Welcome to spikesortingtest.com, a ground truth dataset database for developing spike sorting benchmarks. This dynamic website uses computational biophysics and the NEURON modeling enviroment to model extracellular electrode recordings of neuronal activity on various electrode layouts. The aim is to test spike sorting algorithms and the limits of spike sorting and this website was setup and operated by Catalin Mitelut as part his biophysics and computational neuroscience research at UBC and previously at the Allen Institute in Seattle (see also www.openneuron.com).

From Datasets to Sorting Results

The aim of this website is to support the unbiased development of sorting algorithms, extracellular electrode layouts and neuroscience through blind testing across multiple types of extracellular datasets. You can use this website by downloading and sorting a dataset and uploading your results to get instant feedback on your performance. We provide metrics indicating (i) the percentage of correct spikes in each sorted unit (purity), (ii) the percentage of spikes that each unit captures from its best cell match (completeness) and (iii) a Single Value Metric (SVM) capturing overall sorting quality.

workflow

Download Datasets

The datasets provided come from biophysically detailed in silico simulations (collaboration between the Allen Institute and the Neuroscience Gateway (NSG) ), simultaneous whole cell and extracellular in vitro recordings and previously recorded and sorted in vivo recordings that were synthesized to make for challenging sortable data.

in silico: biophysically detailed, simulated high-density electrode layout recordings

in vitro: simultaneously patched and extracellularly recorded single cells

in vivo: reconstructed recordings from previously sorted in vivo data

datasets

Upload Results

Spike sorting results (i.e. spike rasters for sorted units) must be converted to a comma separated value (.csv or .txt) format containing: spike time (in seconds), unit ID and the maximum amplitude channel of each spike in the unit (important for unit drift). We provide python code for conversion and can also assist to convert your sorted data to the required format.