🧠 Contributing to a BIDS Standard for Multimodal and EEG Hyperscanning

_anna-monnier

Short bio: a "Generative Neurophenomenology" PhD !

I am a PhD student in Psychiatric Sciences at Université de Montréal - CHU Sainte-Justine ! My work is at the intersection of conscious phenomena & psychiatry and my methods of investigation:

• EEG & ECG hyperscanning,
• Social neuroscience,
• and Phenomenology (subjective, first-person methods).

This practice is called Generative Neurophenomenology [1]

Introduction

Brain Imaging Data Structure (BIDS) is the leading standard for organizing neuroimaging data, but it was not originally designed for hyperscanning — the simultaneous recording of two or more brains. This project contributes to filling this gap by documenting and sharing the decisions made to adapt BIDS to a naturalistic piloting EEG hyperscanning dataset.

A multimodal project	EEG + ECG Hyperscanning data, videos + multiple subjective experience metrics
Analyses	Examining 🧠 interpersonal synchrony in relation with the felt experience of 👥 Togetherness (feeling one with a partner) !
Big Cohorte	80 dyads, mother-child pairs, autistic and non-autistic

--> The project presents a critical need for bidsification !!!

📄 Project poster

Why this matters

As hyperscanning studies multiply, the field needs shared conventions for:

• Representing dyadic/group relationships between participants
• Integrating subjective and phenomenological data alongside neural signals
• Handling multi-modal recordings (EEG + ECG + motion + subjective) in a unified structure

This pilot dataset and its accompanying pipeline aim to serve as a concrete, reproducible example that can inform future BIDS extensions for social and interpersonal neuroscience.

Key standardization challenges

Challenge	Our solution	Status
Dyad metadata — BIDS has no concept of a dyad (two linked participants)	dyads.tsv custom file (.bidsignore to pass validator)	🚧 Non-standard, needs community discussion
Subjective ratings per task — subjective ratings tied to specific interaction segments	Injected as a column in events.tsv per task	✅ BIDS-compatible via events extension
Peak connection moments — phenomenologically-anchored ~15s epochs	BIDS annotation in events.tsv (trial_type: peak_connection)	✅ BIDS-compatible
Padding — 2s buffer pre/post task for filter edge artefacts	onset = 2s convention, documented in sidecar JSON	🚧 Convention maison, not yet standardized
Excluded participants — set-aside subjects kept in participants.tsv for stable numbering	processing_status column in participants.tsv	🚧 Not in BIDS spec
XDF → BrainVision conversion — EGI/LSL format not natively in BIDS	Custom converter using pyxdf + pybv + MNE-BIDS	✅ Output is BIDS-compliant

Active community discussions

There is currently no official BIDS Extension Proposal (BEP) for hyperscanning — this is an open gap in the community. Here are the key ongoing discussions this project connects to:

• 📌 GitHub Issue #402 — "Hyperscanning data storage" (opened 2020, still open): the original BIDS discussion on how to represent dyadic data, never finalized due to lack of contributors → bids-standard/bids-specification#402
• 💬 Neurostars — "BIDS structure for longitudinal dyadic data" (2023): community discussion on using the acq entity as dyad identifier, validated by core BIDS developer Rémi Gau → https://neurostars.org/t/bids-structure-for-longitudinal-dyadic-data/26173
• 📖 BIDS Starter Kit FAQ — official BIDS guidance on hyperscanning (using ses-dyadic convention) → https://bids-standard.github.io/bids-starter-kit/FAQ.html
• 🧠 OHBM OSSIG / Brainhack Mattermost — #bids channel with 5000+ members → https://mattermost.brainhack.org

---

2 Goals for brainhack school 2026

🎯 Main goal: becoming autonomous in analysing and adapting my lab's pipelines to my data — and exploring the potential of visualizations!

1. PERSONAL GOAL = To explore inter-brain synchrony patterns between mothers and child (autistic or non-autistic) of the pilot EEG data (9 dyads) I want to to check if their neural dynamics relate to the subjective experience they reported. — So, also in correlation with likert scales. For this mission, the pipeline I am going to adapt to my data is not shared yet with the world, and comes from my PPSP lab

2. SHARED GOAL = To produce a bidsfication Notebook for hyperscanning + annotations of subjective data (that can be useful for researcher practicing neurophenomenology and hyperscanning)

---

Data

Data collected under approved ethics protocols; raw data not publicly shared

The protocole follows 10 tasks

Duration	Task
1 min	⬜ 👁️ Eyes open
1 min	⬜ 🙈 Eyes closed
2 min	🟦 🤚 Spontaneous imitation
2 min	🟩 🗣️ Day planning
1 min	⬜ 👁️ Eyes open
1 min	⬜ 🙈 Eyes closed
2 min	🟩 🗣️ Day planning
2 min	🟦 🤚 Spontaneous imitation
1 min	⬜ 👁️ Eyes open
1 min	⬜ 🙈 Eyes closed

• Participants: 9 pilot dyads (autistic and non-autistic child + mother)
• EEG: Dual 128 EGI HydroCel system (high density hyperscanning), continuous recording during the 10 tasks for LSL (LabRecorder) producing an xdf.
• Subjective measure: Inclusion of the Other in the Self scale (IOS, 1–7 Likert) after each of the 10 tasks (.tsv)

Pipeline overview

flowchart TD
    A[📂 Raw XDF files] --> B[🔍 Quality Check]
    B --> C[📋 BIDSification]
    C --> C1[🧠 EEG conversion\nXDF → BrainVision]
    C --> C2[📊 IOS ratings integration\nevents.tsv]
    C --> C3[📊 Personality Questionnaires ratings integration\nphenotype]

    C1 --> D[⚙️ Preprocessing\nFiltering · PyPREP · ICA]
    C2 --> D
    C3 --> D
    D --> D1[📏 Benchmarking\nQuality metrics validation\nPyPREP/Zapline+/ICA & AR parameters]
    D1 --> E[✂️ Epoching]
    E --> F[🎵 Frequency Band Analysis\nDelta · Theta · Alpha · Beta · Gamma]
    F --> G[🔗 Connectivity\nPLV · Transfer Entropy · AdjCircCorr]
    G --> H[📊 Integration-Correlation/Task\nConnectivity metrics × IOS correlation]
    H --> I[📈 Statistics\nMixed models · Permutation tests · ML classification]

    style A fill:#2d6a4f,color:#fff
    style B fill:#2d6a4f,color:#fff
    style C fill:#2d6a4f,color:#fff
    style C1 fill:#2d6a4f,color:#fff
    style C2 fill:#2d6a4f,color:#fff
    style C3 fill:#e07b00,color:#fff
    style D fill:#1a6b3a,color:#fff
    style D1 fill:#1a6b3a,color:#fff

Loading

---

Deliverables

1. ✅ Jupyter Notebook going through bidsification process for a multimodal hyperscanning project --> and a posteriori annotations is not treated here yet

For the goal of personal exploration goal:

1. ✅ Adapted EEG preprocessing pipeline for my pilot (not shared)
2. ⬜ Inter-brain connectivity analysis (PLV, transfer entropy) for 9 pilot dyads
3. ⬜ Statistical comparison of inter-brain synchrony between groups (autistic/non-autistic dyads) and across the 10 tasks
4. ⬜ Correlation between task-averaged inter-brain synchrony and IOS ratings per task

---

Skills, Tools & Methods

Status	Skill	Tools
✅	Agentic coding — using AI to assist pipeline adaptation	Claude Code
✅	Git & GitHub workflows — branching, pull requests, reproducible science	Git / GitHub
✅	Pipeline adaptation — adapting the SCAALE hyperscanning pipeline to my pilot dataset	ppsp-hyperscanning-pipeline
✅	BIDSification — phenotype annotations, a posteriori BIDS annotations	MNE-BIDS · PyBIDS · PyXDF
✅	EEG preprocessing — filtering, bad channels, ICA, line noise removal	MNE-Python · PyPREP · AutoReject · ICALabel · Zapline+
⬜	Hyperscanning connectivity — PLV, wPLI, transfer entropy, AdjCircCorr	HyPyP
⬜	Statistics for small datasets — mixed models, permutation tests	pandas · numpy · scikit-learn
⬜	Data visualization — connectivity maps, topographies, synchrony plots	matplotlib · seaborn

---

Visualization

When arrived at that stage, the next challenge of my project is to move from average data analysis per task (tasks of 1 or 2 minutes) to identify / vizualise the dynamics of the synchronisation (choice of metrics, directionality, switches, metastability, evolution of regimes of synchrony...)

---

References and acknowledgements

This work is part of a CIHR-funded project (2024–2028)

• reference list to be completed [1] Monnier, A., Adel, L., & Dumas, G. (2025). Now is the time: operationalizing generative neurophenomenology through interpersonal methods. Neuroscience of Consciousness, 2025(1), niaf052.