Real-time GPU-accelerated coupled cardiac system: Integrating bidirectional interactions between living optogenetic monolayers and computational simulations

Figures

• 

  Fig 1. Conceptual overview of the closed-loop coupled cardiac system. (A) A 3D cellular automaton model of the heart showing cardiac waves circulating around an inexcitable region (dashed circle), representing a reentrant circuit. (B) Schematic of a cardiac monolayer grown around a circular obstacle, illustrating a reentrant circuit with a clockwise-propagating wave around the center; the colorbar indicates cellular voltage. (C) Closed-loop system with a fixed delay applied to a cardiac monolayer exhibiting linear wave propagation. (D) Full implementation of the coupled system, where real-time signals from the tissue update the computational model, and model predictions drive adaptive control of reentrant activity. The wave propagates upward in the tissue and downward in the simulation, mimicking a clockwise circulating reentrant wave.

• 

  Fig 2. Overview of the closed-feedback loop system.

• 

  Fig 3. Two-dimensional cellular automata model implemented in WebGL using the Abubu.js library. (A) Simulation output showing the generation of uniform planar wavefronts propagating across the domain from top to bottom. The colormap represents cell voltage ranging from 0 to 1. (B) Simulation output illustrating the formation of spiral wave dynamics, demonstrating the model’s capability to reproduce complex reentrant patterns relevant to cardiac arrhythmias.

• 

  Fig 4. Motion-based wave detection algorithm. (A) Motion is extracted by subtracting the pixel intensity from its value six frames earlier, revealing tissue contractions and wave propagation. (B) Temporal signal from a representative region of interest (ROI, white rectangle) showing two intrinsic peaks corresponding to the contraction and relaxation phases of cardiac cells. (C) Excitation wave timing is determined by detecting peaks in pixel intensity across two regions of interest (ROI 1 and ROI 2), shown schematically. The algorithm calculates conduction velocity and estimates wave arrival time at ROI 3 based on spatial and temporal offsets. This value is transmitted to the feedback-loop controller.

• 

  Fig 5. Real-time optogenetic stimulation via simulation-driven LED control. The simulation reads voltage values at specific coordinates in the 2D texture and, upon detecting activation, transmits a signal to the Node.js server. The server relays this information via serial communication to the Arduino microcontroller, which then triggers a defined pattern on the LED matrix. This enables synchronized light delivery to the cardiac monolayer based on simulated wave dynamics.

• 

  Fig 6. Schematic of the coupled system integrating imaging, optical stimulation, and computational simulation.

• 

  Fig 7. Source of inaccuracy due to signal noise. (A) A frame from simulated waves generated by the projector, demonstrating double peaks corresponding to the contraction and relaxation phases of virtual cardiac cells. (B) The plot shows the averaged optical signal across columns within the region of interest (ROI) for three consecutive frames. Due to the non-uniform shape of detected waves, increasing the detection threshold (dotted line) shifts the estimated timing of signal detection by several milliseconds, thereby introducing variability in wavefront localization and contributing to temporal noise within the system.

• 

  Fig 8. Coupled system with cellular automata model and fixed-delay protocol. (A–F) Snapshots of the cardiac system under closed-loop stimulation with the cellular automaton model. In (A–D), two white rectangles indicate regions of interest used to estimate wave speed, assuming propagation from bottom to top; red arrows show wave direction. In (E–F), simulation snapshots show waves propagating from top to bottom. (G–I) Tissue activity in the absence of coupling, showing wave propagation from left to right, highlighting the capability of the coupled system in modulating the tissue dynamics. (J) Evolution of cycle periods during model-driven feedback in (A–F), exhibiting an alternating pattern. (K–O) Snapshots of the coupled system with fixed-delay stimulation. (K–N) show wave propagation in the tissue, while (O) illustrates the fixed delay applied by the system. (P–R) Tissue activity without coupling, showing wave propagation from the top left to the bottom right, representing intrinsic tissue activity in the absence of fixed-delay intervention. (S) Evolution of cycle periods under fixed-delay stimulation.