Preclinical testing of large-scale brain implants to map vocal cortical networks
We have developed a new paradigm to record cortical activity in freely moving and behaving minipigs while they vocalize (see Palma et al., 2022).
​
This paradigm is now used to validate novel brain implants before human translation and to detail the dynamics of cortical networks underlying vocal production. We are particularly interested in the extent to which these dynamics compare to those observed in human recordings.
Neuromorphic processing of neural signals
Large-scale neural implants generate tremendous amounts of data that need to be processed in real time for rehabilitation applications such as brain-computer interfaces. Embedding the processing of neural signals at the level of recording sites within cortical implants thus become challenging due to heavy computational load and thus high power consumption. We thus develop new approaches based on artificial spiking neural networks to enable online unsupervised processing of neural signals (e.g., spike sorting) compatible with very-low-power neuromorphic hardware.
​
See Zenodo repositories for spike sorting code and artificial test data
Real-time software framework
Closed-loop paradigmes used in brain-computer interfaces require real-time processing of neural signals acquired using multielectrode arrays. For this purpose, we develop a highly optimised software framework that allow user-friendly creation of real-time neural processing pipelines based on a modular interconnection of individual processing blocks. Currently, this software is compatible with the following commercial recording systems: Blackrock Neuroport, Intan RHD and MultiChannelSystems WS2100.
Spatiotemporal mapping of multielectrode array neural data
We are making available the Neuromap software that allows spatiotemporal visualization of multielectrode array data and merging with anatomical images.
​
See the NeuroMap website for free download