braindec — NiCLIP documentation

braindec is the Python package for NiCLIP, a contrastive language–image pre-training model that decodes brain activation maps into cognitive task descriptions from the Cognitive Atlas ontology.

User guide

• Installation
• Quickstart
• Examples

API reference

• API Reference
  • Core prediction
  • Embeddings
  • Cognitive Atlas
  • Data fetching
  • Model
  • Utilities

Installation

pip install "braindec[plotting] @ git+https://github.com/jdkent/brain-decoder.git"

Quickstart

Download the example assets and run functional decoding in a few lines:

from braindec.fetcher import download_bundle, get_data_dir
from braindec.predict import image_to_labels

work_dir = get_data_dir()
download_bundle("example_prediction", destination_root=work_dir)

# … construct paths, load model, then:
task_df = image_to_labels(
    my_activation_map,
    model_path=model_fn,
    vocabulary=vocabulary,
    vocabulary_emb=vocabulary_emb,
    prior_probability=vocabulary_prior,
    topk=10,
    logit_scale=20.0,
)
print(task_df)

See the examples gallery for a full walkthrough.

Citation

Peraza et al. (2025). NiCLIP: Neuroimaging contrastive language-image
pretraining model for predicting text from brain activation images.
bioRxiv. https://doi.org/10.1101/2025.06.14.659706

Indices

• Index

• Module Index