Most EEG machine learning tutorials assume you can write Python: load with MNE, epoch, build a scikit-learn pipeline, write a PyTorch training loop, write the cross-validation, plot the results. For a researcher whose expertise is the neuroscience, that is weeks of engineering before the first result, and every hand-written step is somewhere a leak can creep in.
BCILattice turns each of those steps into something you click and inspect. It does not hide the decisions. It shows them, records them, and keeps the output of every step so you can check it.
What you do without code
| Task | How |
|---|---|
| Import | Pick a file or a whole folder of subjects; confirm label columns; see quality warnings first. |
| Preprocess | ICA Review, Channel Quality, Epoching, Normalization and Augmentation, previewed on one file, then applied to all. |
| Analyse | Topographies, a 3D cortical view, ERSP/ITC time-frequency, channel selection by 20 criteria, dimensionality reduction. |
| Extract features | 71 named feature families across five domains, with a sensible default preselected. |
| Train | Wire blocks on a canvas: classical models, Riemannian classifiers, EEGNet and other deep architectures, transformers. |
| Evaluate | Validation Scheme and Nested CV blocks, per-fold scores, and a results validation tab that flags leaky designs. |
| Report | A structured report built from the run's own numbers, plus a document editor for the prose. |
The canvas helps you wire it
The canvas also suggests the next block. Select a trainer and it offers a matching validation scheme, already wired in one click. Pipeline recipes in the docs give block-by-block wiring for channel selection, dimensionality reduction, cross-validation, leave-one-subject-out and domain adaptation.
When you do want code
The Custom Code block runs Python you write inside the pipeline (it is the one plan-gated block), and the results notebook runs Python in-process against a finished study. Code is there for when you need it, not a requirement to get started.
Questions
Can I analyse EEG without programming?
Yes. Graphical tools such as BCILattice cover preprocessing, artifact removal, epoching, feature extraction, model training and evaluation without writing code.
Is no-code EEG analysis rigorous enough to publish?
It can be more rigorous than hand-written scripts, because every step records its parameters and the evaluation guardrails catch leaky splits and unnested tuning. What matters for publication is that the methods are correct and reported, not how they were typed.
Can I train deep learning models on EEG without code?
Yes. In BCILattice, EEGNet, ShallowConvNet, DeepConvNet, ATCNet, EEGConformer, LSTMs and EEG transformers are canvas blocks, and optimisers, losses, schedulers and early stopping are blocks too.
Read the reference
Related
Free EEG Analysis Software
There is excellent free EEG software. The real cost is in what you have to set up, learn or write before you get a result, and that varies a lot.
EEG Machine Learning
A node canvas of 119 blocks covers the pipeline from design matrix to evaluation — and the evaluation half is built to make an inflated score hard to produce by accident.
BCI for Beginners: How to Get Started with EEG and BCI
You do not need a headset to start. A public dataset, a laptop and a clear first question will teach you more than a new device and no plan.
BCI Analysis Software
Eleven steps sit between a conditioned recording and a labelled design matrix. In BCILattice each one shows you its result before you commit it, and records what it ran with afterwards.