Epic: HBN ERSP - Boy-shot vs Puppy-shot in 'ThePresent' (R3)

[neuromechanist]

Goal

Test whether the first 500 ms of EEG responses differ between shots that open with the boy visible and shots that open with the puppy visible during passive viewing of the Pixar short The Present. Uses HBN-EEG R3 (ds005507). Metric: event-related spectral perturbation (ERSP), aligned to shot onsets.

Hypothesis

Shots opening with the puppy elicit a different early ERSP pattern than shots opening with only the boy. Direction, frequency band, and topography are intentionally unspecified at this stage; they get locked during the Phase 5 plan-mode pass based on literature and cluster QA.

Locked Contrast

Contrast 1 - Boy-only vs Puppy-only:

• Boy-only: has_boy=1 AND has_puppy=0 (20 stim-side shots)
• Puppy-only: has_boy=0 AND has_puppy=1 (15 stim-side shots)
• Mutually exclusive; LLR (log luminance ratio) as regressor of no interest in the group GLM.

See .context/ideas.md for the alternatives considered and rejected.

Dataset

• Local primary: /Volumes/S1/Datasets/HBN/L100/R3_L100_bdf (184 subjects, 100 Hz BDF, full R3 downsampled)
• Full 500 Hz validation: s3://fcp-indi/data/Projects/HBN/BIDS_EEG/cmi_bids_R3 (post-validation only)
• Stimulus events: shot_events.tsv in repo root, 49 trusted rows

Phases

Each phase is a sub-issue, a worktree, a branch, a PR.

• Phase 1 - Preprocess R3 (BIDS import, 1 Hz HPF, cleanline, channel rejection)
• Phase 2 - AMICA + dipfit
• Phase 3 - ICLabel classification and non-brain rejection
• Phase 4 - Shot-event expansion and epoching
• Phase 5 - ERSP precompute and IC clustering
• Phase 6 - Group statistics and figures

Deliverables

• Group-level ERSP maps per condition on 0-500 ms, baseline -500 to -100 ms
• Cluster-level stats across occipital and temporal IC clusters
• Conditionwise ERSP-difference figure at the cluster of maximal effect
• STUDY with _amica, _iclabel, _clustered, _epoched intermediates under derivatives/

Non-goals

• Other HBN movies
• Deep learning
• Custom HED work
• CI/CD (deferred to Week 4)

References

• project_brief.md - authoritative spec
• .context/plan.md - current tasks
• .context/research.md - parameter justifications
• Shirazi et al., HBN-EEG: A Curated Collection, bioRxiv 2024, 10.1101/2024.10.03.615261
• Alexander et al., The Healthy Brain Network, Sci Data 2017, 10.1038/sdata.2017.181