Deciphering Cardiac Neural Network Dynamics: From In-Vivo Recordings to Unsupervised Spike Labeling

The cardiac nervous system continuously modifies the heart's mechanical and electrical functions to meet the body's demand for blood circulation. This closed-loop control mechanism consists of interconnected neural populations that function collaboratively. It has evolved to manage different facets of cardiac function, resulting in cardiopulmonary signals such as blood pressure and respiration. In this work, we focus on the stellate ganglia, a bundle of sympathetic “fight-or-flight” nerves located in the front of the neck, which receives and integrates central, peripheral, and cardiopulmonary information to produce sympathetic responses in various disease pathologies. Using in-vivo extracellular microelectrode recordings, we investigate differences in information transfer between healthy porcines and porcines with chronic heart failure. We introduce two new metrics, cofluctuation and neural specificity, to describe the control of cardiac function by the stellate ganglion. Our analysis reveals the complex and dynamic nature of this control system and provides new insights into the neural mechanisms underlying cardiac function. Additionally, we discuss the challenges of spike sorting in our dataset and our approach to spike sorting as a subset of unsupervised learning.