice via novel decision support systems or via extensions of existing models such as Tumour Control Probability models. In the evaluation phase, the predictability of treatment outcomes allows the new knowledge to be evaluated by comparing predicted and actual outcomes. Conclusion: Personalised or tailored cancer therapy ensures not only that patients receive an optimal treatment, but also that the right resources are being used for the right patients. Rapid Learning approaches combined with evidence based medicine are expected to improve the predictability of outcome and radiotherapy is the ideal field to study the value of Rapid Learning. The next step will be to include patient preferences in the decision making. [Copyright &y& Elsevier]

Abstract: Background: Individualised, isotoxic, accelerated radiotherapy (INDAR) allows the delivery of high biological radiation doses, but the long-term survival associated with this approach is unknown. Methods: Patients with stage III NSCLC in the Netherlands Cancer Registry/Limburg from January 1, 2002 to December 31, 2008 were included. Results: Patients (1002) with stage III NSCLC were diagnosed, of which 938 had T4 and/or N2–N3 disease. Patients treated with curative intent were staged with FDG-PET scans and a contrast-enhanced CT or an MRI of the brain. There were no shifts over time in the patient or tumour characteristics at diagnosis. The number of stage III NSCLC patients remained stable over time, but the proportion treated with palliative intent decreased from 47% in 2002 to 37% in 2008, and the percentage treated with chemo-radiation (RT) increased from 24.6% in 2002 to 47.8% in 2008 (p <0.001). The proportion of surgical patients remained below 5%. Sequential chemotherapy and conventional RT resulted in a median and a 5-year survival of 17.5months and 8.4%, respectively, whereas with sequential chemotherapy and INDAR this was 23.6months and 31%, respectively (p <0.001). Concurrent chemotherapy and INDAR was associated with a median and 2-year survival that was not reached and 66.7%, respectively (p =0.004). Conclusions: The proportion of patients treated with a curative intention with chemo-RT has increased markedly over time of observation. INDAR is associated with longer survival when compared to standard dose RT alone given with or without chemotherapy. [Copyright &y& Elsevier]

Abstract: Purpose: As Radiation-Induced Lung Toxicity (RILT) is dose-limiting for radiotherapy (RT) of lung cancer and current parameters are only moderately associated with RILT, we sought for novel parameters associated with RILT. Patients and methods: In this prospective study, FDG-PET-CT scans were taken on days 0, 7 and 14 after initiation of high-dose RT in 18 patients with stage III non-small cell lung cancer. The maximal Standardized Uptake Value (SUVmax) in the lung outside of the GTV was used as a measure of FDG uptake. At the same time-points, the serum IL-6 concentrations were measured. RILT was defined as dyspnea score ⩾2 (CTCAE3.0). Results: Six of 18 patients developed RILT. Before RT, SUVmax in the lung was not significantly different between patients who developed RILT and those who did not develop RILT. Patients who developed RILT post-radiation had a significant increased SUV on days 7 and 14 during RT, whereas the group that did not experience RILT showed no significant SUV changes. The SUVmax of the lungs increased significantly more in the group that later developed RILT compared to those who did not develop RILT. Neither the IL-6 concentration nor the mean lung dose was associated with RILT. Conclusions: The increase in FDG uptake in the normal lung early during RT was highly associated with the subsequent development of clinical RILT. This may help to identify patients at high risk for RILT at a time when adjustments of the treatment or strategies to prevent RILT are still possible. [Copyright &y& Elsevier]

Abstract: Purpose: To determine what the influence is of dyspnea (CTCAE3.0) before high-dose radiotherapy (RT) on the incidence and severity of subsequent lung toxicity in patients with non-small cell lung cancer (NSCLC). Methods: In 197 patients with stage I-III NSCLC maximal dyspnea scores (CTCAE3.0) were obtained prospectively at three time periods: before RT, the first 6 months post-RT and 6–9 months post-RT. Only patients who were clinically progression-free 12 months or more after RT were included, thus minimizing dyspnea due to tumor progression. Time-trends of dyspnea as a function of baseline dyspnea were investigated and correlated with gender, age, chemotherapy, mean lung dose (MLD), lung function parameters (FeV1 and DLCO), stage, PTV dose, overall treatment time and smoking habits. Results: The proportion developing less, the same or more dyspnea 6–9 months post-treatment according to their baseline dyspnea scores was: Grade 0: none, 82.9%, 17.1%; Grade 1: 21.2%, 51.9%, 26.9%; Grade 2: 27.3%, 54.5%, 18.2%, respectively. Only age was associated with increased dyspnea after RT. Conclusions: Patients with dyspnea before therapy have a realistic chance to improve after high-dose radiotherapy. Reporting only dyspnea at one time-point post-RT is insufficient to determine radiation-induced dyspnea. [Copyright &y& Elsevier]

Purpose: Because individual tumors are heterogeneous, including for 18F-deoxyglucose (FDG) uptake and, most likely, for radioresistance, selective boosting of high FDG uptake zones within the tumor has been suggested. To do this, it is critical to know whether the location of these high FDG uptake patterns within the tumor remain stable during radiotherapy (RT). Methods and Materials: Twenty-three patients with Stage I-III non–small-cell lung cancer underwent repeated FDG positron emission tomography computed tomography scans before radical RT (Day 0) and at Days 7 and 14 of RT. On all scans, the high and low FDG uptake regions were autodelineated using several standardized uptake value thresholds, varying from 34% to 80% of the maximal standardized uptake value. The volumes and overlap fractions of these delineations were calculated to demonstrate the stability of the high FDG uptake regions during RT. Results: The mean overlap fraction of the 34% uptake zones at Day 0 with Days 7 and 14 was 82.8% ± 8.1% and 84.3% ± 7.6%, respectively. The mean overlap fraction of the high uptake zones (60%) was 72.3% ± 15.0% and 71.3% ± 19.7% at Day 0 with Days 7 and 14, respectively. The volumes of the thresholds varied markedly (e.g., at Day 0, the volume of the 60% zone was 16.8 ± 20.3 cm3). In contrast, although the location of the high FDG uptake patterns within the tumor during RT remained stable, the delineated volumes varied markedly. Conclusion: The location of the low and high FDG uptake areas within the tumor remained stable during RT. This knowledge may enable selective boosting of high FDG uptake areas within the tumor. [Copyright &y& Elsevier]

Abstract: Purpose: To integrate PET-CT scans from different hospitals into radiotherapy treatment planning. Methods and materials: A cylindrical phantom with spheres of different diameters was scanned on three different Siemens Biograph PET-CT scanners in three hospitals. The spheres and cylinder were filled with 18F-FDG such that different sphere-to-background (S/B) ratios were obtained. Scans were analyzed using dedicated software for automated delineation based on standardized uptake value (SUV) and using different reconstruction parameters. Results: SUV thresholding curves for different S/B ratios were obtained for the different scanners. Differences in SUV auto-contouring thresholds were found to be significant for PET-CT simulators from different radiotherapy and nuclear medicine departments. A change in PET reconstruction parameters showed a significant effect on the results. Conclusion: Synchronization of PET-CT imaging protocols between cooperating hospitals is important for reliable determination of SUV auto-contouring thresholds. Whenever this goal has been achieved automated SUV delineation based on a S/B ratio using PET-CT images from different institutions can reliably be performed using individually determined threshold curves. [Copyright &y& Elsevier]