All projects

Project report · 2026

SpikeLab

Neural signal analysis without a black box

In one line

SpikeLab turns continuous recordings and pre-sorted NeuroExplorer data into a traceable analysis workflow. Every stage can be inspected, rerun, and exported instead of being locked inside a proprietary interface.

Question

Can one local tool make multi-electrode array analysis reproducible—from raw spike detection to burst and waveform measurements—without hiding the important decisions?

What I built

  • Built an offline Streamlit application for spike detection, waveform analysis, inter-spike intervals, firing rates, burst detection, and electrode comparison.
  • Supported both continuous recordings and already sorted NeuroExplorer spike data so the same measurements can be checked across workflows.
  • Added reproducible exports and deterministic fixtures for burst algorithms, filtered detection, waveform boundaries, and EDF calibration.

Main result

of tested NeuroExplorer results reproduced
95%
analysis areas covered by regression fixtures
4

Result figure

Filtered synthetic voltage trace with threshold crossings, retained detections, and the refractory-suppressed crossing
Synthetic regression fixture: six threshold crossings are found and five are retained. This checks documented software behaviour; it is not biological validation.

What the work showed

  • The tool reproduced the reference NeuroExplorer output in 95% of the cases recorded during the project.
  • Fixed inputs now produce fixed expected outputs, making changes to detection and burst logic easier to audit.
  • The workflow keeps raw traces, detected events, derived metrics, and exports connected, reducing manual hand-offs between tools.

What it does not prove

  • The deterministic fixtures check software behaviour; they do not prove that an algorithm is biologically correct for every preparation.
  • Detection thresholds and burst parameters still require experimental judgement and should be reported with any scientific result.

Conclusion

SpikeLab makes neural-signal analysis easier to inspect and reproduce. Its strongest contribution is not a new biological claim; it is a dependable path from a recording to results that can be checked.