Real-time spike sorting platform for high-density extracellular probes with ground-truth validation and drift correction
James Jaeyoon Jun, Catalin MitelutiD, Chongxi LaiiD, Sergey L. GratiyiD, Costas A. AnastassiouiD, T.D. HarrisiD
Abstract
Electrical recordings from a large array of electrodes give us access to neural population activity with single-cell, single-spike resolution. These recordings contain extracellular spikes which must be correctly detected and assigned to individual neurons. Despite numerous spike-sorting techniques developed in the past, a lack of high-quality ground-truth datasets hinders the validation of spike-sorting approaches. Furthermore, existing approaches requiring manual corrections are not scalable for hours of recordings exceeding 100 channels. To address these issues, we built a comprehensive spike-sorting pipeline that performs reliably under noise and probe drift by incorporating covariance-based features and unsupervised clustering based on fast density-peak finding. We validated performance of our workflow using multiple ground-truth datasets that recently became available. Our software scales linearly and processes up to 1000-channel recording in real-time using a single workstation. Accurate, real-time spike sorting from large recording arrays will enable more precise control of closed-loop feedback experiments and brain-computer interfaces.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 153
- Assay
- Research Focus
- Approach
Abstract summary
Claude Opus 5.5 · xhigh effort · 18 Oct 2026 · from the abstract
Covariance-based features plus unsupervised density-peak clustering sort spikes from up to 1000 channels in real time on one workstation, while tolerating noise and probe drift.
High-density electrode arrays resolve single spikes from many neurons, but assigning them to cells still depends on manual curation that does not scale past about a hundred channels over hours of recording, and sorters have been hard to validate for want of ground-truth data. This pipeline is tested against ground-truth recordings that have recently become available and runs fast enough to be used while an experiment is going on.
Key results
ClaimedA spike-sorting pipeline built on covariance-based features and unsupervised clustering by fast density-peak finding is reported to stay reliable in the presence of recording noise and movement of the probe.
Design described in the abstract; no error rates, drift magnitudes or comparison with other sorting algorithms are given. · Abstract
ClaimedPerformance was checked against multiple ground-truth datasets that had recently become available.
Validation is stated to use multiple recently released ground-truth datasets, none of which is named or counted, and no accuracy, precision or recall figure is reported. · Abstract
SuggestedRuntime grows linearly with recording size, so a single workstation keeps up with a 1000-channel recording as it is acquired.
Stated as linear scaling and real-time processing of up to 1000-channel recordings on one workstation; no hardware specification, recording length or timing table accompanies it. · Abstract
Methods, data and reproducibility
A spike-sorting pipeline for high-density extracellular probes: covariance-based features are computed from detected events and clustered without supervision by fast density-peak finding, with the pipeline designed to absorb noise and probe drift; it was run on ground-truth datasets and on large-channel-count recordings to gauge throughput.
- Ground-truth extracellular recordings used to validate sorting output
- High-density probe recordings processed to test throughput · n = up to 1000 channels
- Assay conditions & compute
- Reported to scale linearly and to sort up to a 1000-channel recording in real time on a single workstation; no processor, memory or GPU details are stated.
- Reproducibility gaps
- The software is not named and no repository is given; the ground-truth datasets are referred to only as recently available, without identifiers.
Limitations (3)
- The authors state that validation of spike sorters is hindered by a shortage of high-quality ground-truth datasets. · stated by the authors
- No accuracy numbers and no comparison with other spike-sorting packages appear, so reliability under noise and drift is asserted rather than quantified. · noted in this summary
- The real-time claim is given without hardware specification, recording duration or the probe geometry it was measured on, so it cannot be reproduced from the abstract. · noted in this summary
Who should read:Read this if you sort spikes from high-density probes at scale, or need sorting fast enough for closed-loop feedback experiments or brain-computer interfaces.
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Spike sorting for large, dense electrode arrays
Cyrille Rossant, Shabnam N Kadir, … Kenneth D. Harris·bioRxiv·Feb 16, 2015
- Fast and accurate spike sorting in vitro and in vivo for up to thousands of electrodes
Pierre Yger, Giulia L. B. Spampinato, … Olivier Marre·bioRxiv·Aug 4, 2016
- Kilosort: realtime spike-sorting for extracellular electrophysiology with hundreds of channels
Marius Pachitariu, Nicholas Steinmetz, … Harris Kenneth D.·bioRxiv·Jun 30, 2016
- Unsupervised spike sorting for large scale, high density multielectrode arrays
Gerrit Hilgen, Martino Sorbaro, … Matthias H. Hennig·Cell ReportsPeer-reviewed·Mar 1, 2017
- Scalable spike sorting across thousands of neurons by modeling neural dynamics with NeuroSort
Lin Liu, Zhengwei Hu, … Xiaojian Li·bioRxiv·Nov 11, 2025
- SpikeForest: reproducible web-facing ground-truth validation of automated neural spike sorters
Jeremy F. Magland, James Jaeyoon Jun, … Alex H. Barnett·eLifePeer-reviewed·May 19, 2020