FluoroSuite

A unified desktop application for benchtop aneurysm fluoroscopy work. It combines the live GigE Vision camera capture, recorded-run playback, and contrast residence analysis into a single PySide6 interface with three tabs:

• Capture — receives the forwarded raw GVSP stream, reconstructs 1024×1024 16-bit frames, applies window/level for viewing, and records each exposure to a .raw run with a JSON sidecar.
• Playback — reviews recorded runs with a scrub/transport bar and window/level.
• Analysis — measures aneurysm contrast residence from a circular ROI and inlet brightness from a fixed rectangular ROI.

The capture engine and on-disk recording format are compatible with the legacy Fluoro daemon, so existing recordings in captures/live/ open directly. Disable the standalone Fluoro live-capture daemon before running this suite, since both bind the same GVSP stream endpoint.

Run

uv sync
uv run fluorosuite

By default the app binds the raw GVSP stream on 10.200.0.1:5802 (the WireGuard endpoint on the workstation, matching the network topology). Override with:

uv run fluorosuite --stream-host 0.0.0.0 --stream-port 5802

Video export

Exported MP4 files use the first usable native H.264 encoder detected for the host: VideoToolbox on macOS, NVENC/QSV/AMF/Media Foundation on Windows, or NVENC/QSV/VAAPI/V4L2 M2M on Linux. Each candidate performs a short hardware encode probe before it is selected; libx264 is used when no GPU backend works. Hardware bitrates scale with frame size and rate from 100 Mbps at 1024×1024 and 30 fps to preserve the quality of the existing CRF 18 exports.

uv run fluorosuite-export --overwrite
uv run fluorosuite-export --encoder software
uv run fluorosuite-export --encoder videotoolbox

Use --encoder with auto, software, videotoolbox, nvenc, qsv, amf, mediafoundation, vaapi, or v4l2m2m. On Linux, set FLUOROSUITE_VAAPI_DEVICE when the desired render node is not the first /dev/dri/renderD* device.

Recording format

Each run is a flat file of concatenated 1024×1024 little-endian 16-bit (14-bit significant) frames, with a JSON sidecar describing geometry, frame count, and timing. The sidecar's data_file field names the corresponding CSV containing time_s, roi_mean, and inlet_mean; contrast and summary metrics are calculated when the data is loaded. Recordings are written to captures/live/.

Contrast residence

Iodinated contrast attenuates X-rays and appears dark, so the residence signal is measured inside the ROI as baseline − current mean brightness. The analysis reports baseline level, peak contrast, time to peak, and residence time.

Architecture

• fluorosuite/capture/ — GVSP stream reassembly (receiver.py) and per-exposure recording (recorder.py).
• fluorosuite/pipeline/ — frontend-neutral pipeline contracts and ROI analysis stages.
• fluorosuite/widgets/ — reusable frame view, visualization panel, transport bar, and collapsible stage drawer.
• fluorosuite/pages/ — the capture, playback, and analysis pages.
• fluorosuite/app.py — the tabbed main window and entry point.

Not for diagnostic use.