Using Fyron In Research
Fyron is intended for clinical data science, imaging AI, and manuscript-oriented workflows where analysis steps must be understandable and reproducible.
It is not a medical device, compliance platform, or replacement for institutional review. Users remain responsible for study approvals, privacy controls, de-identification, validation, and local governance.
Citation
If Fyron supports a study, cite the repository using the metadata in CITATION.cff. Include the package version and, when possible, the commit hash or release tag used for the analysis.
Privacy And Security
Fyron is local-first by default:
- no telemetry is sent on import,
- no data is uploaded unless you explicitly configure a client or CLI command,
- FHIR, DICOMweb, Teable, S3, and document downloads require user-provided endpoints or credentials,
- examples should use synthetic or de-identified data.
Security reporting and healthcare data cautions are documented in the GitHub security policy.
Public Repository Safety
Public issues, pull requests, examples, screenshots, and manifests should contain only synthetic or de-identified data. Do not share:
- credentials, bearer tokens, API keys, private endpoint URLs, or database DSNs,
- patient names, birth dates, medical record numbers, accession numbers, or free-text notes,
- patient-identifying DICOM tags, FHIR resource payloads, or institution-specific study identifiers,
- local paths that reveal private projects, users, hospitals, or unreleased study names.
When asking for help, reduce the problem to the smallest synthetic table, JSON resource, DICOM-like fixture, or folder structure that reproduces the behavior.
Reproducibility Artifacts
For reviewer-facing work, save these artifacts together:
| Artifact | Why it matters |
|---|---|
| cohort CSV or table export | defines who entered the analysis |
| cohort flow table | explains inclusion and exclusion |
| feature matrix schema/QC table | catches leakage, missingness, and drift |
| statistical result tables | preserves estimates, CIs, p-values, and corrections |
| figures | supports manuscript and supplement review |
| provenance manifest | records package version, inputs, outputs, parameters, and hashes |
Synthetic Examples
Fyron examples are designed to run without private systems. Use them to understand expected shapes before connecting real FHIR servers, DICOMweb endpoints, S3 buckets, or Teable workspaces.
Start with the Example Gallery for workflow-oriented tutorials.
Limitations
- Fyron does not certify clinical correctness.
- FHIR profiles, DICOM metadata, segmentation labels, and endpoint definitions can vary by institution.
- Statistical helpers produce inspectable tables, but researchers must choose appropriate designs and assumptions.
- Model validation helpers do not replace external validation or clinical review.
- Curate is a local beta review app. It is useful for folder-based image curation, but it is not an audit system for regulated clinical workflows.