Discovering Features for Neuron-type Identification

Advancements in multichannel recordings of neuron activity present an opportunity to discover novel features of spatially-varying extracellularly-recorded action potentials that are useful for identifying neuron-types. Traditional approaches to classifying neuron-types often rely on computing waveform features based on conventions of single-channel recordings and thus inherit their limitations. However, multichannel spatiotemporal waveforms are the product of signals from underlying current sources being mixed within the extracellular space. We introduce an approach to demix the underlying sources of spatiotemporal waveforms. Using biophysically realistic computational models, we simulate extracellular action potential waveforms and characterize them by the relative prevalence of these underlying sources, which we use as features for identifying neuron-types. We then organize known neuron-types into a hierarchy of neuron-types based on differences in the source prevalences, providing a multi-level classification scheme. This simulation-based approach provides a machine learning strategy for neuron-type identification.

Bio
https://iconlab.asu.edu/crook.html