Medical and industrial imaging increasingly serve as the foundation for digital twins — virtual, patient- or process-specific replicas that support diagnosis, monitoring, design, and decision-making. This minisymposium brings together mathematicians, computational scientists, and engineers working at the intersection of imaging, mathematical modeling, and scientific machine learning to build imaging-based digital twins with real-world impact, spanning biomedical and industrial applications. We will discuss algorithmic advances for digital twins and for solving the underlying governing equations, covering both classical numerical methods and emerging machine-learning-based techniques. Talks will address model-driven, data-driven, and hybrid strategies that combine physics-based simulation with learned representations, together with methods for uncertainty quantification, model-order reduction, and real-time simulation needed to make digital twins deployable in practice. Particular attention will be devoted to the translation of these methods beyond the laboratory: validation against real data, integration into clinical and industrial workflows, and the challenges of scaling personalized or process-specific models to cohorts or production settings. By connecting biomedical applications (e.g., cardiovascular and organ-level digital twins) with broader industrial and sustainability-oriented use cases, the session aims to foster cross-fertilization between academia, industry, and end users, in line with M2P's mission of turning mathematical and computational advances into deployable, reliable quantitative tools.