Bioelectromagnetism empowered by virtual models

This thesis advances a multiscale digital-twin framework for bioelectromagnetics that links image-based modelling, numerical dosimetry, and targeted experiments across cellular, preclinical, and human scales. The overarching goal is to quantify electromagnetic (EM) interactions under realistic conditions and to use predictive simulations to design exposure systems, interpret biological effects, and support safety assessment. The work is organized in three parts that progressively scale from in vitro to in vivo and human-level applications, maintaining methodological continuity throughout. In Part I – Virtual models for in vitro biological systems, the focus is on the microscale, where cell-level mechanisms of EM interaction are explored through detailed digital twins and exposure calibration. High- fidelity 3D digital twins of stem cells (iNSCs and MSCs) were reconstructed from confocal microscopy, explicitly resolving subcellular compartments (plasma membrane, ER, mito- chondria). Microdosimetric analyses show how microsecond pulsed electric fields (μsPEFs) can be tuned, at fixed intensity, via pulse duration to modulate membrane electroporation and intracellular calcium targets, informing the RISEUP stimulation strategy. A numerical calibration method then harmonizes the E-field delivered by different electrode technolo- gies, enabling comparable stimulation conditions across setups and validating equivalence through cell-level dosimetry. Finally, two RF exposure systems (i.e. s-CPW and LoREC) were engineered and validated for FR1 (700 MHz, 3.5 GHz) real-time electrophysiological experiments, combining S-parameter, SAR, and temperature characterizations to guarantee low-noise, traceable RF exposure during electrophysiology. In Part II – Virtual models for in vivo studies, the investigation moves to the tissue and organ scales, addressing preclinical modeling of spinal cord and vertebral targets for therapeutic electroporation. A rat spinal cord-injury digital twin reproduced laminectomy, lesion evolution, and EPB implantation to prospectively assess target E-fields and neurofunctional safety, coupling quasi-static dosimetry with axonal electrophysiology. The section culminates in a large-animal ovine study of electroporation for vertebral metastases using novel coaxial bipolar electrodes, integrating simulations with clinical, imaging, and histology endpoints to quantify ablation dose, temperature rise, and neural safety margins. Together, these studies demonstrate translational modelling that de-risks protocols before and during experimentation. Finally, in Part III – Virtual Models for Human Anatomy and Exposure Assessment attention shifts to the macroscopic level, where anatomically realistic human models are used to study EMF exposure under 5G and neurostimulation scenarios. At high frequencies (FR2), an anatomically realistic multilayer skin model with undulated interfaces reveals morphology- driven near-surface field nonuniformities and conservative local peaks compared with ideal planar slabs, refining mm-wave dosimetry. An enhanced whole-body female model (“Venus”) with upright breast anatomy quantifies posture- and morphology-dependent differences in SAR under 2.45 GHz and 24 GHz plane waves and evaluates occupational exposure during TMS, underscoring the need for anatomically precise models in regulatory and clinical contexts. Across scales and frequencies, the thesis shows that anatomically and technologically realistic digital twins can (i) predict exposure metrics inaccessible to measurement, (ii) standardize stimulation across devices and experiments, and (iii) improve the reliability and translational value of EMF research, from in vitro protocol design to preclinical therapy planning and human exposure evaluation.