A biatrial digital twin integrating electrophysiology, mechanics, and circulation: from physiology to atrial fibrillation

This paper presents a patient-specific, multiscale biatrial digital twin that integrates electrophysiology, mechanics, and circulation to simulate and analyze the physiological and pathological (atrial fibrillation) mechanisms driving atrial function and hemodynamic changes.

The Gist

Imagine your heart's upper chambers (the atria) not just as passive buckets waiting to be filled, but as active, rhythmic squeeze-boxes that help push blood down into the main pumping chambers. When these squeeze-boxes get out of sync, as they do in a condition called Atrial Fibrillation (AF), the whole system loses efficiency.

This paper introduces a super-complex, virtual "Digital Twin" of a human heart's upper chambers. Think of it as a high-definition video game simulation that is so realistic, it doesn't just look like a heart; it acts like one.

Here is a breakdown of what the researchers built and why it matters, using simple analogies:

1. The "Digital Twin": A Heart in a Computer

Usually, scientists study the heart's electricity (the spark) or its mechanics (the squeeze) separately. This team built a model that combines everything into one package:

• The Spark (Electricity): How the electrical signal travels through the heart tissue.
• The Squeeze (Mechanics): How the muscle physically contracts and stretches.
• The Flow (Circulation): How blood moves through the whole body, not just the heart.

The Analogy: Imagine a car engine. Most models only look at the spark plugs (electricity) or the pistons moving (mechanics). This model connects the spark plugs to the pistons, the pistons to the wheels, and the wheels to the road, all in one simulation. If you change the spark, the model instantly shows you how the car's speed and fuel efficiency change.

2. Building the Model: The "Lego" Heart

To make this twin, they took a real CT scan of a patient's heart and turned it into a 3D digital map.

• The Mesh: They covered this map with millions of tiny digital "pixels" (called a mesh).
• The Fibers: Just like wood grain, heart muscle has a specific direction it pulls. They programmed the model to know exactly which way the fibers point in every tiny corner of the heart.
• The Calibration: This is the most important part. They tweaked the model's settings until it behaved exactly like a healthy human heart. It had to pump the right amount of blood, create the right pressure, and squeeze at the right time.

The Analogy: Think of tuning a guitar. If the strings are too loose or too tight, the note is wrong. The researchers "tuned" their digital heart until it played the perfect "healthy heart" note. Only then could they trust it to play "sick" notes later.

3. The "Figure-Eight" Loop: The Gold Standard

One of the hardest things to get right in heart simulations is the Pressure-Volume Loop.

• What it is: A graph that shows how much pressure the heart makes as it changes size.
• The Goal: In a healthy heart, this graph looks like a figure-eight (like the number 8 on its side). It has two loops: one for the "squeezing" phase and one for the "filling" phase.
• The Achievement: Many previous computer models produced messy, squiggly lines. This model successfully recreated the perfect figure-eight, proving it captures the true physics of how the heart works.

4. Testing the "What Ifs": Sensitivity Analysis

Once the healthy twin was built, the researchers started playing "What If?" games. They tweaked specific knobs to see what happened:

• What if the muscle gets stiffer? (Like a rubber band that has lost its stretch). The heart couldn't fill up as much.
• What if the squeeze gets weaker? The heart couldn't push blood out effectively.
• What if the valves get stuck? The flow changed dramatically.

This helps doctors understand which parts of the heart are most critical. It's like testing a car by slightly loosening a bolt or reducing tire pressure to see how much it affects the ride before the car actually breaks.

5. Simulating the Crash: Atrial Fibrillation (AF)

Finally, they used the model to simulate Atrial Fibrillation, a condition where the heart's electrical signals get chaotic, like a crowd of people shouting instead of marching in step.

• The Trigger: They introduced a "rogue" electrical signal (an ectopic beat) near the veins where AF often starts.
• The Result: The orderly electrical wave broke apart into chaotic swirls (rotors).
• The Consequence:
  • The Squeeze Stops: The heart muscle stopped squeezing in a coordinated way. It just quivered.
  • The Pump Fails: Because the "booster pump" (the atria squeezing) stopped working, the heart's overall output dropped by 20%.
  • The Graph Collapses: The beautiful figure-eight loop disappeared and turned into a single, weak loop. The "booster" part of the graph vanished completely.

Why This Matters

This isn't just a cool computer graphic. It's a predictive tool.

• For Doctors: It could help them understand why a specific patient's heart is failing and test treatments (like drugs or ablation) in the computer before trying them on the patient.
• For Science: It proves that you can't understand the heart's pumping power without understanding its electricity, and vice versa. They are inextricably linked.

In Summary:
The researchers built a virtual heart that is so accurate it mimics the real thing's electrical sparks, muscle squeezes, and blood flow. They showed that when the electricity goes haywire (AF), the mechanical squeeze collapses, and the heart's efficiency plummets. This "Digital Twin" gives us a powerful new way to study heart disease and potentially save lives by testing solutions in the virtual world first.

Technical Summary

1. Problem Statement

Atrial electromechanics is critical for regulating ventricular filling and global hemodynamics, yet modeling it consistently across scales remains a significant challenge. Existing computational models often suffer from specific limitations:

• Electrophysiology (EP) models focus on arrhythmia mechanisms but lack mechanical and hemodynamic coupling.
• Computational Fluid Dynamics (CFD) models study hemodynamics but ignore tissue mechanics and electrical activation.
• Electromechanical models often fail to reproduce realistic atrial pressure–volume (PV) loops (specifically the "figure-eight" morphology) or lack a closed-loop circulatory system, leading to unrealistic boundary conditions.
• There is a lack of validated frameworks that can simultaneously reproduce regional activation times, chamber volumes, ejection fractions (EF), and PV loop characteristics in both healthy and pathological (Atrial Fibrillation - AF) states.

2. Methodology

The authors developed a multiscale, multiphysics biatrial digital twin that couples a 3D electromechanical model of the atria with a 0D closed-loop circulatory model.

A. Anatomical and Geometric Modeling

• Data Source: Patient-specific geometry reconstructed from contrast-enhanced CT scans of a 56-year-old male.
• Meshing: A two-mesh strategy was employed to balance accuracy and cost:
  • Fine Mesh (Electrophysiology): ~16.7 million tetrahedra (avg edge 0.267 mm) to resolve steep electrical gradients.
  • Coarse Mesh (Mechanics): ~0.7 million tetrahedra (avg edge 0.799 mm) for solid mechanics.
• Heterogeneity: The atria were subdivided into 48 anatomical regions. Wall thickness was explicitly varied based on imaging evidence. Myocardial fiber orientations were assigned using a rule-based approach (longitudinal in walls/bundles, circumferential in appendages).

B. Mathematical Framework

The model integrates three physical domains:

1. Electrophysiology (3D): Modeled using the Monodomain equation with the Courtemanche-Ramirez-Nattel (CRN) human atrial ionic model.
   • Heterogeneity: Regional variations in conduction (diffusion tensor) and ionic conductances (e.g., $I_{K1}$, $I_{to}$, $I_{CaL}$) were applied to 11 conduction regions and 9 ionic regions.
2. Solid Mechanics (3D): Governed by the balance of linear momentum.
   • Passive Mechanics: Transversely isotropic Holzapfel-Ogden hyperelastic law.
   • Active Mechanics: Land et al. excitation-contraction model driven by intracellular calcium ($[Ca^{2+}]_i$) and fiber stretch.
   • Boundary Conditions: Robin (spring) conditions applied to venous openings, valve annuli, and pericardium to allow physiological motion; Neumann (pressure) conditions applied to endocardium coupled to the 0D circulation.
3. Hemodynamics (0D): A closed-loop circulatory model representing systemic and pulmonary circuits, as well as all four cardiac chambers.
   • Coupling: The 3D atrial volumes act as Lagrange multipliers, ensuring consistency between 3D deformation and 0D hemodynamic pressures/flows.

C. Calibration and Simulation

• Calibration: Parameters (diffusion coefficients, active tension $T_{ref}$, elastances, resistances) were tuned to match clinical biomarkers: regional Local Activation Times (LATs), atrial volumes ($V_{min}, V_{max}, V_{preAC}$), and phasic Ejection Fractions (Reservoir, Conduit, Booster).
• Sensitivity Analysis: A local sensitivity analysis was performed on 8 key parameters (active tension, passive stiffness, pericardial stiffness, ventricular elastances) to quantify their impact on atrial function.
• AF Scenario: Persistent AF was induced via an S1–S2 protocol (premature ectopic stimulus near pulmonary veins) combined with electrical remodeling (scaling of ionic currents to mimic chronic AF).

3. Key Contributions

1. Fully Coupled Framework: Integration of 3D biatrial electrophysiology, solid mechanics, and a 0D closed-loop circulation in a single, strongly coupled digital twin.
2. Realistic PV Loops: The model successfully reproduces the characteristic figure-eight atrial pressure–volume loops (distinct A-loop for active contraction and V-loop for passive filling), a feature often missing in previous studies.
3. Systematic Sensitivity Analysis: Quantification of how specific mechanical and hemodynamic parameters influence atrial functional biomarkers, distinguishing the roles of active contractility vs. passive stiffness vs. ventricular loading.
4. Pathological Propagation: Demonstration of how electrical remodeling in AF propagates across scales, leading to mechanical failure (loss of booster pump) and hemodynamic decline (reduced Cardiac Output).

4. Key Results

Physiological Calibration

• Activation: Calibrated regional diffusion coefficients successfully reproduced physiological LATs within experimental ranges. Homogeneous diffusion resulted in significant delays (>40–70 ms).
• Mechanics: The calibrated model matched reference atrial volumes and phasic EFs (Reservoir, Conduit, Booster) for both left and right atria.
• PV Loops: The calibrated simulation produced clear figure-eight loops. Uncalibrated simulations failed to separate the A-loop and V-loop, producing unrealistic morphologies.

Sensitivity Analysis Findings

• Active Tension ($T_{ref}$): Primarily modulates the booster pump (increasing EF and reducing end-systolic volume) with minimal effect on reservoir/conduit phases.
• Passive Stiffness ($a, a_f$): Primarily affects the reservoir phase. Increased stiffness reduces maximum volume and narrows the V-loop, mimicking impaired passive filling (e.g., fibrosis).
• Pericardial Stiffness ($k_{peri}$): Constrains expansion, reducing both reservoir volume and the work done during the booster phase (reduced A-loop area).
• Ventricular Elastance:
  • Increased active ventricular elastance reduced atrial work (smaller A-loop) as the ventricle became more efficient at filling.
  • Increased passive ventricular elastance (stiff ventricle) forced the atria to compensate by increasing active work (larger A-loop) to maintain Cardiac Output.

Atrial Fibrillation (AF) Application

• Induction: An ectopic stimulus near the pulmonary veins triggered unidirectional conduction block and reentry, forming two stable rotors in the left atrium.
• Electromechanical Impact:
  • Calcium/Tension: Spatially averaged intracellular calcium peak decreased by 55%, and active tension decreased by 74%.
  • Mechanics: The "booster" phase disappeared. Atrial volume traces showed no late diastolic decrease.
  • PV Loops: The figure-eight morphology collapsed into a single V-loop, indicating a total loss of effective atrial contraction.
• Hemodynamics:
  • Loss of the "atrial kick" (late diastolic flow) at the mitral and tricuspid valves.
  • Cardiac Output (CO) Reduction: A 20.4% reduction in CO was observed, consistent with clinical estimates (20–25%).

5. Significance

This work provides a robust, validated platform for investigating the multiscale consequences of atrial pathology. By successfully capturing the transition from electrical disorganization to mechanical failure and global hemodynamic compromise, the framework:

• Offers a tool to test personalized therapies (e.g., ablation strategies, drug effects) in silico.
• Clarifies the mechanistic link between atrial stiffness, ventricular loading, and atrial contribution to cardiac output.
• Demonstrates that accurate modeling of atrial function requires the simultaneous consideration of electrophysiology, tissue mechanics, and closed-loop hemodynamics, particularly for understanding the hemodynamic impact of AF.

Limitations noted by authors: The model currently lacks mechano-electric feedback (stretch-induced electrical changes), uses a single anatomy (not fully patient-specific for validation), and assumes regular ventricular rhythm during AF (ignoring irregular ventricular response). Future work will address these gaps.