Search

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.

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.

The registration of multi-source radiation oncology data is a time-consuming and labour-intensive procedure. The standardisation of data collection offers the possibility for the acquisition of quality data for research and clinical purposes. With this study, we present an overview of the different tumour group data lists in the Dutch national proton therapy registry. Furthermore, as a representative example of the workings of these different tumour-specific knowledge graphs, we present the FAIR (Findable, Accessible, Interoperable, Reusable) data principles-compliant knowledge graph approach describing the head and neck tumour variables using radiotherapy domain ontologies and semantic web technologies. Our goal is to provide the radiotherapy community with a flexible and interoperable data model for data exchange between centres. We highlight data variables that are needed for models used in the model-based approach (MBA), which ensures a fair selection of patients that will benefit most from proton therapy. [ABSTRACT FROM AUTHOR]

Purpose Multiple models have been developed to predict pathologic complete response ( pCR) in locally advanced rectal cancer patients. Unfortunately, validation of these models normally omit the implications of cohort differences on prediction model performance. In this work, we will perform a prospective validation of three pCR models, including information whether this validation will target transferability or reproducibility (cohort differences) of the given models. Methods We applied a novel methodology, the cohort differences model, to predict whether a patient belongs to the training or to the validation cohort. If the cohort differences model performs well, it would suggest a large difference in cohort characteristics meaning we would validate the transferability of the model rather than reproducibility. We tested our method in a prospective validation of three existing models for pCR prediction in 154 patients. Results Our results showed a large difference between training and validation cohort for one of the three tested models [Area under the Receiver Operating Curve ( AUC) cohort differences model: 0.85], signaling the validation leans towards transferability. Two out of three models had a lower AUC for validation (0.66 and 0.58), one model showed a higher AUC in the validation cohort (0.70). Discussion We have successfully applied a new methodology in the validation of three prediction models, which allows us to indicate if a validation targeted transferability (large differences between training/validation cohort) or reproducibility (small cohort differences). [ABSTRACT FROM AUTHOR]