Measured vs simulated portal images for low MU fields on three accelerator types: possible consequences for 2D portal dosimetry.

PURPOSE: As external beam treatment plans become more dynamic and the dose to normal tissue is further constrained, treatments may consist of a larger number of beams, each delivering smaller doses (or monitor units, MU), in, e.g., volumetric modulated arc therapy (VMAT). Electronic portal imaging devices (EPID) may be used to verify external beam treatments on integrated fractions as well as in a more time dependent manner such as field by field. For treatment verification performed during a fraction (e.g., individual fields or VMAT control points), the lower limit of EPID measurement capability becomes important. The authors quantified the signal and timing accuracy of EPID images for low MU intensity modulated radiotherapy (IMRT) and conformal fields. METHODS: EPID images were collected from three different vendor's accelerators for low MU fields and compared to expected images. Simulations were performed to replicate the EPID acquisition pattern and to enhance the understanding of EPID readout schemes. RESULTS: Large discrepancies between observed and predicted images were noted due to an under-response to single low MU fields. It is shown that a variability of up to 37% can be observed for low MU fields in clinically used EPID acquisition modes and that the majority of this variability can be accounted for by the readout scheme, integration, and timing of EPID acquisitions. Simulations have confirmed the causes of the discrepancies. The occurrence and extent of the variation has been estimated for clinical settings. CONCLUSIONS: Incorrect absolute EPID signals collected for low MU fields in external beam treatments will negatively affect quantitative applications such as individual field based EPID dosimetry, typically appearing as an underdose, unless corrections to currently employed EPID readout schemes are made.

An Evaluation of Atlas Selection Methods for Atlas-Based Automatic Segmentation in Radiotherapy Treatment Planning.

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.

Can Atlas-Based Auto-Segmentation Ever Be Perfect? Insights From Extreme Value Theory.

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=316 ) and thoracic ( N=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.

Radiation Oncology Terminology Linker: A Step Towards a Linked Data Knowledge Base.

Performing image feature extraction in radiation oncology is often dependent on the organ and tumor delineations provided by clinical staff. These delineation names are free text DICOM metadata fields resulting in undefined information, which requires effort to use in large-scale image feature extraction efforts. In this work we present a scale-able solution to overcome these naming convention challenges with a REST service using Semantic Web technology to convert this information to linked data. As a proof of concept an open source software is used to compute radiation oncology image features. The results of this work can be found in a public Bitbucket repository.

Clinical evaluation of atlas and deep learning based automatic contouring for lung cancer.

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.

Comparative evaluation of autocontouring in clinical practice: A practical method using the Turing test.

PURPOSE: Automated techniques for estimating the contours of organs and structures in medical images have become more widespread and a variety of measures are available for assessing their quality. Quantitative measures of geometric agreement, for example, overlap with a gold-standard delineation, are popular but may not predict the level of clinical acceptance for the contouring method. Therefore, surrogate measures that relate more directly to the clinical judgment of contours, and to the way they are used in routine workflows, need to be developed. The purpose of this study is to propose a method (inspired by the Turing Test) for providing contour quality measures that directly draw upon practitioners' assessments of manual and automatic contours. This approach assumes that an inability to distinguish automatically produced contours from those of clinical experts would indicate that the contours are of sufficient quality for clinical use. In turn, it is anticipated that such contours would receive less manual editing prior to being accepted for clinical use. In this study, an initial assessment of this approach is performed with radiation oncologists and therapists. METHODS: Eight clinical observers were presented with thoracic organ-at-risk contours through a web interface and were asked to determine if they were automatically generated or manually delineated. The accuracy of the visual determination was assessed, and the proportion of contours for which the source was misclassified recorded. Contours of six different organs in a clinical workflow were for 20 patient cases. The time required to edit autocontours to a clinically acceptable standard was also measured, as a gold standard of clinical utility. Established quantitative measures of autocontouring performance, such as Dice similarity coefficient with respect to the original clinical contour and the misclassification rate accessed with the proposed framework, were evaluated as surrogates of the editing time measured. RESULTS: The misclassification rates for each organ were: esophagus 30.0%, heart 22.9%, left lung 51.2%, right lung 58.5%, mediastinum envelope 43.9%, and spinal cord 46.8%. The time savings resulting from editing the autocontours compared to the standard clinical workflow were 12%, 25%, 43%, 77%, 46%, and 50%, respectively, for these organs. The median Dice similarity coefficients between the clinical contours and the autocontours were 0.46, 0.90, 0.98, 0.98, 0.94, and 0.86, respectively, for these organs. CONCLUSIONS: A better correspondence with time saving was observed for the misclassification rate than the quantitative contour measures explored. From this, we conclude that the inability to accurately judge the source of a contour indicates a reduced need for editing and therefore a greater time saving overall. Hence, task-based assessments of contouring performance may be considered as an additional way of evaluating the clinical utility of autosegmentation methods.

Machine learning algorithms for outcome prediction in (chemo)radiotherapy: An empirical comparison of classifiers.

PURPOSE: Machine 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. METHODS: We 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). RESULTS: Random 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. CONCLUSION: Random 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.

Big Data in radiation therapy: challenges and opportunities.

Data collected and generated by radiation oncology can be classified by the Volume, Variety, Velocity and Veracity (4Vs) of Big Data because they are spread across different care providers and not easily shared owing to patient privacy protection. The magnitude of the 4Vs is substantial in oncology, especially owing to imaging modalities and unclear data definitions. To create useful models ideally all data of all care providers are understood and learned from; however, this presents challenges in the guise of poor data quality, patient privacy concerns, geographical spread, interoperability and large volume. In radiation oncology, there are many efforts to collect data for research and innovation purposes. Clinical trials are the gold standard when proving any hypothesis that directly affects the patient. Collecting data in registries with strict predefined rules is also a common approach to find answers. A third approach is to develop data stores that can be used by modern machine learning techniques to provide new insights or answer hypotheses. We believe all three approaches have their strengths and weaknesses, but they should all strive to create Findable, Accessible, Interoperable, Reusable (FAIR) data. To learn from these data, we need distributed learning techniques, sending machine learning algorithms to FAIR data stores around the world, learning from trial data, registries and routine clinical data rather than trying to centralize all data. To improve and personalize medicine, rapid learning platforms must be able to process FAIR "Big Data" to evaluate current clinical practice and to guide further innovation.

Radiomics: the bridge between medical imaging and personalized medicine.

Radiomics, the high-throughput mining of quantitative image features from standard-of-care medical imaging that enables data to be extracted and applied within clinical-decision support systems to improve diagnostic, prognostic, and predictive accuracy, is gaining importance in cancer research. Radiomic analysis exploits sophisticated image analysis tools and the rapid development and validation of medical imaging data that uses image-based signatures for precision diagnosis and treatment, providing a powerful tool in modern medicine. Herein, we describe the process of radiomics, its pitfalls, challenges, opportunities, and its capacity to improve clinical decision making, emphasizing the utility for patients with cancer. Currently, the field of radiomics lacks standardized evaluation of both the scientific integrity and the clinical relevance of the numerous published radiomics investigations resulting from the rapid growth of this area. Rigorous evaluation criteria and reporting guidelines need to be established in order for radiomics to mature as a discipline. Herein, we provide guidance for investigations to meet this urgent need in the field of radiomics.

Decision support systems for personalized and participative radiation oncology.

A paradigm shift from current population based medicine to personalized and participative medicine is underway. This transition is being supported by the development of clinical decision support systems based on prediction models of treatment outcome. In radiation oncology, these models 'learn' using advanced and innovative information technologies (ideally in a distributed fashion - please watch the animation: http://youtu.be/ZDJFOxpwqEA) from all available/appropriate medical data (clinical, treatment, imaging, biological/genetic, etc.) to achieve the highest possible accuracy with respect to prediction of tumor response and normal tissue toxicity. In this position paper, we deliver an overview of the factors that are associated with outcome in radiation oncology and discuss the methodology behind the development of accurate prediction models, which is a multi-faceted process. Subsequent to initial development/validation and clinical introduction, decision support systems should be constantly re-evaluated (through quality assurance procedures) in different patient datasets in order to refine and re-optimize the models, ensuring the continuous utility of the models. In the reasonably near future, decision support systems will be fully integrated within the clinic, with data and knowledge being shared in a standardized, dynamic, and potentially global manner enabling truly personalized and participative medicine.

Implementation of a rapid learning platform: Predicting 2-year survival in laryngeal carcinoma patients in a clinical setting.

BACKGROUND AND PURPOSE: To improve quality and personalization of oncology health care, decision aid tools are needed to advise physicians and patients. The aim of this work is to demonstrate the clinical relevance of a survival prediction model as a first step to multi institutional rapid learning and compare this to a clinical trial dataset. MATERIALS AND METHODS: Data extraction and mining tools were used to collect uncurated input parameters from Illawarra Cancer Care Centre's (clinical cohort) oncology information system. Prognosis categories previously established from the Maastricht Radiation Oncology (training cohort) dataset, were applied to the clinical cohort and the radiotherapy only arm of the RTOG-9111 (trial cohort). RESULTS: Data mining identified 125 laryngeal carcinoma patients, ending up with 52 patients in the clinical cohort who were eligible to be evaluated by the model to predict 2-year survival and 177 for the trial cohort. The model was able to classify patients and predict survival in the clinical cohort, but for the trial cohort it failed to do so. CONCLUSIONS: The technical infrastructure and model is able to support the prognosis prediction of laryngeal carcinoma patients in a clinical cohort. The model does not perform well for the highly selective patient population in the trial cohort.

Development and evaluation of an online three-level proton vs photon decision support prototype for head and neck cancer - Comparison of dose, toxicity and cost-effectiveness.

To quantitatively assess the effectiveness of proton therapy for individual patients, we developed a prototype for an online platform for proton decision support (PRODECIS) comparing photon and proton treatments on dose metric, toxicity and cost-effectiveness levels. An evaluation was performed with 23 head and neck cancer datasets.

Set-up verification and 2-dimensional electronic portal imaging device dosimetry during breath hold compared with free breathing in breast cancer radiation therapy.

PURPOSE: To compare set-up and 2-dimensional (2D) electronic portal imaging device (EPID) dosimetry data of breast cancer patients treated during voluntary moderately deep inspiration breath hold (vmDIBH) and free breathing (FB). METHODS AND MATERIALS: Set-up data were analyzed for 29 and 51 consecutively treated patients, irradiated during FB and vmDIBH, respectively. Of the 51 vmDIBH patients, the first 25 had undergone an extra trained computed tomography (CT) scan and used an additional "breathing stick" (vmDIBH_trained). The last 26 patients did not use the breathing stick and did not undergo a trained CT (vmDIBH_untrained). The delivered 2D transit dose was measured with EPID in 15 FB and 28 vmDIBH patients and compared with a 2D predicted dose by calculating global gamma values γ using 5% and 5 mm as dose difference and distance-to-agreement criteria, respectively. Measurements with a percentage of pixels with an absolute gamma value > 1 (|γ| > 1) greater than 10% were classified as deviating. RESULTS: Only small, sub-millimeter differences were seen in the set-up data between the different patient groups. The mean of means, systematic error, and random error ranged from - 0.6 mm to 3.3 mm. The percentage of pixels with |γ| > 1 for all patients was 9.8% (2-25.8). No statistically significant differences were observed between the patient groups. In total, 38% of the gamma images were classified as deviating: 43.6% in vmDIBH_untrained patients compared with 38.0% in vmDIBH_trained patients and 33.3% in FB patients (P > .05). CONCLUSION: Both set-up and 2D EPID dosimetry data indicate that reproducibility of radiation therapy for patients treated during FB and vmDIBH is similar. Small but not significant differences in 2D EPID dosimetry were observed. Further investigation with 3-dimensional EPID dosimetry is recommended to investigate the clinical relevance of deviant gamma images.

Towards a semantic PACS: Using Semantic Web technology to represent imaging data.

The DICOM standard is ubiquitous within medicine. However, improved DICOM semantics would significantly enhance search operations. Furthermore, databases of current PACS systems are not flexible enough for the demands within image analysis research. In this paper, we investigated if we can use Semantic Web technology, to store and represent metadata of DICOM image files, as well as linking additional computational results to image metadata. Therefore, we developed a proof of concept containing two applications: one to store commonly used DICOM metadata in an RDF repository, and one to calculate imaging biomarkers based on DICOM images, and store the biomarker values in an RDF repository. This enabled us to search for all patients with a gross tumor volume calculated to be larger than 50 cc. We have shown that we can successfully store the DICOM metadata in an RDF repository and are refining our proof of concept with regards to volume naming, value representation, and the applications themselves.