Review: Sorting and Clustering Techniques for High Density Electrode Array Recordings

isabel delgado ruz
Neuronal activity is known to be sparse, since only small populations of neurons showing above threshold spiking activity has been recorded on different cortical areas (imaging techniques and electrophysiology). However, existing sorting and clustering algorithms tend to discard neurons that show little activity (low firing rate, temporally sparse activity) or spikes from neurons that deviate from the mean (outliers from main cluster on high dimensional space). On the other hand, High Density
more » ... ectrode Arrays allows for spike sorting and clustering which recovers units of small amplitude (voltage) that can not be sorted with other recording technologies. Here we review the different approaches followed on the literature for spike detection, feature extraction, sorting and clustering. We also discuss the importance of developing a simple architecture allowing for new developments to plug-in on existing analysis workflow. Additionally analogue aspects of the Action Potential (and sub-threshold activity) have been ignored on population analysis, sensory response characterisation and neural code studies. Mostly given the technological restrictions and access to technology, since for wire electrodes, or even tetrodes the view from different neuron's spikes turns out to be binary given the restricted area and spatial resolution. Although these simplification has been enough for most tasks so far, they are certainly not enough for advanced and realistic characterisation of how information is processed and transmitted in the brain. Several open-source projects are available for processing electrophysiological data, but their inter-operation is restricted since an standardization and modular designs are still needed. Here we propose a framework's model in order to freely develop modules that can be connected to each other, allowing for flexibility on the analysis process. Since processing required is heavily dependant on the dataset, experimental conditions and research question. Designing independent and reusable modules [...]
doi:10.6084/m9.figshare.16989808.v1 fatcat:hz45hnxtd5dgfognwmlt72s4ei