In WP2, available modelling components in the heart-brain axis were inventorized and modelling pipelines built for the two use cases in the project. Corresponding verification and validation activities to build the credibility were designed. Advances in the individual models are: the 3D fluid simulations of the left atria, coupled with (1) an 0D model of the coagulation cascade to predict thrombus formation; (2) an 0D-1D model of the cardiovascular circulation and (3) 3D model of cerebral blood flow; the development of models of thrombus fracture and thrombus constitutive models to provide insights into the thrombectomy modelling pipeline; and innovative constitutive multi-scale models of the vessel wall remodelling that will provide in-silico indices for clinical decisions related to the HS use case.
WP3 performed a comprehensive review of existing clinical knowledge on thrombus aetiology, imaging, experiments, and development and testing of knowledge- and data-driven models. We also worked on models aimed at determining stroke aetiology using multimodal parameters. The methodology to execute finite element (FE) models for the three mechanical thrombectomy procedures—stent retriever, aspiration, and combined approaches—was developed. Previous experimental trials were replicated using FE simulations of the corresponding thrombectomy procedures. A verification study and convergence tests were done for the coupled 0D blood flow ↔ 3D perfusion model.
WP4 aims to develop data-driven disease management models to predict disease progression before and after intracranial aneurysm (IA) rupture and patient outcome according to treatments. Three generations of models will be developed (1) The 1st generation Bayesian Network (BN) model will predict disability of patients diagnosed with unruptured IA. (2) the 2nd generation BN model will associate clinical, radiological and genomics factors to new factors representing injury, repair and vessel wall conditions. (3) the 3rd generation BN model will integrate information associated with IA rupture and subarachnoid hemorrhage (SAH), and IA treatment to provide a fully integrated disease model of the disease management.
WP5 focuses on developing and validating five subject-specific Digital Twin Heart (DTH) simulation pipelines for stroke and aneurysm treatment. Key achievements include: (1) Full development strategies for three of the five core pipelines have been completed. (2) Significant progress on automated pipeline for extracting patient-specific vessel geometries from CTA scans. The first running simulations for the three main thrombectomy treatments have been accomplished. (3) successful demonstration of a novel co-registration method to fuse high-resolution images with 3D scans. In parallel, a physics-based path-planning algorithm for coil deployment was developed, and uncertainty quantification campaigns for both coiling and flow diverter simulations have been planned and initiated. (4) To support subject-specific modelling, a statistical algorithm for selecting representative patient boundary conditions has been created. A BN model is now under development to quantify biological uncertainty, supported by the curation of clinical data, the preparation of a reference population dataset, and a new collaboration to integrate genetic data.
In WP6 we want to generate proof of clinical value of the personalised IS Treatment Selection DTH (starting M49). We also determine the adaptation, perception and change in behaviours of patients and clinical professionals who utilize digital twins. We have investigated and conceptualized the role of emotions (e.g. shame) in designing the interface between a user and a simulation predicting on health. And we studied the conceptualisation of agency in the context of AI-based technology and introduce an alternative conceptual framework offering a more nuanced account of the forms and modalities through which human agency can be represented and mediated by a Digital Twin.
In WP7 we established procedures related to development and management of GEMINI model code, and a preliminary plan of VVUQ activities for computational models. We set up a Github repository, assembled a catalogue of software components and models, and published three technical deliverables which also serve as tutorials for Project participants. We created a Software Delivery Team and had several tutorials explaining the use of HPC resources, optimization of computational models, and automated VVUQ procedures.
