The classification of analog signals from individual neurons, in a sea of brain action potential, is a critical first step in network data analysis. Tetrodes produce large, multi-dimensional datasets, which then require cluster analysis to determine the stochastic events from individual neurons. As the number of tetrodes increases, the hours and effort required for manual spike sorting is prohibitive – in time, labor cost, and delayed results to qualify experiment results.
This massive amount of data, time and effort, makes obvious the need for fully automated spike sorting methods with statistics and algorithms capable of analyzing multi-dimensional spike parameters. The algorithms must produce results that rarely require user intervention and cleanup - splitting and combining clusters for satisfactory results even with overlapping clusters with unclear borders.
Every sorted data set contains Type I and Type II classification errors - it is the reality of low signal-to-noise of distant neurons in microwire recordings. Two researchers can sort the same data set and then argue about the results! Only huge close neurons are obvious for classification.
The challenge: a good algorithm with solid results that can be trusted to the point of not needing to “touch up” each electrode’s clusters and allowing results to be analyzed before the next recording session. Ideally, setting “score level cutoffs” based on a researcher’s experience with the signals will allow one to be comfortable with the classifications. Seeing cluster results and/or summary data analysis before the next recording session increases research productivity and quality.
SNAP Sorter software provides a fully automated mathematical approach that identifies clusters in multi-dimensional space through recursion, tackling the multi-dimensionality of the data. Recursion is paired with an approach to dimensional evaluation in which each dimension of a dataset is examined for its informational importance for clustering: the dimensions offering greater informational importance are given added weight during recursive clustering. To combat strong background activity, SNAP takes an iterative approach of data filtering according to a signal-to-noise ratio metric. SNAP then finds cluster cores that are thereafter expanded to include complete clusters.
SNAP Sorter is optimized for use on tetrode data, providing fast, consistent auto-clustering results:
SNAP eliminates hours of manual spike sorting on each data set - with better results than manual cutting. Data analysis can begin immediately after SNAP completes auto-clustering, with results available the same day. This provides you with access to results before your next day’s recording session - allowing you to analyze your data, assess the integrity of your experiment, and adjust/change experiment protocols and tetrode depths between sessions to achieve the quality results you need.
Note: SNAP Sorter is compatible with Neuralynx’s SpikeSort 3D (SS3D) software for final review and touch up, if needed. (However, the tetrode spike and cluster definitions are so accurate with SNAP that, in most cases, the results rarely require touch up with SS3D.)
Developed and used daily by MIT rodent and NHP labs, the SNAP Sorter algorithm for tetrode cluster analysis is patent pending by MIT, and licensed exclusively by Neuralynx.