A browser-based research platform for estimating, improving, and reporting the environmental footprint of radiology departments and clinical AI.
Research software · active development
CEDARS is an open research platform. A manuscript describing CEDARS has been submitted for peer review. The software, interface, defaults, and methods remain under active development and may evolve as validation and evidence improve. The CEDARS Score and EcoLabel are research outputs, not an external certification.
Live application: cedarsleaf.com
About the CEDARS Collaborative: international collaborators, institutions, and project background →
CEDARS is a multi-person, multi-institutional, international collaboration spanning radiology, clinical AI, sustainability, and health systems. Collaborator affiliations identify contributors and do not imply institutional endorsement or sponsorship.
CEDARS turns radiology sustainability data into a structured workflow:
| Step | Purpose |
|---|---|
| 1 · Input | Describe a Radiology Department or an AI model / informatics workload |
| 2 · Score & EcoLabel | Interpret the current footprint using the CEDARS Score, Rating, and standardized label |
| 3 · Improve | Test prospective changes separately from the current state and compare projected effects |
| 4 · Report (& Share) | Review reporting details, prepare CEDARS materials, and preserve or share the assessment |
No account or installation is required for the public site.
CEDARS deliberately separates:
- Clinical AI — how an AI system is used in a Radiology Department. Local use can add compute while also changing imaging operations, such as scan time, avoided studies, or contrast use.
- AI Model & Informatics — the lifecycle characteristics of the AI system itself, including model/task context, training, inference, compute location, deployment assumptions, provenance, and model comparison.
The technical model is described once; local Department use is configured separately.
CEDARS can estimate and contextualize equipment energy, utilization, energy/carbon per study, electricity emissions, modeled Scope 3 categories, storage/archiving, clinical AI, contrast-media and resource indicators, current practices, and projected intervention scenarios. Local measured values can replace defaults where supported.
CEDARS supports one selected model or like-for-like candidate comparison; single-task imaging AI, LLM/foundation-model, and agentic workloads; separate training and inference compute contexts; provenance; deployment-volume and amortized per-study reporting; and an AI Research Label. CEDARS does not predict model performance.
The Improve workspace separates current practice from future scenarios. Opportunities are ranked from the current assessment when CEDARS can model them; guidance-only AI/informatics checks remain explicitly non-quantified unless sufficient inputs exist.
Reporting is organized around:
- Review & finalize
- Prepare your CEDARS materials
- Preserve or share
Outputs include EcoLabels, reporting text, structured methods/reproducibility fields, and portable assessment files.
CEDARS is literature-informed, but not every model parameter has equally strong evidence. Some values are published measurements; others are transparent estimates, proxies, or illustrative defaults where direct evidence is limited.
- The authoritative assumptions, parameter provenance, evidence limitations, and full bibliography are maintained in
sources.md. - Relevant references are also linked directly from the public interface where they affect interpretation or guidance.
- Local measurements should replace defaults for reporting of record whenever available.
- Model performance is user-supplied and is not predicted by CEDARS.
Keeping the detailed bibliography in sources.md avoids duplicating partial reference lists that can drift out of date.
Normal calculator use is browser-local by default.
| Option | Purpose |
|---|---|
| Save on this device | Browser-local working copy |
| Complete CEDARS file | Portable, versioned .cedars.json assessment for backup or exact transfer |
| Reproducible link | Compact current calculator/current AI-model configuration; not a complete multi-model backup |
| Contribute this assessment | Optional research submission after explicit review and consent |
A reproducible URL is not private: anyone with the full link can open the encoded configuration. Names, email addresses, consent choices, and patient-identifiable information are not intentionally placed in the shareable URL.
Research contribution is separate from normal saving and sharing. Until the optional contribution service is deployed/configured, that action remains disabled.
The calculation logic is kept in testable modules rather than duplicated in the interface:
| Module | Responsibility |
|---|---|
calc.js |
CEDARS Score/Rating and regional carbon/cost helpers |
model.js |
Department fleet, dashboard, clinical effects, storage, and intervention calculations |
urlstate.js |
Validation and encode/decode of shareable calculator state |
ailabel.js |
AI model record → AI Research Label calculations and reporting fields |
aiRecords.js |
Canonical AI model records and Department-use persistence |
Vitest regression tests pin fixed inputs to known outputs, and the production build runs through GitHub Actions before deployment. package-lock.json is committed for reproducible dependency resolution.
Requirements: Node ≥22.12.
git clone https://github.com/takinci/cedars.git
cd cedars/frontend
npm install
npm run devOpen http://localhost:5173.
npm test
npm run buildCurrent frontend stack includes React 18, Vite 8, Chart.js, Lucide React, and Vitest.
sources.md— scientific assumptions, provenance, evidence limitations, and referencesCHANGELOG.md— user-facing changesROADMAP.md— active/planned workCITATION.cff— software citation metadata- About the CEDARS Collaborative — collaborators, institutions, and project background
cloudflare/README.md— optional research-contribution backend- GitHub Issues — bugs, feedback, and feature suggestions
The public About CEDARS page lists current collaborators. Displayed affiliations identify collaborators and do not imply institutional review, endorsement, or sponsorship.
A manuscript describing CEDARS has been submitted for peer review. Until a publication citation is available, please cite the software using GitHub's Cite this repository function, which reads CITATION.cff.
When the associated manuscript is published, the preferred article citation will be added to CITATION.cff.
- Software/source code: Apache License 2.0
- Content, methodology, documentation, and generated labels/badges: CC BY 4.0
Suggested content attribution: CEDARS (cedarsleaf.com), CC BY 4.0.
The names CEDARS, the leaf mark, and CEDARS Score / Rating identify this project; Apache-2.0 does not grant trademark rights.
CEDARS is a research and estimation tool. Outputs may include literature-based estimates, modeled values, proxies, and user-entered data. They are not medical, clinical, financial, regulatory, or certification advice and are provided as-is without warranty.
For scientific or operational reporting of record, review the assumptions in sources.md and replace defaults with appropriate local measured data where available.
Radiology sustainability · Clinical AI · Transparent environmental reporting