Harnessing data science to advance radiation oncology

A joint physics and radiobiology DREAM team vision - Towards better response prediction models to advance radiotherapy

Radiotherapy developed empirically through experience balancing tumour control and normal tissue toxicities. Early simple mathematical models formalized this practical knowledge and enabled effective cancer treatment to date. Remarkable advances in technology, computing, and experimental biology now create opportunities to incorporate this knowledge into enhanced computational models. The ESTRO DREAM (Dose Response, Experiment, Analysis, Modelling) workshop brought together experts across disciplines to pursue the vision of personalized radiotherapy for optimal outcomes through advanced modelling. The ultimate vision is leveraging quantitative models dynamically during therapy to ultimately achieve truly adaptive and biologically guided radiotherapy at the population as well as individual patient-based levels. This requires the generation of models that inform response-based adaptations, individually optimized delivery and enable biological monitoring to provide decision support to clinicians. The goal is expanding to models that can drive the realization of personalized therapy for optimal outcomes. This position paper provides their propositions that describe how innovations in biology, physics, mathematics, and data science including AI could inform models and improve predictions. It consolidates the DREAM team's consensus on scientific priorities and organizational requirements. Scientifically, it stresses the need for rigorous, multifaceted model development, comprehensive validation and clinical applicability and significance. Organizationally, it reinforces the prerequisites of interdisciplinary research and collaboration between physicians, medical physicists, radiobiologists, and computational scientists throughout model development. Solely by a shared understanding of clinical needs, biological mechanisms, and computational methods, more informed models can be created. Future research environment and support must facilitate this integrative method of operation across multiple disciplines.

Essential parameters needed for a U-Net-based segmentation of individual bones on planning CT images in the head and neck region using limited datasets for radiotherapy application

Objective.The field of radiotherapy is highly marked by the lack of datasets even with the availability of public datasets. Our study uses a very limited dataset to provide insights on essential parameters needed to automatically and accurately segment individual bones on planning CT images of head and neck cancer patients.Approach.The study was conducted using 30 planning CT images of real patients acquired from 5 different cohorts. 15 cases from 4 cohorts were randomly selected as training and validation datasets while the remaining were used as test datasets. Four experimental sets were formulated to explore parameters such as background patch reduction, class-dependent augmentation and incorporation of a weight map on the loss function.Main results.Our best experimental scenario resulted in a mean Dice score of 0.93 ± 0.06 for other bones (skull, mandible, scapulae, clavicles, humeri and hyoid), 0.93 ± 0.02 for ribs and 0.88 ± 0.03 for vertebrae on 7 test cases from the same cohorts as the training datasets. We compared our proposed solution approach to a retrained nnU-Net and obtained comparable results for vertebral bones while outperforming in the correct identification of the left and right instances of ribs, scapulae, humeri and clavicles. Furthermore, we evaluated the generalization capability of our proposed model on a new cohort and the mean Dice score yielded 0.96 ± 0.10 for other bones, 0.95 ± 0.07 for ribs and 0.81 ± 0.19 for vertebrae on 8 test cases.Significance.With these insights, we are challenging the utilization of an automatic and accurate bone segmentation tool into the clinical routine of radiotherapy despite the limited training datasets.

End-to-end framework for automated collection of large multicentre radiotherapy datasets demonstrated in a Danish Breast Cancer Group cohort

Large Digital Imaging and Communications in Medicine (DICOM) datasets are key to support research and the development of machine learning technology in radiotherapy (RT). However, the tools for multi-centre data collection, curation and standardisation are not readily available. Automated batch DICOM export solutions were demonstrated for a multicentre setup. A Python solution, Collaborative DICOM analysis for RT (CORDIAL-RT) was developed for curation, standardisation, and analysis of the collected data. The setup was demonstrated in the DBCG RT-Nation study, where 86% (n = 7748) of treatments in the inclusion period were collected and quality assured, supporting the applicability of the end-to-end framework.

Prediction of Radiation-induced Lymphopenia following Exposure of the Thoracic Region and Associated Risk of Infections and Mortality

AIMS

Large blood volumes are irradiated when the heart is exposed to radiation. The mean heart dose (MHD) may be a good surrogate for circulating lymphocytes exposure. We investigated the association between MHD and radiation-induced lymphopenia and explored the impact of the end-of-radiation-therapy (EoRT) lymphocyte count on clinical outcomes.

MATERIALS AND METHODS

In total, 915 patients were analysed: 303 patients with breast cancer and 612 with intrathoracic tumours: oesophageal cancer (291), non-small cell lung cancer (265) and small cell lung cancer (56). Heart contours were generated using an interactive deep learning delineation process and an individual dose volume histogram for each heart was obtained. A dose volume histogram for the body was extracted from the clinical systems. We compared different models analysing the effect of heart dosimetry on the EoRT lymphocyte count using multivariable linear regression and assessed goodness of fit. We published interactive nomograms for the best models. The association of the degree of EoRT lymphopenia with clinical outcomes (overall survival, cancer treatment failure and infection) was investigated.

RESULTS

An increasing low dose bath to the body and MHD were associated with a low EoRT lymphocyte count. The best models for intrathoracic tumours included dosimetric parameters, age, gender, number of fractions, concomitant chemotherapy and pre-treatment lymphocyte count. Models for patients with breast cancer showed no improvement when adding dosimetric variables to the clinical predictors. EoRT lymphopenia grade ≥3 was associated with decreased survival and increased risk of infections among patients with intrathoracic tumours.

CONCLUSION

Among patients with intrathoracic tumours, radiation exposure to the heart contributes to lymphopenia and low levels of peripheral lymphocytes after radiotherapy are associated with worse clinical outcomes.

Driving quality improvement through better data: The story of New Zealand's radiation oncology collection

Aotearoa/New Zealand is one of the first nations in the world to develop a comprehensive, high-quality collection of radiation therapy data (the Radiation Oncology Collection, ROC) that is able to report on treatment delivery by health region, patient demographics and service provider. This has been guided by radiation therapy leaders, who have been instrumental in overseeing the establishment of clear and robust data definitions, a centralised database and outputs delivered via an online tool. In this paper, we detail the development of the ROC, provide examples of variation in practice identified from the ROC and how these changed over time, then consider the ramifications of the ROC in the wider context of cancer care quality improvement. In addition to a review of relevant literature, primary data were sourced from the ROC on radiation therapy provided nationally in New Zealand between 2017 and 2020. The total intervention rate, number of fractions and doses are reported for select cancers by way of examples of national variation in practice. Results from the ROC have highlighted areas of treatment variation and have prompted increased uptake of hypofractionation for curative prostate and breast cancer treatment and for palliation of bone metastases. Future development of the ROC will increase its use for quality improvement and ultimately link to a real time cancer services database.

RootPainter3D: Interactive-machine-learning enables rapid and accurate contouring for radiotherapy

PURPOSE

Organ-at-risk contouring is still a bottleneck in radiotherapy, with many deep learning methods falling short of promised results when evaluated on clinical data. We investigate the accuracy and time-savings resulting from the use of an interactive-machine-learning method for an organ-at-risk contouring task.

METHODS

We implement an open-source interactive-machine-learning software application that facilitates corrective-annotation for deep-learning generated contours on X-ray CT images. A trained-physician contoured 933 hearts using our software by delineating the first image, starting model training, and then correcting the model predictions for all subsequent images. These corrections were added into the training data, which was used for continuously training the assisting model. From the 933 hearts, the same physician also contoured the first 10 and last 10 in Eclipse (Varian) to enable comparison in terms of accuracy and duration.

RESULTS

We find strong agreement with manual delineations, with a dice score of 0.95. The annotations created using corrective-annotation also take less time to create as more images are annotated, resulting in substantial time savings compared to manual methods. After 923 images had been delineated, hearts took 2 min and 2 s to delineate on average, which includes time to evaluate the initial model prediction and assign the needed corrections, compared to 7 min and 1 s when delineating manually.

CONCLUSIONS

Our experiment demonstrates that interactive-machine-learning with corrective-annotation provides a fast and accessible way for non computer-scientists to train deep-learning models to segment their own structures of interest as part of routine clinical workflows.

Regression Analysis of Factors Based on Cluster Analysis of Acute Radiation Pneumonia due to Radiation Therapy for Lung Cancer

We conducted in this paper a regression analysis of factors associated with acute radiation pneumonia due to radiation therapy for lung cancer utilizing cluster analysis to explore the predictive effects of clinical and dosimetry factors on grade ≥2 radiation pneumonia due to radiation therapy for lung cancer and to further refine the effect of the ratio of the volume of the primary foci to the volume of the lung lobes in which they are located on radiation pneumonia, to refine the factors that are clinically effective in predicting the occurrence of grade ≥2 radiation pneumonia. This will provide a basis for better guiding lung cancer radiation therapy, reducing the occurrence of grade ≥2 radiation pneumonia, and improving the safety of radiotherapy. Based on the characteristics of the selected surveillance data, the experimental simulation of the factors of acute radiation pneumonia due to lung cancer radiation therapy was performed based on three signal detection methods using fuzzy mean clustering algorithm with drug names as the target and adverse drug reactions as the characteristics, and the drugs were classified into three categories. The method was then designed and used to determine the classification correctness evaluation function as the best signal detection method. The factor classification and risk feature identification of acute radiation pneumonia due to radiation therapy for lung cancer based on ADR were achieved by using cluster analysis and feature extraction techniques, which provided a referenceable method for establishing the factor classification mechanism of acute radiation pneumonia due to radiation therapy for lung cancer and a new idea for reuse of ADR surveillance report data resources.

Hematological toxicity in patients with solid malignant tumors treated with radiation - Temporal analysis, dose response and impact on survival

PURPOSE

To describe the kinetics of the peripheral blood components after radiotherapy, to examine radiation exposure vs. End-of-Radiation-Therapy (EoRT) counts and to associate the EoRT lymphocyte count with death and cancer treatment failure.

MATERIALS AND METHODS

Cohort study of patients who received curative intent radiotherapy for solid tumor diagnoses from 2009-2016 at Rigshospitalet, Copenhagen and had available 3D radiation exposure data. We illustrated peripheral blood count kinetics within 12 months before and after radiotherapy start and analyzed the impact of the irradiated body volume. We investigated overall survival and cancer treatment failure according to EoRT lymphopenia using Cox regression analyses.

RESULTS

We analyzed 4055 patients with both pre-treatment and EoRT platelet counts and 2318 patients who also had neutrophil and lymphocyte counts. Only the lymphocyte decline after radiotherapy start was clinically relevant and remained low one year after radiotherapy. The higher the volume of the body exposed to radiation, the lower the EoRT blood counts. Female gender (p < 0.001), number of fractions (p = 0.010), dose-volume (p < 0.001) and concomitant use of chemotherapy, particularly the platinum compounds (p < 0.001) were independently associated with a lower EoRT lymphocyte count. Patients with head and neck cancer had the lowest EoRT lymphocyte count. Patients with lymphopenia had a higher risk of death in the year after radiotherapy, compared with patients with no lymphopenia.

CONCLUSION

Radiation schemes with fewer fractions and radiation techniques allowing reduction of the volume of the body exposed to radiation could be expected to better preserve patients' immune function.

Towards a cancer mission in Horizon Europe: recommendations

A comprehensive translational cancer research approach focused on personalized and precision medicine, and covering the entire cancer research-care-prevention continuum has the potential to achieve in 2030 a 10-year cancer-specific survival for 75% of patients diagnosed in European Union (EU) member states with a well-developed healthcare system. Concerted actions across this continuum that spans from basic and preclinical research through clinical and prevention research to outcomes research, along with the establishment of interconnected high-quality infrastructures for translational research, clinical and prevention trials and outcomes research, will ensure that science-driven and social innovations benefit patients and individuals at risk across the EU. European infrastructures involving comprehensive cancer centres (CCCs) and CCC-like entities will provide researchers with access to the required critical mass of patients, biological materials and technological resources and can bridge research with healthcare systems. Here, we prioritize research areas to ensure a balanced research portfolio and provide recommendations for achieving key targets. Meeting these targets will require harmonization of EU and national priorities and policies, improved research coordination at the national, regional and EU level and increasingly efficient and flexible funding mechanisms. Long-term support by the EU and commitment of Member States to specialized schemes are also needed for the establishment and sustainability of trans-border infrastructures and networks. In addition to effectively engaging policymakers, all relevant stakeholders within the entire continuum should consensually inform policy through evidence-based advice.