UnitBrowser - A Tool to Evaluate and Post-Process Units Sorted by Automatic Spike Sorting Algorithms
-
1
Eberhard Karls Universität, Tübingen, Center for Integrative Neuroscience, Germany
-
2
ETH Zürich, Department of Biosystems Science and Engineering, Switzerland
-
3
ETH Zurich, Department of Biosystems Science and Engineering, Switzerland
Motivation
High-density multi-electrode arrays (HD-MEAs) with thousands of recording electrodes are a powerful tool in systems neuroscience. However, the enormous amount of data created by these devices also creates enormous challenges with respect to automatized and efficient pre-processing of the raw data – such as spike sorting. Biological analysis involving neuronal data is highly dependent on proper spike sorting, it is therefore imperative that the spikes assigned to a neuron are complete and contain minimal contamination. Current spike sorting algorithms can have drawbacks such as their dependence on templates for the spike waveform of a neuron, which may lead to splitting of a neuron by the sorter as the waveform might change over time due to tissue drifts or due to changes in experimental conditions. However, the vast amount of data recorded with HD-MEAs poses a challenge for manual inspection, even if only the final spike sorting result is checked. A tool which allows to screen hundreds of sorted units quickly and judge their sorting quality is desperately needed. We are developing UnitBrowser, a graphical tool, which allows for efficient quantitative and qualitative manual evaluation of identified units from an automatic spike sorter. In addition to assessing the spike sorting quality, UnitBrowser allows the user to manually apply several post-processing steps like merging and refinement of units.
Methods
UnitBrowser uses a MATLAB graphical interface. The user can load spike times of a unit and their corresponding waveforms. The tool computes several quality metrics such as the inter spike interval (ISI) histogram, mean waveform with standard deviations, etc. In addition, waveforms can be projected onto several 2D views based on waveform data on selected electrodes. These views include principal components, time, min-max, slices of waveforms, etc. Such projection views allow the user to identify clusters of spikes not belonging to a unit, or identify gaps of missing spikes for that unit. UnitBrowser offers several post-processing steps, including deleting spikes from a unit, merging units, and adding unclassified but detected spikes to a unit. Furthermore, the tool short-lists the possible candidate units that may require merging, or units that are duplicated or contain duplicate spikes. A particular feature of UnitBrowser is its capability to take high spatial density of the recording into account and visualize spike waveforms across the 2-dimensional layout of the array.
Results
For several experiments done in our lab, we assessed the units sorted by an unpublished automatic spike sorter (based on mean shift algorithm for clustering on a PCA-based feature set in a local neighborhood of electrodes). In our experiments, we recorded extracellular activity from mouse retina using HiDens MEA recording from 126 electrodes (Frey et al., 2010). The quality of sorting varied across experiments depending on factors such as experiment duration and changing experimental conditions. Via the UnitBrowser, we were able to visualize the sorted units and observed the following:
• Changing the ambient luminance levels in our experiments evidently resulted in altered spike waveforms, so that spikes from a large number of neurons were split into several units across the altered ambient conditions
• In very long experiments (>10 hours) we found an increased percentage of incomplete units and those that required merging as the spike waveforms changed over time possibly due to tissue movement.
The UnitBrowser offers functionality to correct for these issues. It can automatically suggest other units that may need to be merged. In addition, it has features that can be used to clean a contaminated unit. We found that such visual inspection can greatly increase the yield of high-quality units from automatic sorting. A post-processing sorting step, as enabled by the UnitBrowser, can be useful especially in long experiments that are done under unstable experimental conditions, such as experiments involving changes in ambient luminance, temperature, or pharmacological conditions.
Discussion
UnitBrowser allows effectively and quickly assessing the quality of large numbers of units from automated spike sorting algorithms. A visual representation of the quality of units using this tool can ensure properly sorted units in individual experiments. In addition, it may aid in adapting and improving automated spike sorting algorithms. For example, an algorithm that adapts to non-stationary data (Franke, Natora, Boucsein, Munk, & Obermayer, 2010) could be used as a possible solution to overcome splitting of units over time and across experimental conditions.
Conclusion
With MEAs having a large number of recording electrodes, one has to rely on automatic spike sorting algorithms. However, contamination or incomplete information in spike sorting could alter the interpretation of biological results considerably. A post-processing step can therefore be valuable to assess the quality of sorted units, and to apply appropriate corrections, if necessary. Our proposed tool enables us to (1) evaluate the quality of individual units, (2) look at various projections of either individual units or groups of units on selected electrodes and (3) apply necessary post-processing corrections. In its current state, the UnitBrowser offers features to automatically short-list and identify units that are candidates for merging with the currently inspected unit. Future work will focus on adding additional features, such as automatically short-listing units that require post-processing in the first place.
References
Franke, F., Natora, M., Boucsein, C., Munk, M. H. J., & Obermayer, K. (2010). An online spike detection and spike classification algorithm capable of instantaneous resolution of overlapping spikes. Journal of Computational Neuroscience, 29(1-2), 127–148.
Frey, U., Sedivy, J., Heer, F., Pedron, R., Ballini, M., Mueller, J., … Hierlemann, A. (2010). Switch-Matrix-Based High-Density Microelectrode Array in CMOS Technology. Solid-State Circuits, IEEE Journal of, 45(2), 467–482.
Keywords:
Retina,
spike sorting,
high density multi-electrode arrays,
sorting quality
Conference:
MEA Meeting 2016 |
10th International Meeting on Substrate-Integrated Electrode Arrays, Reutlingen, Germany, 28 Jun - 1 Jul, 2016.
Presentation Type:
Poster Presentation
Topic:
MEA Meeting 2016
Citation:
Idrees
S,
Franke
F,
Diggelmann
R,
Hierlemann
A and
Münch
T
(2016). UnitBrowser - A Tool to Evaluate and Post-Process Units Sorted by Automatic Spike Sorting Algorithms.
Front. Neurosci.
Conference Abstract:
MEA Meeting 2016 |
10th International Meeting on Substrate-Integrated Electrode Arrays.
doi: 10.3389/conf.fnins.2016.93.00054
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
22 Jun 2016;
Published Online:
24 Jun 2016.
*
Correspondence:
Dr. Saad Idrees, Eberhard Karls Universität, Tübingen, Center for Integrative Neuroscience, Tübingen, Germany, saad.idrees@cin.uni-tuebingen.de