How ayushranjan-17/skinscan-ai works
SkinScan.AI is a point-of-care dermatology triage and explainable-AI application, not an autonomous diagnostic system. A clinician supplies a lesion image and patient context through the web app; the app requests backend analysis, which preprocesses the image, runs multimodal inference, and returns risk stratification and visual explanations. The interface presents findings and supports reports, scan history, and biopsy feedback. Training and evaluation are a separate supporting capability. Edges reflect sampled imports/calls or README workflow statements; unsampled details are not inferred.
Main entry points: App.jsx Scan Workflow · api.js Backend Client · app.py Analysis API · inference.py Inference Engine
Components 23
Clinician Web App
- Scan Workflow
src/App.jsx - Image & Context
src/components/DiagnosticWorkbench.jsx - Backend Client
src/utils/api.js - Demo Processing
src/utils/imageProcessing.js - Demo Cases
src/data/benchmarks.js - Guided Onboarding
src/components/GuidedOnboarding.jsx
Inference Service
- Analysis API
backend/app.py - Inference Engine
backend/inference.py - Image Preprocessing
ml_pipeline/dullrazor.py - Accounts & Scans
backend/database.py
Clinical Decision Support
- Risk Stratification
ml_pipeline/risk_stratification.py - Risk Summary
src/components/RiskGaugeCard.jsx - Verification Workflow
src/components/VerificationWorkflowCard.jsx - Visual Explanation
src/components/GradCamVisualizer.jsx - ABCD Analysis
src/components/AbcdAnalysisCard.jsx - Clinical Report
src/components/ClinicalReportModal.jsx - Scan History
src/components/ScanHistoryDrawer.jsx - Biopsy Feedback
src/components/HitlFeedbackPanel.jsx - Research Metrics
src/components/ResearchMetricsDashboard.jsx
Model Development
- Hybrid Model Training
ml_pipeline/train_hybrid_model.py - Model Evaluation
ml_pipeline/evaluate_metrics.py - Holdout Preparation
ml_pipeline/holdout_common.py
Other components
- Clinician
Connections 22
- Clinician → Scan Workflow: starts scan
README.md - Scan Workflow → Image & Context: collects input
src/App.jsx - Scan Workflow → Backend Client: requests analysis
src/App.jsx - Scan Workflow → Demo Processing: uses demo fallback
src/App.jsx - Image & Context → Demo Cases: offers demo cases
src/components/DiagnosticWorkbench.jsx - Backend Client → Analysis API: calls analysis API
src/utils/api.js - Analysis API → Inference Engine: runs inference
backend/app.py - Analysis API → Accounts & Scans: persists scans
backend/app.py - Inference Engine → Image Preprocessing: preprocesses image
backend/inference.py - Inference Engine → Risk Stratification: stratifies risk
backend/inference.py - Scan Workflow → Risk Summary: renders findings
src/App.jsx - Scan Workflow → Verification Workflow: renders workflow
src/App.jsx - Scan Workflow → Visual Explanation: renders explanation
src/App.jsx - Scan Workflow → ABCD Analysis: renders analysis
src/App.jsx - Scan Workflow → Clinical Report: offers report
src/App.jsx - Scan Workflow → Scan History: offers history
src/App.jsx - Scan Workflow → Biopsy Feedback: offers feedback
src/App.jsx - Backend Client → Accounts & Scans: retrieves history
- Guided Onboarding → Clinician: guides use
- Research Metrics → Model Evaluation: shows measured results
README.md - Hybrid Model Training → Model Evaluation: evaluates model
ml_pipeline/train_hybrid_model.py - Holdout Preparation → Model Evaluation: supplies evaluation data
ml_pipeline/evaluate_v4_holdout.py