RODIN: Radiation Oncology Data Interoperability Network for an AI and Digital Twins Driven Future

 

Chairs:

• Laia Humbert-Vidan

• Charles Mayo

 

Description

Radiation oncology is a highly data-intensive field, where ability to collect, aggregate and share clinical information is critical for advancing patient care and research. Rapid growth and democratization of analytics technologies including predictive AI,  generative AI and digital twins are transforming our ability to use data to improve patient care and accelerate discovery. Reaching the potential of these technologies is tied to the volume, quality and scope of the data feeding them.

Construction of large, comprehensive, interoperable “real world” datasets can support advanced analytic techniques that inform evidence-based decision making and enhance patient management, potentially improving outcomes through greater access to diverse and representative data. However, real-world data is often inconsistently coded, with limited standardisation both across institutions and even among practitioners within the same institution. These inconsistencies hinder the potential for automated learning from the vast volumes of data routinely archived in clinical systems and cancer care databases.

Combining AI and other technologies along with standardisation based approaches  is essential to enable the accurate and automated extraction,  use and sharing of clinically relevant data from existing systems. By aligning with the FAIR principles (Findability, Accessibility, Interoperability, and Reusability) standardisation not only facilitates more efficient data aggregation and sharing across institutions but also supports scalable analytic methods, enhancing the utility of clinical datasets for both research and patient care. The combination will result in automations enabling more time spent using data and  less time aggregation and curation.

Motivation

This workshop aims to convene experts who are actively combining work with real-world radiation oncology data, AI and contributing to data standardisation or interoperability initiatives. By bringing together clinicians, medical physicists, data scientists, IT specialists, data managers and other professionals with practical experience in these areas, participants will collaboratively work on actionable steps towards  interoperable data ecosystems to improve the landscape for  AI development and digital twin models in radiation oncology.

The result of convening this community of data experts using AI will be increasing the scale and richness of datasets.  It will spur development of artificial intelligence methods and patient-specific digital twin models designed to simulate treatment response and toxicity risk. Increasing aggregation of real world evidence data sets used with federated analytics approaches based on standardized ontologies  will enable impactful discoveries, including identifying patterns between treatment effectiveness and toxicity, expanding knowledge on rare cancers, benchmarking clinical performance and radiotherapy plan quality, as well as facilitating technology evaluation.

 

Expected outcomes:

• Develop a shared understanding of ongoing initiatives working toward interoperable data frameworks in radiation oncology, including interoperability and limitations of  current standards, vendor system infrastructures, and implementation efforts across institutions and groups.

• Identify key limitations, requirements, risks, and feasible approaches addressing these limitations for data access and sharing in multi-institutional settings, and for the development and implementation of interoperable frameworks, including strategies to transform existing datasets into federated research infrastructures.

• Define the scope and outline of a roadmap article on recommendations for improving data infrastructure to support multi-institutional research and clinical collaboration in radiation oncology.  The topics in the article will include  combined use of distributed learning, ontologies, data standardisation, technologies including AI and interoperability frameworks.

• Initiate the development of a collaborative program of pilot projects  to test interoperable radiation oncology data infrastructures, including federated learning and distributed analytics across centres. Use preliminary data from the collaboration to establish the basis for a joint multi-institutional funding proposal (e.g., EU COST Action or similar international initiatives) to expand availability of tools, methods and federated data sets to clinics and researchers.