skills.rest
CUHK-AIM-Group
CUHK-AIM-GroupOFFICIAL
Explore all skills by this author →
WebsiteGitHub

nsd-skill

Orchestrates BIDS validation, multimodal processing, and stimulus extraction for the Natural Scenes Dataset.

89|5|Updated Mar 25, 2026
Search Tags:#quality control#neuroimaging#coco#fmri#nsd#natural scenes dataset#bids validation
One-click install
npx skills add https://github.com/CUHK-AIM-Group/NeuroDiscovery --skill nsd-skill-cuhk-aim-group
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: nsd-skill
Source: https://github.com/CUHK-AIM-Group/NeuroDiscovery/tree/main/skills/nsd-skill
Command: npx skills add https://github.com/CUHK-AIM-Group/NeuroDiscovery --skill nsd-skill-cuhk-aim-group

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pandas, and includes scripts (resource) components.

What problem does it solve? Working with the Natural Scenes Dataset (NSD) requires coordinating BIDS validation, 7T structural and task-fMRI processing, COCO stimulus metadata extraction, and quality control across 8 subjects with ~30-40 sessions each, which is error-prone when done manually. ## Core Features & Use Cases - BIDS Validation: Verify NSD directory structure, subject completeness, session counts, and stimulus file presence with a compliance report via scripts/validate_nsd.py. - Multimodal Processing Delegation: Route T1w structural MRI to smri-skill and task-fMRI (natural scene viewing) to fmri-skill through a confirmed, numbered execution plan. - Stimulus Metadata Extraction: Merge NSD trial events with COCO captions and object categories into a stimulus metadata CSV for stimulus-response and voxel-wise encoding analyses. - QC Summaries: Generate per-subject quality control reports with framewise displacement metrics and exclusion lists using 7T-appropriate thresholds. - Use Case: A visual neuroscience researcher downloads NSD data and asks for end-to-end processing; the skill validates the BIDS layout, delegates sMRI and fMRI pipelines, extracts COCO stimulus annotations, and produces QC summaries in a clean nsd_output/ directory. ## Quick Start Ask the agent to run the full NSD pipeline on your local BIDS directory, for example: "Validate and process my NSD data at /data/NSD, extract the COCO stimulus metadata, and generate QC summaries."

Frequently Asked Questions about nsd-skill

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I validate NSD BIDS structure before processing?▼

Run the validate_nsd.py script with --input pointing to your NSD BIDS directory and --output for the report path. It checks subject completeness, T1w and task-fMRI presence, session counts, and stimulus events files, then writes a CSV compliance report.

How to extract COCO stimulus metadata from NSD trials?▼

Use the extract_nsd_stimulus.py script with --nsd-dir and optionally --coco-dir pointing to COCO annotations. It merges trial-level events with nsd_stiminfo.tsv and COCO captions and categories into a single stimulus metadata CSV.

What fMRI processing does the Natural Scenes Dataset require?▼

NSD task-fMRI involves natural scene viewing with a fixation task across ~30-40 sessions per subject. The skill delegates preprocessing and voxel-wise encoding to fmri-skill, since standard task GLM often does not apply to this dense sampling design.

What QC thresholds are used for 7T fMRI motion exclusion?▼

The QC script computes mean and max framewise displacement from fMRIPrep confounds files and flags subjects exceeding a default FD mean threshold of 0.3 mm, which is typical for 7T data. The threshold is configurable via --fd-threshold.

What are the limitations of processing NSD data?▼

NSD is a high-resolution 7T dataset with ~30 hours of fMRI per subject, requiring significant compute resources and storage. The skill is orchestration-only, so actual preprocessing depends on the availability of smri-skill, fmri-skill, and claw-shell.

Related Skills

Explore More Education & Research Skills→
bcmcpherbcmcpher

bids

Validate and organize BIDS-compliant neuroimaging dataset structures.

Community
Advanced
NeuroAIHubNeuroAIHub
34

Neural Population Decoding Analysis

Design and validate multivariate neural decoding analyses for EEG, MEG, or fMRI population data.

Official
Advanced
CUHK-AIM-GroupCUHK-AIM-Group
78

nibabel-skill

Load and validate NIfTI files, inspect affines, and convert voxel coordinates.

Official
Intermediate
KaiserWhoLearnsKaiserWhoLearns

fuzzy-match

Match strings across datasets to reconcile entities with spelling variations.

Community
Intermediate