Skip to content

Neuroinformatics

The neuroinformatics plugin covers neuroscience data standards (Brain Imaging Data Structure (BIDS), Hierarchical Event Descriptors (HED)), experiment design (PsychoPy, Lab Streaming Layer (LSL)), and dataset validation.

The plugin map

A conversion skill sits at the center; a validator agent defends its output; experiment design covers the data-collection side that happens before conversion:

Neuroinformatics plugin map: bids-conversion at the center, defended by the bids-validator agent, with experiment-design covering data collection

One command, six steps, to a validated dataset

Converting raw recordings into a shareable BIDS dataset is a fixed sequence:

BIDS conversion flow: Inventory, Scaffold, Convert files, JSON sidecars, TSV tables, Validate

  1. Inventory: source data formats, subjects, channels
  2. Scaffold: dataset_description.json, participants.tsv
  3. Convert files: BrainVision, EEGLAB .set, EDF, BDF, and others
  4. JSON sidecars: SamplingFrequency, EEGReference, and other required metadata
  5. TSV tables: channels, events, electrodes
  6. Validate: bids-validator

Why the layout is fixed

BIDS's predictable folders and filenames are what let EEGLAB, MNE-Python, bids-validator, BIDS Apps, OpenNeuro, and NEMAR all read a dataset without dataset-specific glue code:

BIDS tree: predictable dataset_description.json, participants.tsv, and per-subject eeg/ folders with .set signals, .json sidecars, and .tsv tables

Why BIDS: one predictable layout with machine-readable sidecars lets EEGLAB, MNE-Python, bids-validator, BIDS Apps, OpenNeuro, and NEMAR all read the same data unchanged

A standard structure turns "my data" into "reusable data": the same folder and filename conventions apply to every BIDS dataset, anywhere, which is what makes cross-dataset tools and mega-analysis pooling possible in the first place.

HED annotation

HED annotation, tagging events with a standardized, machine-readable vocabulary for the what of an event, not just its timing, lives inside the bids-conversion and experiment-design skills' reference material rather than as a separate skill. Conceptually, going from a plain-English event description to a schema-valid HED tag string is a parse → tag → validate pipeline:

HEDit pipeline: parse natural language into structured facts, tag by retrieving HED schema nodes, then validate against the official HED validator with feedback looped back into re-tagging

The HED schema is the contract; nothing in that pipeline invents vocabulary outside it.

Skills

  • bids-conversion: guided conversion to BIDS for EEG, EMG, MEG, fMRI, and other modalities
  • experiment-design: PsychoPy + LSL experiment scaffolding, feeding BIDS-compatible output back into conversion
  • bids-validator agent: autonomous validate-and-fix: runs bids-validator, diagnoses errors, and applies corrections

Try it

"Convert ./raw-data to BIDS format, modality EEG, task rest"
"Validate the BIDS dataset at ./bids-dataset"
"Design a visual oddball ERP paradigm with 2 conditions"

Learn more

The Agentic Research Course week 9, "Neuroinformatics," covers BIDS, HED, and experiment design hands-on.