i4h-catheter-navigation-digital-twin

von nvidia

Erstellen Sie einen digitalen Zwilling der patientenspezifischen Gefäßstruktur aus CT (Vorverarbeitung + Segmentierung). Verwenden Sie dies, wenn Sie aufgefordert werden, CT vorzuverarbeiten, Gefäße zu segmentieren, die Mittellinie zu extrahieren oder vorzubereiten…

npx skills add https://github.com/nvidia/skills --skill i4h-catheter-navigation-digital-twin

i4h Catheter Navigation - Digital Twin

Purpose

Download or locate a CT volume, preprocess it to an attenuation cache, and segment the arterial tree into vessel mask + centerline - the vasculature digital twin required for patient-specific viewport and DRR runs.

Base Code

ROOT="${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>/dev/null)}"
if [ ! -d "$ROOT/workflows/catheter_navigation" ]; then
  ROOT="${I4H_WORKFLOWS:-$HOME/i4h-workflows}"
  [ -d "$ROOT/workflows/catheter_navigation" ] || git clone https://github.com/isaac-for-healthcare/i4h-workflows "$ROOT"
fi
export I4H_WORKFLOWS="$ROOT"; cd "$ROOT"

Basics

  • Output cache layout: --output-dir / --ct-dir (e.g. /tmp/ct_cache) holds mu_volume.npy, metadata.json, and after segmentation vessel mask + centerline artifacts.
  • Contrast-enhanced CTA subjects work best; TotalSegmentator small subset (~3.2 GB) is the documented public dataset.
  • Comply with the dataset license; no patient data is committed to the repo.

Run

Run the steps below in order. Each step is a separate bash call; variables persist in the local agent's tmux session.

Step 1 - resolve paths

REPO_ROOT="${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>/dev/null)}"; [ -d "$REPO_ROOT/workflows/catheter_navigation" ] || REPO_ROOT="$HOME/i4h-workflows"
WF_ROOT="${REPO_ROOT}/workflows/catheter_navigation"
RUN_DIR="${WF_ROOT}/runs/digital_twin_$(date +%Y%m%d_%H%M%S)"
mkdir -p "${RUN_DIR}/logs"
ln -sfn "${RUN_DIR}" "${WF_ROOT}/runs/.latest"

# User-supplied or downloaded subject directory (must contain ct.nii.gz + segmentations/)
SUBJ="${SUBJ:-}"
CACHE="${CACHE:-/tmp/ct_cache}"

if [ -z "${SUBJ}" ] || [ ! -f "${SUBJ}/ct.nii.gz" ]; then
  echo "digital-twin: set SUBJ to an extracted TotalSegmentator subject (got '${SUBJ:-<unset>}')." >&2
  echo "Example: SUBJ=/path/to/Totalsegmentator_dataset_small_v201/s0011" >&2
  exit 1
fi

Step 2 - download dataset (skip if SUBJ already exists)

Only run when the user has no CT data yet.

curl -L "https://www.dropbox.com/scl/fi/pee5yxebfxrhz007cbuy5/Totalsegmentator_dataset_small_v201.zip?rlkey=osvfk02jc4lw5gr6uhrldtb9e&dl=1" \
  -o "${RUN_DIR}/Totalsegmentator_dataset_small_v201.zip"
unzip "${RUN_DIR}/Totalsegmentator_dataset_small_v201.zip" -d "${RUN_DIR}/Totalsegmentator_dataset_small_v201"
ls "${RUN_DIR}/Totalsegmentator_dataset_small_v201"
# Then set SUBJ to one extracted subject before continuing.

Step 3 - preprocess CT

"${REPO_ROOT}/i4h" run catheter_navigation preprocess_ct --local \
  --run-args="--nifti ${SUBJ}/ct.nii.gz --output-dir ${CACHE} --save-hu" \
  2>&1 | tee "${RUN_DIR}/logs/preprocess_ct.log"

Step 4 - segment vessels

"${REPO_ROOT}/i4h" run catheter_navigation segment_vessels --local \
  --run-args="--ct-dir ${CACHE} --ts-gt-dir ${SUBJ}/segmentations" \
  2>&1 | tee "${RUN_DIR}/logs/segment_vessels.log"

Verify

test -f "${CACHE}/mu_volume.npy"
test -f "${CACHE}/metadata.json"
ls -la "${CACHE}"

Notes

  • SUBJ must point at one extracted subject with ct.nii.gz and segmentations/ (TotalSegmentator layout).
  • CACHE is reused by [[i4h-catheter-navigation-viewport]] and cache-based [[i4h-catheter-navigation-render-drr]].
  • Segmentation is CPU/GPU mixed and may take several minutes depending on volume size.

Prerequisites

  • [[i4h-catheter-navigation-setup]] completed (imports and CLI work).
  • A CT NIfTI and matching vessel segmentations (or TotalSegmentator subject).
  • = 32 GB RAM recommended for large volumes.

Limitations

  • Does not ship data; user must download or provide their own CT.
  • Zenodo mirror is throttled; prefer the Dropbox URL in Step 2.

Troubleshooting

  • Error: SUBJ unset or missing ct.nii.gz - Fix: download Step 2 dataset or set SUBJ to an existing subject path.
  • Error: segment_vessels fails on --ts-gt-dir - Fix: confirm ${SUBJ}/segmentations exists (TotalSegmentator ground truth).
  • Error: out of memory during preprocess - Fix: use a smaller subject or increase swap; close other GPU/CPU workloads.

Final Response

Report CACHE path, key artifacts present, log paths under RUN_DIR, and recommend [[i4h-catheter-navigation-viewport]] or [[i4h-catheter-navigation-render-drr]] next.

Mehr Skills von nvidia

compileiq-debug
nvidia
Verwenden, wenn etwas nicht stimmt: Search() hängt, alle Evaluierungen geben INVALID_SCORE zurück, Scores verbessern sich nicht, jede Konfiguration liefert dieselbe Zahl, ptxas-Fehler…
create-github-pr
nvidia
Erstelle GitHub-Pull-Requests mit der gh CLI. Verwende, wenn der Benutzer einen neuen PR erstellen, Code zur Überprüfung einreichen oder einen Pull-Request öffnen möchte. Auslöser-Schlüsselwörter -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Scannt andere offene Issues, um solche zu finden, die ein bestimmter PR möglicherweise ebenfalls behebt oder versehentlich kaputt macht. Gibt benachbarte Fix-Möglichkeiten und Widerspruchsrisiken mit Datei:Zeile… aus.
fhir-basics
nvidia
Bringt Agenten bei, wie FHIR R4 APIs funktionieren, welche Ressourcen verfügbar sind, wie man sie mit Suchparametern abfragt und wie man alle Antwortformate korrekt parst…
compileiq-validate-result
nvidia
Verwende NACH Abschluss einer Suche und VOR dem Einfordern eines Speedups oder dem Versand eines ACF. Lädt die dump_results CSV, extrahiert Top-K-Kandidaten (Einzelziel)...
changelog-audit
nvidia
Auditiere die CHANGELOG.md vor einem Release: stelle verlorene Einträge wieder her, sortiere nach Benutzerauswirkung, verfeinere die Sprache der Einträge, führe Zeilenumbrüche durch und (im Release-Branch-Modus) erhöhe die Vergleichsnummer…
maintain-dynamic-plugins
nvidia
Verwalte NeMo Relay dynamische Plugin-Lader, Manifeste, Rust native SDKs, gRPC Worker-Protokoll, Python Worker-SDK, Dokumentation, Tests und Abdeckung des Release-Workflows
dgx-diagnose
nvidia
Diagnostizieren Sie häufige DGX Station GB300-Probleme – CUDA-Abstürze, falsche GPU-Zuweisung, vLLM/SGLang-Container-Fehler, MIG-Status-Probleme, NVLink/Fabric-Manager-Fehler,…