Objectives

Complex conditions such as cardiovascular disorders and cancer require the integration of diverse patient data to support accurate diagnosis and treatment decisions. TWIN-X aims to improve diagnostic accuracy through the development of explainable and trustworthy digital twins for oncology and cardiology. By creating unified, clinically meaningful representations of patients, the project will establish new foundations for AI-driven, patient-centred healthcare.

Scroll down for more information and click on the box

• Move digital twins from fragile prototypes to transparent, modular and transferable patient models.

  Despite their growing potential, current digital twin approaches still face significant challenges. Many solutions struggle with scalability, interoperability, and clinical validation. Concerns regarding data quality, privacy, and ethical considerations further limit their adoption in clinical practice. In addition, medical AI systems often fail to capture the full clinical context and to meaningfully connect information from diverse data sources.

  TWIN-X addresses these challenges by creating a reusable framework that transforms unstructured clinical reports, medical imaging, and digital pathology data into integrated patient representations. By fascilitating the seamless integration of diverse clinical data across languages and healthcare settings, TWIN-X will lay the foundation for explainable, and clinically applicable AI systems.

• Deliver generative AI models made in Europe that are robust, trustworthy and ethical.

  TWIN-X develops methods that make AI systems more transparent, reliable, and fair. The project creates frameworks that help explain how medical AI reaches its conclusions, enabling clinicians and researchers to better understand, assess, and trust AI-generated insights. In addition, TWIN-X establishes methodologies to evaluate AI systems in terms of their limitations, alignment with ethical values, explainability, trustworthiness, and potential biases. By advancing responsible AI development, the project contributes to build a sovereign European research ecosystem that supports the safe and effective integration of innovative digital technologies into healthcare.

• Develop an AI model that imposes clinical semantics prior to modelling.

  TWIN-X relies on a so-called Structure-First Approach that organises patient information in a hierarchical way. Before being processed by AI models, patient data are structured according to clinically meaningful categories such as organ systems and disease pathways. The information is also organised to capture the progression of health events over time and the status of clinical findings, including whether they have been confirmed or ruled out. By establishing these structured patient representations, TWIN-X enables a more comprehensive analysis of medical information and stronger connections between different data sources, ultimately supporting more reliable clinical decision-making.

• Demonstrate the value of integrated patient representations in clinical use-cases.

  TWIN-X will test and validate the potential of digital twins in two clinical domains: oncology and cardiology. By leveraging AI-driven patient representations, the project aims to enhance the understanding of liver, lung, and upper gastrointestinal cancers, as well as improve cardiovascular risk assessment and disease management. If successful, these pilot studies will demonstrate the value of generative AI models in supporting clinical decision-making and pave the way for their application to a broader range of cancer and cardiovascular conditions.

• Create an open toolkit that enables researchers to use, evaluate, and build upon the TWIN-X AI technologies in a transparent and reliable way.

  Guided by the principles of findability, interoperability, and reusability, TWIN-X is developing an open-access toolkit that makes advanced AI technologies available to other researchers. The project will establish a scientific infrastructure for training, testing, and sharing trustworthy generative models in the field of biomedicine. By supporting researchers to build on existing technologies and adapt them to their specific needs, TWIN-X promotes a collaborative approach to the responsible development of AI.