Why this exists

Signal Studio started with a question nobody could answer well: why is working with neural recordings still this hard, when the methods themselves have been settled for years?

The same script, written again

We were PhD students in the United States and in France, working on electroencephalography and intracranial recordings in cognitive neuroscience. Different labs, different countries, the same routine: open a new dataset, and start rewriting the preprocessing script that somebody in the lab had already written three times before.

Each version worked. Each version was slightly different. And nobody could say exactly what any of them did without reading every line, because the decisions that changed the result were buried in function arguments spread across four files.

The part that bothered us

It was not the effort. Writing analysis code is part of the work, and it should be. What bothered us was the opacity. When you inherit a colleague's pipeline, you inherit a set of choices you cannot see: which window function, which transition width, whether the filter was applied before or after epoching, what happened to the rejected trials.

Those choices change results. Sometimes a little, sometimes enough to change a conclusion. And in a field where reproducibility is a live problem, we were all shipping analyses whose steps could not easily be inspected by the person reading the paper.

Who it was closed to

The other thing we kept running into: the tools assumed you could program. Not everybody in cognitive neuroscience can, and there is no reason they should have to. A clinician who understands seizure networks better than any of us should not be locked out of an analysis because it requires fluency in a scripting language.

The knowledge in this field is not evenly distributed with the programming skill. That gap costs the field real work that never gets done.

What we built instead

Signal Studio turns each processing step into a card you can open. The parameters are visible. The operations inside are visible. A pipeline is a file you can hand to a colleague, and they get the same run you had, not an approximation of it.

And every built-in card is checked against an independent reference implementation on real recordings, because "it compiles" is not the same as "it computes the right thing". The agreement figures are published in the validation method.

Where it is now

It is in public beta, developed by NeuraCrypt. It is unfinished, and it is being used on real data. If you want to help push it forward, the beta programme is open.