Below we have defined preliminary spike sorting metrics and anonymized results from all sorts submitted to this project.

Sorting Metrics

Purity
(Sorted Unit based metric)

The (largest) # of spikes in a unit sourced from a unique cell (i.e. best match) divided by the total spikes in unit.

Completeness
(Cell based metric)

The (largest) # of spikes in a unit sourced from a unique cell (i.e. best match) divided by the total spikes in matching cell.


U_i = Set of spikes from sorted unit i
C_j = Set of spikes from ground truth cell j
|U_i| = Number of spikes in set

We have also developed a Single Value Metric (SVM) which provides a single value to capture sorting results. As this metric is under development, we are looking for feedback from users that consider the metric may not be representative of their sorting results.

As the databse grows we would like to identify the best software and operators so please feel free to send us your result (.jpg file) from your best sort. Importantly, these high quality sorts will help us identify the limits of spike sorting for a particular dataset.


Global Results

Silico_0
Silico_1