The activity of the motor cortices (M1) exhibits characteristic temporal and statistical properties that emerge from the interaction of large neuronal populations. Rather than reproducing the activity of individual neurons, an important challenge is to develop neural networks that generate population dynamics matching those observed in biological recordings. Such biologically realistic models are essential for understanding cortical computation and for developing neuromorphic brain–computer interfaces that interact naturally with neural tissue.

This project aims to train a spiking neural network on chip to reproduce the statistical properties of motor cortical populations recorded from rodents during motor behavior. Instead of matching spike trains neuron by neuron, the network will be optimized to reproduce key electrophysiological descriptors of cortical activity, including the mean bursting rate (MBR), median inter-spike interval (ISI), mean firing rate (MFR) distribution, inter-burst interval (IBI), and burst length. The resulting network will therefore emulate the collective dynamics of the biological networks while remaining suitable for deployment on neuromorphic hardware.

 

Methodology

1. Analyze extracellular recordings from the rat primary motor cortex and extract statistical descriptors of neuronal activity, including MBR, median ISI, MFR distribution, IBI, and burst duration.

2. Develop and train a spiking neural network whose parameters are optimized to reproduce the same statistical properties observed in the biological recordings.

3. Compare the biological and artificial networks using both neuron-level and population-level metrics, evaluating the similarity of firing statistics, burst organization, and temporal dynamics.

4. Investigate how network architecture and connectivity influence the emergence of biologically realistic activity, identifying minimal network configurations capable of reproducing natural motor cortical dynamics.

 

Available Material

• Extracellular recordings from rat primary motor cortex.

• Analysis pipeline for extracting neuronal firing statistics.

• Spiking neural network simulation framework.

• Access to neuromorphic hardware for deployment and validation.

 

Requirements

• Python programming.

• Interest in computational neuroscience, spiking neural networks, and neuromorphic computing.

 

Contacts

Elisa Donati (elisa@ini.uzh.ch)

 

References

1. Chiappalone, M., & Levi, T. (2026). Advancing neuroengineering with Neuromorphic Twins. Nature Communications, 17(1), 1938.

2. Richter, O., Wu, C., Whatley, A. M., Köstinger, G., Nielsen, C., Qiao, N., & Indiveri, G. (2024). DYNAP-SE2: a scalable multi-core dynamic neuromorphic asynchronous spiking neural network processor. Neuromorphic computing and engineering, 4(1), 014003.