Human Digital Twin

Engineering problem

The Human Digital Twin is a computational physiology workbench for studying how interacting physiological systems behave across spatial and temporal scales. It connects whole-body state, systemic circulation, pharmacokinetics, spatial tissue transport, microvascular flow and exploratory immune-response models through a reusable simulation architecture.

The aim is not to reproduce a particular person. It is to make coupled physiological models inspectable, configurable and testable in one environment.

Computational architecture

The framework separates model state, numerical solvers, experiment configuration and visualization. Systemic compartments exchange state with optional spatial tissue samples, while the interface provides anatomical inspection, time-series plots, experiment setup and simulation playback.

Pulse Physiology Engine 4.4.0 can supply quantities such as cardiac output, heart rate, blood pressure, organ perfusion, blood volume, ventilation and oxygen-related variables. Rest, exercise and recovery schedules can then provide time-dependent parameters and boundary conditions for downstream models. Native Pulse integration runs locally; the deployed browser workbench does not execute Pulse natively on Vercel.

Mathematical foundations

Physiologically based pharmacokinetics

A representative organ mass balance is

$$\frac{dM_i}{dt}=Q_i(C_{in,i}-C_{out,i})-CL_iC_i+E_i.$$

Here, $M_i$ is drug mass in organ $i$, $Q_i$ is organ blood flow, $C_{in,i}$ and $C_{out,i}$ are inlet and outlet concentrations, $CL_i$ is a clearance term, $C_i$ is the organ concentration and $E_i$ represents exchange with another modeled domain. The implementation supports 13 well-mixed organ representations with arterial, venous and portal circulation, tissue partitioning, clearance and closed-circulation mass accounting.

Spatial tissue transport

Spatial samples use a finite-volume form of the transport equation

$$\frac{\partial C}{\partial t}=\nabla\cdot(D\nabla C)-R(C)+S.$$

Cell-centered Cartesian control volumes represent diffusion, permeability, partition-mediated exchange, optional binding, degradation and intracellular transport. Numerical tissue domains exchange material conservatively with their associated systemic compartments.

Microvascular flow

Network segments use a simplified Poiseuille relationship,

$$Q=\frac{\pi r^4\Delta P}{8\mu L},$$

to relate pressure difference, vessel radius, viscosity and segment length. Conductance-based bifurcating networks provide pressure-driven flow and vascular-to-tissue exchange. This is a reduced network model, not fully resolved cardiovascular CFD.

Multiscale coupling

Systemic circulation and spatial samples advance on coordinated time schedules. Physiological states provide boundary conditions, while conservative exchange transfers mass between well-mixed and spatial domains. Synchronization keeps systemic concentrations, tissue fields and model-specific state changes consistent over time.

Exploratory immune and hematological models

The workbench includes blood-cell turnover and population-balance differential equations for neutrophil and monocyte dynamics, simplified lymphocyte trafficking and activation, local spatial immune-cell behavior and illustrative IL-6 stimulus experiments. These are exploratory computational models rather than clinically validated immune physiology.

Implemented capabilities

• Interactive anatomy with organ, vessel, skeleton and tissue inspection
• Physiological state variables and numerical time-series plots
• Spatial transport fields and microvascular networks
• Experiment configuration and simulation playback
• Reusable model and solver interfaces
• Rest, exercise and recovery schedules

Verification and limitations

Automated tests, controlled numerical experiments, mass-conservation checks and model-specific accounting are used to examine implementation behavior. Verification focuses on whether equations and couplings behave as intended under known conditions; it does not establish clinical validity.

The framework is not a clinically validated representation of an individual human. Physiological detail varies by subsystem and several immune, vascular and tissue models are deliberately reduced to support computational experiments.

Engineering applications

The project demonstrates work in computational physiology, PBPK modeling, spatial drug transport, multiscale coupling, numerical model integration and scientific software architecture.

Technical resources

The live workbench is the public application. Its source repository is private.