OpenScar is an AI-native research and workflow framework for scar and keloid care, designed to turn longitudinal clinical observations into structured datasets, reusable agent skills, and clinician-reviewable decision support assets.
Scar and keloid care is common, visual, longitudinal, and highly variable. The same lesion may change across months of injections, laser treatment, surgery, radiotherapy, pressure therapy, silicone therapy, or observation. Yet most clinical records do not capture scar morphology, treatment exposure, image quality, patient-reported symptoms, and recurrence outcomes in a reusable structure.
OpenScar starts from a practical thesis:
Plastic surgery can build durable AI assets by standardizing scar phenotype data before training ambitious models.
The first milestone is not a diagnostic product. It is a public, cautious, and extensible starter kit for building scar-focused datasets, workflows, and research tools.
- Scar and keloid outcomes are difficult to compare across clinics because images, timing, scales, and treatment documentation are inconsistent.
- Recurrence risk is often discussed qualitatively, while structured longitudinal risk factors are rarely captured in a machine-readable format.
- Scar photos contain useful signals, but they are vulnerable to lighting, angle, distance, color calibration, edema, makeup, and healing-phase bias.
- Molecular scar biology is advancing through single-cell and spatial data, but it is rarely connected to clinic-ready phenotype frameworks.
| Module | Purpose |
|---|---|
| ScarBank | Minimal structured schema for patients, lesions, visits, images, treatments, and outcomes. |
| ScarPhoto | Standardized image capture and quality-control workflow for scar documentation. |
| ScarTrajectory | Longitudinal representation of lesion change across time and interventions. |
| KeloidRisk | Research-only recurrence and treatment-response modeling workspace. |
| ScarClaw | Agent skills for literature review, intake, follow-up, photo standardization, and response review. |
| ScarAtlas | Future bridge from clinical phenotype to single-cell, spatial, and fibrosis biology resources. |
- Roadmap: staged build plan from GitHub asset to dataset, demo, and manuscript.
- Project Page: static GitHub Pages-ready overview for OpenScar Starter.
- Data Schema: minimum viable schema for structured scar and keloid data capture.
- Mock Data Guide: synthetic case files and validation workflow for the first schema check.
- Demo Workflow: end-to-end mock case workflow and report generation.
- Pilot Registry SOP: retrospective 20-50 case workflow for governed registry creation.
- Case Abstraction Form: human-facing checklist mapped to the six-table schema.
- Registry CSV Templates: empty six-table CSV headers for pilot data entry.
- Prospective Intake Pack: clinic-ready capture workflow for new scar and keloid visits.
- Clinician Intake Form: structured baseline, follow-up, treatment, and outcome capture.
- Patient Symptom Questionnaire: patient-reported pain, itch, satisfaction, function, and appearance fields.
- Photo Capture Checklist: standardized ScarPhoto image capture and metadata workflow.
- OpenScar Data Dictionary: field-level dictionary for REDCap, Excel, or registry setup.
- Working Agreement: step-completion rule for agent-assisted project development.
- Literature Seed: first literature-map categories and search prompts.
- Skill Map: 10 planned OpenScar skills for clinical, imaging, research, and operations workflows.
- Public Positioning: GitHub-ready one-liner, audience, and safety positioning.
- Release Notes: suggested v0.1 release description, topics, and next release candidates.
- ScarClaw skill pack:
- Build a standardized scar/keloid registry template for a plastic surgery team.
- Convert retrospective cases into a clean research table for recurrence or treatment-response analysis.
- Validate the first schema against synthetic mock data before using any real case material.
- Launch a governed 20-50 case retrospective pilot using the Pilot Registry SOP.
- Start prospective scar and keloid capture using the Prospective Intake Pack.
- Create agent workflows for scar photo standardization, follow-up intake, and evidence-aware literature review.
- Prepare a public project page and methods paper around AI-native scar phenotyping.
OpenScar Starter is not a medical device, diagnostic system, treatment recommendation engine, or substitute for licensed clinicians. It is a research and workflow-support scaffold. Any patient-specific interpretation, diagnosis, treatment selection, escalation decision, or consent process must remain under qualified clinical supervision.
- Review the schema against 3-5 real scar and keloid charts inside an approved internal environment.
- Choose 20-50 retrospective cases for a pilot registry.
- Use the case abstraction form to measure which fields are available retrospectively.
- Test the prospective intake forms on 10 new clinic cases under approved governance.
- Test the ScarClaw skill pack on synthetic or approved de-identified retrospective cases.
Run the synthetic mock dataset validator from the repository root:
node open-scar-starter/scripts/validate-mock-data.mjsValidate the OpenScar skill pack:
node open-scar-starter/scripts/validate-openscar-skills.mjsValidate the empty registry CSV templates:
node open-scar-starter/scripts/validate-registry-template.mjsValidate the REDCap/Excel-ready data dictionary:
node open-scar-starter/scripts/validate-data-dictionary.mjsGenerate the synthetic end-to-end demo report:
node open-scar-starter/scripts/generate-demo-report.mjsThe generated report is written to demo/mock-case-workflow.md.
Suggested repository description:
AI-native scar and keloid phenotyping framework with registry schema, synthetic mock data, validation scripts, and clinician-reviewable workflow skills.
Suggested topics: scar, keloid, plastic-surgery, clinical-ai, medical-ai, phenotyping, registry, agent-skills, ai4medicine, digital-health.