dev_plan.md

QC-Studio MVP Scope

Desgin Overview

Key requirements: datatypes / formats

• Support visualization of curated data

  • Raw BIDS (TBD: how to handle relative paths not starting from derivatives)
  • MRIQC

• Support visualization of processed data from these pipelines

  • fMRIPrep
  • Freesurfer / fsqc
  • QSIPrep
  • QSIRecon

Key requirements: UI

• Visualize 3D MRI using niivue
• Display flat image montages
• Display IQMs (optional for MVP)

Constraints

• User has full access to data either locally or via ssh
• QC UI is populated based on files listed in the pipeline_qc.json (fixed schema)
  • Does allow custom “qc-task” definitions.
• Only single base image and overlay in niivue panel
• Only SVGs or PNGs for montage viewer. No HTMLs.
• Only pass | fail | uncertain ratings supported

Tasks

• Configs

  • Generate pipeline_qc.json (see sample_qc.json)

• UI-orchestrator (see ui.py)

  • Handle argparse
  • Web-app init / global parameters
  • Calls to data handler
  • Calls to layout

• UI-data-handler

  • Write Pydantic json parser (see models.py)
  • Write data loaders (see utils.py)
    • MRI
    • SVGs
    • TSVs (IQMs)
  • Handle chunking for pagination (TODO)
    • n_subjects per page
    • n_QC tasks per page

• UI-layout (see layout.py)

  • Overall layout manager → handle 4 display panels
  • Niivue streamlit plug-in to support base + overlay images
  • SVG panel
  • IQM panel (optional for MVP)
  • Rating panel

• Write _qc_scores.tsv (see layout.py)

  • Handle overwrite / append

Pipelines to support

• mriqc
• freesurfer
• fmriprep
• qsiprep
• qsirecon
• agitation