A decision aid for intensity-modulated radiation-therapy plan selection in prostate cancer based on a prognostic Bayesian network and a Markov model

Improving the Quality of Care in Radiation Oncology using Artificial Intelligence

Radiation therapy is a complex process involving multiple professionals and steps from simulation to treatment planning to delivery, and these procedures are prone to error. Additionally, the imaging and treatment delivery equipment in radiotherapy is highly complex and interconnected and represents another risk point in the quality of care. Numerous quality assurance tasks are carried out to ensure quality and to detect and prevent potential errors in the process of care. Recent developments in artificial intelligence provide potential tools to the radiation oncology community to improve the efficiency and performance of quality assurance efforts. Targets for artificial intelligence enhancement include the quality assurance of treatment plans, target and tissue structure delineation used in the plans, delivery of the plans and the radiotherapy delivery equipment itself. Here we review recent developments of artificial intelligence applications that aim to improve quality assurance processes in radiation therapy and discuss some of the challenges and limitations that require further development work to realise the potential of artificial intelligence for quality assurance.

A novel method for predicting the progression rate of ALS disease based on automatic generation of probabilistic causal chains

Causal discovery is considered as a major concept in biomedical informatics contributing to diagnosis, therapy, and prognosis of diseases. Probabilistic causality approaches in epidemiology and medicine is a common method for finding relationships between pathogen and disease, environment and disease, and adverse events and drugs. Bayesian Network (BN) is one of the common approaches for probabilistic causality, which is widely used in health-care and biomedical science. Since in many biomedical applications we deal with temporal dataset, the temporal extension of BNs called Dynamic Bayesian network (DBN) is used for such applications. DBNs define probabilistic relationships between parameters in consecutive time points in the form of a graph and have been successfully used in many biomedical applications. In this paper, a novel method was introduced for finding probabilistic causal chains from a temporal dataset with the help of entropy and causal tendency measures. In this method, first, Causal Features Dependency (CFD) matrix is created on the basis of parameters changes in consecutive events of a phenomenon, and then the probabilistic causal graph is constructed from this matrix based on entropy criteria. At the next step, a set of probabilistic causal chains of the corresponding causal graph is constructed by a novel polynomial-time heuristic. Finally, the causal chains are used for predicting the future trend of the phenomenon. The proposed model was applied to the Pooled Resource Open-Access Clinical Trials (PRO-ACT) dataset related to Amyotrophic Lateral Sclerosis (ALS) disease, in order to predict the progression rate of this disease. The results of comparison with Bayesian tree, random forest, support vector regression, linear regression, and multivariate regression show that the proposed algorithm can compete with these methods and in some cases outperforms other algorithms. This study revealed that probabilistic causality is an appropriate approach for predicting the future states of chronic diseases with unknown cause.

Artificial intelligence in radiotherapy

Artificial intelligence (AI) has already been implemented widely in the medical field in the recent years. This paper first reviews the background of AI and radiotherapy. Then it explores the basic concepts of different AI algorithms and machine learning methods, such as neural networks, that are available to us today and how they are being implemented in radiotherapy and diagnostic processes, such as medical imaging, treatment planning, patient simulation, quality assurance and radiation dose delivery. It also explores the ongoing research on AI methods that are to be implemented in radiotherapy in the future. The review shows very promising progress and future for AI to be widely used in various areas of radiotherapy. However, basing on various concerns such as availability and security of using big data, and further work on polishing and testing AI algorithms, it is found that we may not ready to use AI primarily in radiotherapy at the moment.

Ontologies in radiation oncology

Ontologies are a formal, computer-compatible method for representing scientific knowledge about a given domain. They provide a standardized vocabulary, taxonomy and set of relations between concepts. When formatted in a standard way, they can be read and reasoned upon by computers as well as by humans. At the 2019 International Conference on the Use of Computers in Radiation Therapy, there was a session devoted to ontologies in radiation therapy. This paper is a compilation of the material presented, and is meant as an introduction to the subject. This is done by means of a didactic introduction to the topic followed by a series of applications in radiation therapy. The goal of this article is to provide the medical physicist and related professionals with sufficient background that they can understand their construction as well as their practical uses.

A Three-Part Bayesian Network for Modeling Dwelling Fires and Their Impact upon People and Property

In the United Kingdom, dwelling fires are responsible for the majority of all fire-related fatalities. The development of these incidents involves the interaction of a multitude of variables that combine in many different ways. Consequently, assessment of dwelling fire risk can be complex, which often results in ambiguity during fire safety planning and decision making. In this article, a three-part Bayesian network model is proposed to study dwelling fires from ignition through to extinguishment in order to improve confidence in dwelling fire safety assessment. The model incorporates both hard and soft data, delivering posterior probabilities for selected outcomes. Case studies demonstrate how the model functions and provide evidence of its use for planning and accident investigation.

Applying Probabilistic Decision Models to Clinical Trial Design

Clinical trial design most often focuses on a single or several related outcomes with corresponding calculations of statistical power. We consider a clinical trial to be a decision problem, often with competing outcomes. Using a current controversy in the treatment of HPV-positive head and neck cancer, we apply several different probabilistic methods to help define the range of outcomes given different possible trial designs. Our model incorporates the uncertainties in the disease process and treatment response and the inhomogeneities in the patient population. Instead of expected utility, we have used a Markov model to calculate quality adjusted life expectancy as a maximization objective. Monte Carlo simulations over realistic ranges of parameters are used to explore different trial scenarios given the possible ranges of parameters. This modeling approach can be used to better inform the initial trial design so that it will more likely achieve clinical relevance.

Big Data Analytics for Prostate Radiotherapy

Radiation therapy is a first-line treatment option for localized prostate cancer and radiation-induced normal tissue damage are often the main limiting factor for modern radiotherapy regimens. Conversely, under-dosing of target volumes in an attempt to spare adjacent healthy tissues limits the likelihood of achieving local, long-term control. Thus, the ability to generate personalized data-driven risk profiles for radiotherapy outcomes would provide valuable prognostic information to help guide both clinicians and patients alike. Big data applied to radiation oncology promises to deliver better understanding of outcomes by harvesting and integrating heterogeneous data types, including patient-specific clinical parameters, treatment-related dose-volume metrics, and biological risk factors. When taken together, such variables make up the basis for a multi-dimensional space (the "RadoncSpace") in which the presented modeling techniques search in order to identify significant predictors. Herein, we review outcome modeling and big data-mining techniques for both tumor control and radiotherapy-induced normal tissue effects. We apply many of the presented modeling approaches onto a cohort of hypofractionated prostate cancer patients taking into account different data types and a large heterogeneous mix of physical and biological parameters. Cross-validation techniques are also reviewed for the refinement of the proposed framework architecture and checking individual model performance. We conclude by considering advanced modeling techniques that borrow concepts from big data analytics, such as machine learning and artificial intelligence, before discussing the potential future impact of systems radiobiology approaches.

Bayesian network ensemble as a multivariate strategy to predict radiation pneumonitis risk

PURPOSE

Prediction of radiation pneumonitis (RP) has been shown to be challenging due to the involvement of a variety of factors including dose-volume metrics and radiosensitivity biomarkers. Some of these factors are highly correlated and might affect prediction results when combined. Bayesian network (BN) provides a probabilistic framework to represent variable dependencies in a directed acyclic graph. The aim of this study is to integrate the BN framework and a systems' biology approach to detect possible interactions among RP risk factors and exploit these relationships to enhance both the understanding and prediction of RP.

METHODS

The authors studied 54 nonsmall-cell lung cancer patients who received curative 3D-conformal radiotherapy. Nineteen RP events were observed (common toxicity criteria for adverse events grade 2 or higher). Serum concentration of the following four candidate biomarkers were measured at baseline and midtreatment: alpha-2-macroglobulin, angiotensin converting enzyme (ACE), transforming growth factor, interleukin-6. Dose-volumetric and clinical parameters were also included as covariates. Feature selection was performed using a Markov blanket approach based on the Koller-Sahami filter. The Markov chain Monte Carlo technique estimated the posterior distribution of BN graphs built from the observed data of the selected variables and causality constraints. RP probability was estimated using a limited number of high posterior graphs (ensemble) and was averaged for the final RP estimate using Bayes' rule. A resampling method based on bootstrapping was applied to model training and validation in order to control under- and overfit pitfalls.

RESULTS

RP prediction power of the BN ensemble approach reached its optimum at a size of 200. The optimized performance of the BN model recorded an area under the receiver operating characteristic curve (AUC) of 0.83, which was significantly higher than multivariate logistic regression (0.77), mean heart dose (0.69), and a pre-to-midtreatment change in ACE (0.66). When RP prediction was made only with pretreatment information, the AUC ranged from 0.76 to 0.81 depending on the ensemble size. Bootstrap validation of graph features in the ensemble quantified confidence of association between variables in the graphs where ten interactions were statistically significant.

CONCLUSIONS

The presented BN methodology provides the flexibility to model hierarchical interactions between RP covariates, which is applied to probabilistic inference on RP. The authors' preliminary results demonstrate that such framework combined with an ensemble method can possibly improve prediction of RP under real-life clinical circumstances such as missing data or treatment plan adaptation.

Personalized treatment planning with a model of radiation therapy outcomes for use in multiobjective optimization of IMRT plans for prostate cancer

PURPOSE

To build a new treatment planning approach that extends beyond radiation transport and IMRT optimization by modeling the radiation therapy process and prognostic indicators for more outcome-focused decision making.

METHODS

An in-house treatment planning system was modified to include multiobjective inverse planning, a probabilistic outcome model, and a multi-attribute decision aid. A genetic algorithm generated a set of plans embodying trade-offs between the separate objectives. An influence diagram network modeled the radiation therapy process of prostate cancer using expert opinion, results of clinical trials, and published research. A Markov model calculated a quality adjusted life expectancy (QALE), which was the endpoint for ranking plans.

RESULTS

The Multiobjective Evolutionary Algorithm (MOEA) was designed to produce an approximation of the Pareto Front representing optimal tradeoffs for IMRT plans. Prognostic information from the dosimetrics of the plans, and from patient-specific clinical variables were combined by the influence diagram. QALEs were calculated for each plan for each set of patient characteristics. Sensitivity analyses were conducted to explore changes in outcomes for variations in patient characteristics and dosimetric variables. The model calculated life expectancies that were in agreement with an independent clinical study.

CONCLUSIONS

The radiation therapy model proposed has integrated a number of different physical, biological and clinical models into a more comprehensive model. It illustrates a number of the critical aspects of treatment planning that can be improved and represents a more detailed description of the therapy process. A Markov model was implemented to provide a stronger connection between dosimetric variables and clinical outcomes and could provide a practical, quantitative method for making difficult clinical decisions.

Bayesian network models for error detection in radiotherapy plans

The purpose of this study is to design and develop a probabilistic network for detecting errors in radiotherapy plans for use at the time of initial plan verification. Our group has initiated a multi-pronged approach to reduce these errors. We report on our development of Bayesian models of radiotherapy plans. Bayesian networks consist of joint probability distributions that define the probability of one event, given some set of other known information. Using the networks, we find the probability of obtaining certain radiotherapy parameters, given a set of initial clinical information. A low probability in a propagated network then corresponds to potential errors to be flagged for investigation. To build our networks we first interviewed medical physicists and other domain experts to identify the relevant radiotherapy concepts and their associated interdependencies and to construct a network topology. Next, to populate the network's conditional probability tables, we used the Hugin Expert software to learn parameter distributions from a subset of de-identified data derived from a radiation oncology based clinical information database system. These data represent 4990 unique prescription cases over a 5 year period. Under test case scenarios with approximately 1.5% introduced error rates, network performance produced areas under the ROC curve of 0.88, 0.98, and 0.89 for the lung, brain and female breast cancer error detection networks, respectively. Comparison of the brain network to human experts performance (AUC of 0.90 ± 0.01) shows the Bayes network model performs better than domain experts under the same test conditions. Our results demonstrate the feasibility and effectiveness of comprehensive probabilistic models as part of decision support systems for improved detection of errors in initial radiotherapy plan verification procedures.

Emergency department triaging of admitted stroke patients--a Bayesian Network analysis

This study uses hospital administrative data to ascertain the differences in the patient characteristics, process and outcomes of care between the Emergency Department (ED) triage categories of patients admitted from an ED presentation into a large metropolitan teaching hospital with a Stroke Care Unit. Bayesian Networks (BNs) derived from the administrative data were used to provide the descriptive models. Nearly half the patients in each stroke subtype were triaged as 'Urgent' (to be seen within 30 minutes). With a decrease in the urgency of triage categories, the proportion admitted within 8 hours decreased dramatically and the proportion of formal discharge increased. Notably, 45% of transient ischaemic attacks (TIAs) were categorized as 'Semi-urgent' (to be attended within 60 minutes), indicating an opportunity to improve emergency assessment of TIAs. The results illustrate the utility of hospital administrative data and the applicability of BNs for review of the current triage practices and subsequent impact.

Formal-Transfer In and Out of Stroke Care Units

The positive impact of stroke care units (SCUs) on patient outcome has been previously reported. In this study, long-term stroke patients that are formally admitted to teaching-hospitals are compared with and without SCUs. The authors focus on the patients’ experience with ongoing care or formal transfers following current care as this cohort is often high users of the system with associated high costs. Bayesian Networks were employed to analyze routinely collected public-hospital administrative data. The results illustrate that the teaching-hospitals with SCUs, while achieving shorter length of stay, in fact deal with younger patients with lower overall patient complexity than non-SCU teaching-hospitals. Other differences include SCUs predominantly treating subarachnoid hemorrhages whereas the non-SCUs treat more cerebral infarctions. This study illustrates the power of Bayesian Networks to expose the nature of caseload and outcomes recorded in hospital-administrative data as a means to gain insight on current practice and create opportunities for benchmarking and improving care.

A Bayesian network approach for modeling local failure in lung cancer

Locally advanced non-small cell lung cancer (NSCLC) patients suffer from a high local failure rate following radiotherapy. Despite many efforts to develop new dose-volume models for early detection of tumor local failure, there was no reported significant improvement in their application prospectively. Based on recent studies of biomarker proteins' role in hypoxia and inflammation in predicting tumor response to radiotherapy, we hypothesize that combining physical and biological factors with a suitable framework could improve the overall prediction. To test this hypothesis, we propose a graphical Bayesian network framework for predicting local failure in lung cancer. The proposed approach was tested using two different datasets of locally advanced NSCLC patients treated with radiotherapy. The first dataset was collected retrospectively, which comprises clinical and dosimetric variables only. The second dataset was collected prospectively in which in addition to clinical and dosimetric information, blood was drawn from the patients at various time points to extract candidate biomarkers as well. Our preliminary results show that the proposed method can be used as an efficient method to develop predictive models of local failure in these patients and to interpret relationships among the different variables in the models. We also demonstrate the potential use of heterogeneous physical and biological variables to improve the model prediction. With the first dataset, we achieved better performance compared with competing Bayesian-based classifiers. With the second dataset, the combined model had a slightly higher performance compared to individual physical and biological models, with the biological variables making the largest contribution. Our preliminary results highlight the potential of the proposed integrated approach for predicting post-radiotherapy local failure in NSCLC patients.

A hierarchical evolutionary algorithm for multiobjective optimization in IMRT

PURPOSE

The current inverse planning methods for intensity modulated radiation therapy (IMRT) are limited because they are not designed to explore the trade-offs between the competing objectives of tumor and normal tissues. The goal was to develop an efficient multiobjective optimization algorithm that was flexible enough to handle any form of objective function and that resulted in a set of Pareto optimal plans.

METHODS

A hierarchical evolutionary multiobjective algorithm designed to quickly generate a small diverse Pareto optimal set of IMRT plans that meet all clinical constraints and reflect the optimal trade-offs in any radiation therapy plan was developed. The top level of the hierarchical algorithm is a multiobjective evolutionary algorithm (MOEA). The genes of the individuals generated in the MOEA are the parameters that define the penalty function minimized during an accelerated deterministic IMRT optimization that represents the bottom level of the hierarchy. The MOEA incorporates clinical criteria to restrict the search space through protocol objectives and then uses Pareto optimality among the fitness objectives to select individuals. The population size is not fixed, but a specialized niche effect, domination advantage, is used to control the population and plan diversity. The number of fitness objectives is kept to a minimum for greater selective pressure, but the number of genes is expanded for flexibility that allows a better approximation of the Pareto front.

RESULTS

The MOEA improvements were evaluated for two example prostate cases with one target and two organs at risk (OARs). The population of plans generated by the modified MOEA was closer to the Pareto front than populations of plans generated using a standard genetic algorithm package. Statistical significance of the method was established by compiling the results of 25 multiobjective optimizations using each method. From these sets of 12-15 plans, any random plan selected from a MOEA population had a 11.3% +/- 0.7% chance of dominating any random plan selected by a standard genetic package with 0.04% +/- 0.02% chance of domination in reverse. By implementing domination advantage and protocol objectives, small and diverse populations of clinically acceptable plans that approximated the Pareto front could be generated in a fraction of 1 h. Acceleration techniques implemented on both levels of the hierarchical algorithm resulted in short, practical runtimes for multiobjective optimizations.

CONCLUSIONS

The MOEA produces a diverse Pareto optimal set of plans that meet all dosimetric protocol criteria in a feasible amount of time. The final goal is to improve practical aspects of the algorithm and integrate it with a decision analysis tool or human interface for selection of the IMRT plan with the best possible balance of successful treatment of the target with low OAR dose and low risk of complication for any specific patient situation.

The ESTRO Breur Lecture 2009. From population to voxel-based radiotherapy: exploiting intra-tumour and intra-organ heterogeneity for advanced treatment of non-small cell lung cancer

Evidence is accumulating that radiotherapy of non-small cell lung cancer patients can be optimized by escalating the tumour dose until the normal tissue tolerances are met. To further improve the therapeutic ratio between tumour control probability and the risk of normal tissue complications, we firstly need to exploit inter patient variation. This variation arises, e.g. from differences in tumour shape and size, lung function and genetic factors. Secondly improvement is achieved by taking into account intra-tumour and intra-organ heterogeneity derived from molecular and functional imaging. Additional radiation dose must be delivered to those parts of the tumour that need it the most, e.g. because of increased radio-resistance or reduced therapeutic drug uptake, and away from regions inside the lung that are most prone to complication. As the delivery of these treatments plans is very sensitive for geometrical uncertainties, probabilistic treatment planning is needed to generate robust treatment plans. The administration of these complicated dose distributions requires a quality assurance procedure that can evaluate the treatment delivery and, if necessary, adapt the treatment plan during radiotherapy.

Evaluation of an automated method to assist with error detection in the ACCORD central laboratory

Background Errors in clinical laboratory data are rare but their potential high cost both in terms of harm to the subject as well as diluted statistical power results in a significant workload for experts, who must review large volumes of data in the search for these errors.

Purpose The current research objective is to develop and evaluate a method to assist in detecting potential errors in laboratory data for an interventional clinical trial, such as Action to Control Cardiovascular Risk in Diabetes, where treatment effects may be influenced by errors in the data.

Methods Utilizing data from a clinical trial investigating the effect of intensive glycemic control on major cardiovascular disease events, we constructed an algorithm that conducts probabilistic error detection called a ‘Bayesian network’. Using a synthetic error model, errors were introduced into a testing dataset, and the Bayesian network’s performance in identifying those errors was compared to laboratory experts. For each laboratory result we used the Bayesian network to compute the probability, the measured value was erroneous. This probability was then used to compute a receiver operating characteristic (ROC) curve. Three laboratory experts were recruited and took a survey consisting of 200 laboratory results. The task was to evaluate if these results were erroneous or not and to provide a confidence rating on a 6-point subjective probability scale.

Results The Bayesian network’s overall area under the ROC curve was calculated to be 0.79, whereas the three laboratory experts had areas under the ROC curve of 0.73, 0.73, and 0.72. Perfect error prediction and random guessing yield a ROC of 1.00 and 0.50, respectively. This difference in performance was statistically significant for all three experts. Human experts were also generally overconfident in their ability to detect errors.

Limitations The model is, by design, specific to a novel intervention in a specific diabetic population and, therefore, the specific Bayesian network discussed may not generalize to other interventions and populations. In addition, the study is limited by the small number of expert eligible to complete the survey.

Conclusions The results of this study suggest that continuous Bayesian networks, suitably constructed, may serve as an effective tool to assist experts in the review of voluminous laboratory data by flagging unlikely results for more thorough review. Clinical Trials 2010; 7: 380—389. http://ctj.sagepub.com

Comparison of Bayesian network and support vector machine models for two-year survival prediction in lung cancer patients treated with radiotherapy

PURPOSE

Classic statistical and machine learning models such as support vector machines (SVMs) can be used to predict cancer outcome, but often only perform well if all the input variables are known, which is unlikely in the medical domain. Bayesian network (BN) models have a natural ability to reason under uncertainty and might handle missing data better. In this study, the authors hypothesize that a BN model can predict two-year survival in non-small cell lung cancer (NSCLC) patients as accurately as SVM, but will predict survival more accurately when data are missing.

METHODS

A BN and SVM model were trained on 322 inoperable NSCLC patients treated with radiotherapy from Maastricht and validated in three independent data sets of 35, 47, and 33 patients from Ghent, Leuven, and Toronto. Missing variables occurred in the data set with only 37, 28, and 24 patients having a complete data set.

RESULTS

The BN model structure and parameter learning identified gross tumor volume size, performance status, and number of positive lymph nodes on a PET as prognostic factors for two-year survival. When validated in the full validation set of Ghent, Leuven, and Toronto, the BN model had an AUC of 0.77, 0.72, and 0.70, respectively. A SVM model based on the same variables had an overall worse performance (AUC 0.71, 0.68, and 0.69) especially in the Ghent set, which had the highest percentage of missing the important GTV size data. When only patients with complete data sets were considered, the BN and SVM model performed more alike.

CONCLUSIONS

Within the limitations of this study, the hypothesis is supported that BN models are better at handling missing data than SVM models and are therefore more suitable for the medical domain. Future works have to focus on improving the BN performance by including more patients, more variables, and more diversity.

Role of positron emission tomography in the treatment of occult disease in head-and-neck cancer: a modeling approach

PURPOSE

To determine under what conditions positron emission tomography (PET) imaging will be useful in decisions regarding the use of radiotherapy for the treatment of clinically occult lymph node metastases in head-and-neck cancer.

METHODS AND MATERIALS

A decision model of PET imaging and its downstream effects on radiotherapy outcomes was constructed using an influence diagram. This model included the sensitivity and specificity of PET, as well as the type and stage of the primary tumor. These parameters were varied to determine the optimal strategy for imaging and therapy for different clinical situations. Maximum expected utility was the metric by which different actions were ranked.

RESULTS

For primary tumors with a low probability of lymph node metastases, the sensitivity of PET should be maximized, and 50 Gy should be delivered if PET is positive and 0 Gy if negative. As the probability for lymph node metastases increases, PET imaging becomes unnecessary in some situations, and the optimal dose to the lymph nodes increases. The model needed to include the causes of certain health states to predict current clinical practice.

CONCLUSION

The model demonstrated the ability to reproduce expected outcomes for a range of tumors and provided recommendations for different clinical situations. The differences between the optimal policies and current clinical practice are likely due to a disparity between stated clinical decision processes and actual decision making by clinicians.