Spike Sorting Algorithms Comparison
-
1
Czech Technical University in Prague, FEE - Dept. of Cybernetics, Czechia
-
2
Charles University , First Medical Faculty, Czechia
Introduction: One of the key problems occurring in neuroinformatics today is the ability to properly extract and sort different spike trains from extracellular recordings. Whilst the spike detection itself is relatively straightforward in up to low signal-to-noise conditions, the spike sorting is proven to be quite a challenge even when only low noise is present. To face such nuisances spike-sorting algorithms employ advanced feature extraction and clustering techniques. As there is already a number of such algorithms at hand, an independent performance comparison is presented to ease researcher's decision, which one to use in his particular work. The comparison is made among four algorithms in total - KlustaKwik [1], Qsort [2], Spike2 (commercial [3]), WaveClus [4]. Various types of artificially generated signals are used to make the comparison as objective as possible - including signals made by our proposed method. Extracellular signals recorded from patients during Deep Brain Stimulation surgery is then used to show how these algorithms perform on real data.
Results: Spike2 and WaveClus underestimated the number of neurons when not operated by an expert. On the other hand, KluskaKwik overestimated the number of neurons. In KlustaKwik and Spike2, spikes from different neurons with similar spike shapes may not be distinguished properly. Qsort was the most reliable sorting algorithm when noise level was about 0.1. However, for higher noise levels (above 0.1), WaveClus performed better. For higher noise levels (>.25), spike shapes could not be reconstructed properly and all the methods we compared failed.
Conclusion: Our comparison of spike sorting algorithms revealed that none of the algorithms used today is optimal. Their performance depended not only on the quality of the signal being processed, but also on operator's experience.
Acknowledgment
This work has been supported from the Czech Ministry of Education: research program MŠM 0021620849 and from the Grant Agency of Czech Republic: grant 309/09/1145.
Figure 1: Visualization of results of different clustering algorithms for one sample signal with noise level 0.1 taken from [4]. The first 1s segment of the signal is depicted for illustration purposes. For each clustering algorithm all spikes in each cluster are displayed with i) highlighted computed mean spike shape and ii) the total number of spikes. PCA visualization of clusters in each algorithm's features space in 2D is shown. As a reference, original clusters are shown using the a priori information from signal generation.
References
1. K. D. Harris et al. "Accuracy of tetrode spike separation as determined by simultaneous intracellular and extracellular measurements.", Journal of Neurophysiology, vol. 84, no. 1, pp. 401-414.
2. U. Rutishauser et al. "Online detection and sorting of extracellularly recorded action potentials in human medial temporal lobe recordings, in vivo.", Journal of Neuroscience Methods, no. 154, pp. 204-224, 2006.
3. Cambridge Electronic Design, "Spike2", online; http://www.ced.co.uk/pru.shtml?spk6wglu.htm.
4. R. Q. Quiroga et al. "Unsupervised spike detection and sorting with wavelets and superparamagnetic clustering.", Neural Computation, vol. 16, no. 8, pp. 1661-1687, 2004.
Conference:
Neuroinformatics 2009, Pilsen, Czechia, 6 Sep - 8 Sep, 2009.
Presentation Type:
Poster Presentation
Topic:
Computational neuroscience
Citation:
Wild
J,
Sieger
T,
Novak
D and
Jech
R
(2019). Spike Sorting Algorithms Comparison.
Front. Neuroinform.
Conference Abstract:
Neuroinformatics 2009.
doi: 10.3389/conf.neuro.11.2009.08.007
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:
21 May 2009;
Published Online:
09 May 2019.
*
Correspondence:
Jiri Wild, Czech Technical University in Prague, FEE - Dept. of Cybernetics, Prague, Czechia, wildj1@fel.cvut.cz