# SynAirG MAISI Bronchoscopy Dataset

Synthetic-only bronchoscopy dataset generated from MAISI chest CT volumes and
packaged as synchronized virtual bronchoscope episodes.

## MAISI Instantiation

- Backend family: MAISI-compatible NV-Generate CT wrapper.
- Version tag: `rflow-ct`.
- Default infer template: `maisi_thoracic_airway_infer.json`.
- Region: chest.
- Anatomy request: trachea, airway, left lung upper lobe, left lung lower lobe, right lung upper lobe, right lung middle lobe, right lung lower lobe, heart.
- Default output size: 512x512x256 voxels.
- Default spacing: 0.75x0.75x1.22656 mm.
- Source mask mode: none for the bulk path; airway masks are derived after
  generation from the paired MAISI label map.
- Airway label request: 57x132 with support for
  MAISI output-label remapping metadata.

## Contents

- LeRobot/Open-H style `data/chunk-*/episode_*.parquet` files.
- Video observations under `videos/`: depth video, normal video, PPS video.
- Per-case medical artifacts under `metadata/cases/<case_id>/`: CT NIfTI, MAISI label NIfTI,
  airway mask NIfTI, mesh geometry, source scope path JSON, TUM trajectory, pose JSONL,
  rendered-condition MP4s, composite review MP4, and case description.
- In geometry-first batches, photorealistic RGB is intentionally deferred to a second sweep; the
  composite review marks the RGB panel as unavailable.

## Synchronization

Rows use relative timestamps starting at 0.0 seconds. `frame_index`, `timestamp`, video frames, and
TUM trajectory rows are generated from the same sampled bronchoscope path at 20 Hz.

## Task

navigate a flexible bronchoscope through synthetic human airways

## Conditioning and PBR-Based Inference

The depth, normal, PPS, and airway-mask streams are aligned to the same camera
path and timestamps. They can be converted into ControlNet-style image or video
conditions for PBR-based inference systems such as BronchoGen or a custom
SynAirG renderer. The paired mesh, centerline graph, and TUM pose file keep the
conditioning frames tied to physical airway geometry.

## Synthetic Use Notice

These cases are synthetic generated artifacts. They are not for clinical
interpretation, clinical deployment, autonomous diagnosis, or regulatory
submission.

Generated episodes: 1000