← All projects
01 · Biocomputing · Open source

NeuroBridge

An open-source analysis platform for brain-organoid multi-electrode array recordings. Built solo in fourteen days. Shipped 23 April 2026.

neurocomputers.io / app / fs437
FinalSpark fs437
2,612,408 spikes · 32 ch · 118h continuous
organoid · MEA · 2026-04
Mean rate
17.4 Hz
σ 4.2
Bursts / min
3.8
Bakkum 2013
NCI
0.412
composite
Φ (irreducibility)
0.087
Tononi 2004

Spike raster — first 60s

window 0–60s · channels 1–12
CH 1
24.1
CH 2
18.6
CH 3
22.3
CH 4
14.0
CH 5
26.7
CH 6
17.2
CH 7
10.4
CH 8
20.9
CH 9
13.1
CH 10
19.5
CH 11
15.6
CH 12
23.4

Methods

9 active
Burst detectionBakkum 2013ON
Functional connectivitySchreiber 2000ON
Criticality fitClauset 2009ON
IIT Φ approx (Queyranne)Tononi 2004ON
Kuramoto metastabilityShanahan 2010ON
Predictive coding proxyFriston 2005ON
Sleep–wake HMMSaponati 2023ON

Thesis.

Living-neuron compute is moving from research curiosity to commercial substrate, and the field has a tooling gap. Labs working with multi-electrode arrays each rebuild the same analysis pipeline in slightly different ways, with no shared baseline against which to compare results.

I'm not a domain expert. I'm a founder who reads the peer-reviewed literature directly and ships what the field needs — with AI-assisted engineering as a force multiplier. NeuroBridge is the first artefact of that approach.

What's inside.

Nine peer-reviewed analysis methods, each shipped with the citation to its original source:

  • Burst detection — Bakkum 2013
  • Functional connectivity via transfer entropy — Schreiber 2000
  • Criticality fitting — Clauset 2009
  • IIT Φ approximation via Queyranne MIP — Tononi 2004
  • Kuramoto metastability
  • Predictive coding proxy
  • Sleep–wake HMM staging
  • Composite scoring & comparative profiling against ten reference neural systems
  • Surrogate-based significance testing with Bonferroni correction

137 REST API endpoints. ~12,000 lines of Python. Stack: Python 3.12 / FastAPI / NumPy / SciPy on the backend, Next.js 16 + React 19 + D3.js on the frontend. Deployed on Vercel with a VPS running Caddy and PM2 for the API.

Validation.

Validated end-to-end on FinalSpark's public fs437 dataset:

2.6M
Spikes processed
32
Channels
118h
Continuous recording
4
Organoids

A methods preprint is forthcoming on arXiv this April.

Design priorities.

Three calls I made deliberately, because the field punishes overclaiming:

  • Citation chain. Every method names its source paper in the UI and the API response.
  • Honest labels. The composite score is "Network Complexity Index", not "IQ". Φ is "network irreducibility measure", not "consciousness".
  • Auditable end-to-end. MIT license, source on GitHub, package on PyPI. Anyone can rerun the pipeline on their own data and check my work.

Get involved.

Open to dataset exchanges, method validation, and joint methodological work with academic and industry labs working on MEA, organoid, or cell-culture electrophysiology. If that's you — write me directly.

Get in touch
Next project

Content production network →

All projects