Advancing Real-World Evaluation in Multi-Organ Hemodynamic Modelling

Provide a summary of the challenge purpose. This should include a general introduction in the topic from both a biomedical as well as from a technical point of view and clearly state the envisioned technical and/or biomedical impact of the challenge. High-resolution, large-coverage hemodynamic characterization is essential for understanding the pathophysiology of cardiovascular disease, portal hypertension, pulmonary circulation disorders, and renal vascular abnormalities. However, in real-world clinical practice, comprehensive acquisition of such information remains fundamentally constrained. Although four-dimensional (4D) flow MRI enables in vivo, time-resolved measurement of blood flow velocity, its prolonged acquisition time, high operational cost, and limitations in spatial resolution and noise robustness hinder routine and large-scale deployment. In contrast, vascular structural information derived from modalities such as magnetic resonance angiography (MRA) or computed tomography angiography (CTA), together with sparse hemodynamic boundary conditions—such as cross-sectional mean velocity or flow rate obtained from fast phase-contrast MRI or Doppler ultrasound—are far more accessible in clinical workflows. Accurately inferring high-resolution, spatiotemporal hemodynamic fields from such low-cost, easy-to-access inputs, and validating these estimates in vivo, remains a central open challenge in contemporary hemodynamic modelling. The CMRx4DFlow2027 challenge is designed to address this gap by explicitly focusing on computational fluid dynamics (CFD) and physics-informed neural networks (PINNs) as complementary methodological frameworks for hemodynamic inference. The challenge aims to systematically explore the capability of these approaches in reconstructing full 4D velocity fields under imaging-derived vascular geometries and increasingly sparse boundary condition constraints. Unlike prior evaluation paradigms that rely on numerical simulations or in vitro experiments as ground truth, this challenge positions 4D flow MRI as an in vivo reference for validation, emphasizing statistical consistency and physical plausibility rather than assuming noise-free absolute truth. To support robust evaluation at scale, the challenge leverages a large, multi-center 4D flow MRI dataset comprising more than 300 cases collected from more than five imaging centers and spanning three major vascular territories. The challenge is organized into three tasks: Task 1: Hemodynamic modelling with well-defined boundary conditions Develop methods capable of accurately reconstructing full spatiotemporal blood flow fields under reliable vascular geometry and sufficiently specified boundary conditions, establishing an upper-bound benchmark for in vivo hemodynamic inference. Task 2: Hemodynamic modelling with sparse boundary information Develop robust modeling approaches that can recover physiologically plausible blood flow fields from vascular geometry with limited and low-cost boundary measurements, reflecting realistic clinical acquisition constraints. Task 3: Unified hemodynamic modeling under variable boundary sparsity Develop a single, unified model that can stably infer full 4D hemodynamic fields across varying levels and patterns of sparse boundary information, enabling scalable and deployable hemodynamic modeling. Through this structured task design, CMRx4DFlow2027 aims to establish a realistic and clinically grounded benchmark for hemodynamic inference with in vivo validation. By systematically comparing traditional CFD and emerging physics-informed learning methods across varying levels of information availability and generalization complexity, the challenge seeks to advance methodological understanding and accelerate the translation of high-resolution hemodynamic modelling into both clinical and research settings. Building upon the organizational experience and strong community engagement established through prior events (CMRxMotion2022; CMRxRecon2023–2026), and in particular extending the scope of the CMRxRecon2026: 4D Flow Imaging Challenge, CMRx4DFlow2027 is designed to further expand the application of advanced hemodynamic modeling in real-world clinical contexts. This initiative specifically aims to enable institutions that lack direct access to 4D Flow MRI acquisition to actively participate in hemodynamic research through modeling, inference, and validation frameworks. The organizational structure, challenge framework, and evaluation protocols of CMRx4DFlow2027 will be built upon the proven foundations of previous challenges (CMRxMotion2022; CMRxRecon2023–2026), ensuring continuity in governance, technical rigor, and community standards. Across these events, we have established a strong track record, collectively attracting more than 500 participating teams worldwide, which provides a solid basis for the successful execution of this challenge. By lowering technical and infrastructural barriers to advanced hemodynamic analysis, CMRx4DFlow2027 aims to broaden access to state-of-the-art vascular modeling, stimulate methodological innovation, and bridge existing gaps in hemodynamic research capabilities across institutions and regions.