A comprehensive system for DICOM classification, sorting, anonymization, and BIDS export

Developed at Karolinska Institutet
Department of Clinical Neuroscience, Neuroradiology

NILS classifies MRI series using six orthogonal axes, each backed by its own YAML-driven detector:

Specialized branch pipelines handle multi-output acquisitions (SWI, SyMRI, EPIMix/NeuroMix, MP2RAGE) — provenance detection runs first and routes to the correct branch.

• Extraction — Import DICOM metadata with adaptive batching and per-instance stack creation
• Sorting — 4-step pipeline: Checkup → Stack Fingerprint → Classification → Completion
• QC Review — Draft-based quality control with Cornerstone.js DICOM viewer and rules engine
• Anonymization — De-identify with configurable ID strategies and compression
• BIDS Export — Self-describing filenames, cross-cohort resolution, field strength filtering

• CSV import for subjects, events, diseases, observation types, and identifiers
• Longitudinal tracking: link imaging sessions to diagnoses and clinical timelines
• Preview/validate before applying any import

• Docker & Docker Compose
• 4GB RAM minimum (8GB recommended)

# Clone the repository
git clone https://github.com/NeuroGranberg/NILS.git
cd NILS

# Start services with your DICOM data
./scripts/manage.sh start --data /path/to/your/dicom/data

# Access the web interface
open http://localhost:5173

Full documentation available at: neurogranberg.github.io/NILS

• Concepts - Core data models and terminology
• Cohort Operations - Extraction, Sorting, Anonymization, Export
• Classification - The six-axis detection system
• QC & Viewer - Quality control and image review

# Start with localhost access (default - secure)
./scripts/manage.sh start --data /srv/dicom

# Mount multiple data directories
./scripts/manage.sh start \
  --data /srv/dicom/ct \
  --data /srv/dicom/mr

# Clean start with network access
./scripts/manage.sh start --clean --forward --data /srv/dicom

# Stop services
./scripts/manage.sh stop

SSH tunnel (recommended for default mode):

ssh -L 5173:localhost:5173 user@server
# Then open http://localhost:5173 locally

Tailscale (with --forward mode):

http://your-server.ts.net:5173

Environment variables in .env:

┌───────────────────────────────────────────────────────┐
│                   Docker Network                       │
│  ┌──────────┐  ┌───────────┐  ┌───────────────────┐  │
│  │    db    │  │ metadata  │  │     backend       │  │
│  │ postgres │  │    db     │  │  FastAPI + Async  │  │
│  └──────────┘  └───────────┘  └───────────────────┘  │
│       │              │                ▲               │
│       │              │       ┌────────┴────────┐     │
│       └──────────────┘       │    frontend     │     │
│     Dual Database System     │  Vite + React   │     │
│  (app state + DICOM metadata)│  + Cornerstone  │     │
│                              └─────────────────┘     │
└───────────────────────────────────────────────────────┘

# Frontend unit tests
./scripts/manage.sh test-frontend

# Backend unit tests
./scripts/manage.sh test-backend

MIT License - see LICENSE file.

If you use NILS in your research, please cite:

Chamyani, N. (2025-2026). NILS - Neuroimaging Intelligent Linked System. Karolinska Institutet, Department of Clinical Neuroscience. https://github.com/NeuroGranberg/NILS

Karolinska Institutet, Department of Clinical Neuroscience, Neuroradiology