Digital twins represent an emerging frontier towards enabling scalable pathways for personalized assessment of medical decisions, device operation, and etiology/diagnosis of diseases. Fundamentally, a digital twin must distinguish itself from a traditional computer simulation through robust, and often dynamic, integration with data. Yet, the inherently multi-modal nature of medical data can often pose specific challenges for development of digital twins, especially when interfacing the data with high-resolution physiology simulations for a specific application. Addressing this can require custom computational, data assimilation and data-driven modelling approaches, possibly going beyond traditional avenues in the field. Here, we describe frameworks and pipelines for incorporating mixed-modal data from ultrasound, CT, angiography, perfusion imaging, clinical records, and treatment/outcomes to drive a custom computational pipeline to develop patient-specific digital twins. We will demonstrate the workflow in the context of applications in stroke. Specifically, we showcase examples of digital twin-based investigations in two specific stroke applications. Our first example caters digital twin development towards standard-of-care stroke procedure for improving diagnosis of Embolic Strokes of Undetermined Sources. Our second example involves developing tools for pre-surgical assessment of stroke risks in heart failure patients who are selected for treatment using a Left Ventricular Assist Device. These examples illustrate approaches for incorporating mixed-modality data: challenges and opportunities associated with incomplete data; small scale modelling loops to translate data into usable inter-connects with simulations; and potential for AI/ML frameworks for acceleration. We will also outline how the resulting digital twin models can be used to enable key personalized insights in terms of stroke applications. We will close by sharing our ongoing efforts on releasing some of the underlying tools and data openly with the broader scientific community.