The timing and workflow of VITAL is such that most of our activities in the first 2-3 years of the project are of a scientific-technical nature, relating to model development and the development of (AI-enforced) pipelines and workflows to individualise the models, converting the multi-scale and multi-organ models into human virtual twins. We made very good progress on multiple facets:
1) Model development and model personalisation:
- multi-scale cardiac model development, with implementation and verification of a micro-structurally informed and multi-scale cardiac growth and remodelling model in three different software platforms (Abaqus, Alya and pyGrowth); models are being extended with mechano-chemical coupling, linking a logic-gated intracellular signalling network with cardiac growth and remodelling;
- multi-scale arterial and venous models, with coupling of strain-energy and constituent based constitutive models of arterial tissue to tube-laws used in 1D models and development of advanced arterial growth- and remodelling algorithms.
- development and calibration of a generic sympathetic neuron model which will allow accounting for nervous control of the heart and vasculature (and other organs); development of sino-atrial node cell model and a sympathetic control model to regulate microvascular constriction and regulation;
- modelling hormonal control, with the development of a bond graph CellML model for the human vascular system and organ-level vascular models that link down to exchange mechanisms for hormones and other blood solutes at the tissue capillary level, and bond graph models for the release and targeting of relevant hormones;
- developing a bond-graph cardiovascular system model coupled to the kidney for volume control with intracellular homeostasis for kidney epithelial cells;
- automated AI-assisted generation of multi-scale organ models through ApiNATOMY, an open-source, ontology-anchored knowledge management infrastructure in which every anatomical conduit is a nested, layered object. The kidney is being used as the first application of this new framework.
- integration and coupling of 0D/1D/3D models and software codes, with a portfolio of benchmark problems for cardiac and arterial (growth and remodelling) models, and pipelines for model calibration, uncertainty quantification and model reduction;
- virtual cohort data generation, with a.o. the generation of age- and sex-specific in silico pulse waves and extracted indices matched to age- and sex-specific in vivo data from UK Biobank.
- development and validation of PK/PD models with several drugs relevant for VITAL and a strategy to couple the PK/PD model to the VITAL multiscale framework;
- the CALAMARI wearable sensor system for 24hr hemodynamic monitoring has been tested and released. The documentation and administration for use in clinical studies has been prepared, and approval is pending. We have streamlined the data processing workflow for feeding into the VITAL models, with a portfolio of algorithms for biomarker extraction from pulse wave signals;
- the development of model development and personalization pipelines with
o automation of the process of combining CellML modules into a system model, then doing parameter identification to calibrate to clinical data;
o integration of multicentric and multimodal CMR, ECG and echocardiography data to build patient-specific 3D–0D electromechanical models;
o semi-automated pipelines for personalising models
To date, over reporting period 1 and 2, VITAL and VITAL-related results have been published in about journal publications, and over 30 presentations at international conferences.
During reporting period 2, we also organized 2 successful trainings schools in Delft, Netherlands (March 2025) and Auckland, New Zealand (February 2026).
2) Model application and validation
Dedicated clinical studies within which our digital twin models will be used are yet to start. In preparation, we worked on defining Minimum Viable Products for the VITAL use cases, connecting the computational modelling workflows to the clinical studies, establishing concrete validation targets for assessing whether model-derived biomarkers and predictions provide information on disease state or treatment response beyond the routinely available clinical measurements. Models are also being tested on already available data. It was demonstrated in a cohort of 15 patients with Left branch Bundle Block, that patient-specific cardiac electromechanical models reproduced baseline cardiac function, and that personalised models can be subsequently used to evaluate treatment response, providing an initial in-silico evaluation of alternative pacing strategies within the same patient-specific cohort.