🤝 The Network Problem

💥 A neuron that fires is not the same as a neuron that communicates.

⚠️ The BCI field has operated for decades as though it were. The standard way to evaluate tissue health near an implanted electrode is to count cells and measure impedance. Neurons present, spikes detectable, device biocompatible. That framework misses something fundamental.

🧱 Neurons don’t encode information alone. They encode it through coordinated activity with their neighbors, through precise timing relationships in local circuits. A neuron that fires but has lost its functional coupling to the surrounding network is, from a decoder’s perspective, noise.

💑 We tested this directly. Using chronically implanted arrays in mouse cortex, we tracked whether pairs of neurons maintained correlated firing patterns over weeks. Functional connectivity, how reliably one neuron’s activity predicts another’s, degraded substantially near the electrode, even among neurons that remained individually detectable. They were alive. Spiking. But uncoupled from their neighbors. The local circuit was fragmented.

📉 The degradation was layer-dependent. Different cortical layers showed different vulnerability profiles, meaning the failure is structured, not random. Any intervention treating all cortical tissue as equivalent will miss the mark.

🎞 This reframes what “biocompatibility” should mean for neural interfaces. The traditional definition (does the tissue survive?) is necessary but insufficient. What matters is functional biocompatibility: does the tissue still compute? (FDA definition of biocompatibility is functional)

📊 Our data say that for tissue closest to the electrode, the answer is progressively no, and this functional degradation occurs before the structural markers the field relies on would raise concern. By the time histology looks bad, the network has been impaired for weeks.

🗓 Connect this to earlier posts. Oligodendrocyte injury, vascular disruption, lysosomal dysfunction, each can independently degrade functional connectivity without killing the neuron. Together they create a microenvironment where neurons survive as individuals but fail as a network.

🎥 Recording quality cannot be evaluated by spike count alone. We need metrics that capture whether recorded populations retain their network structure. A channel that detects a neuron is not the same as a channel that detects a neuron doing something meaningful.

Longitudinal Functional connectivity: https://lnkd.in/ePjwGw2S
Alive but Silenced: https://lnkd.in/e6bgMBAt
Review: https://lnkd.in/eYi75Sxi

#Neuroscience #Neurotechnology #BrainComputerInterface #NeuralEngineering #Bioengineering #FunctionalPerformance