Features
AI-powered cardiac diagnostics. 49 conditions. 50+ API endpoints. Built for clinicians who need answers.
49-Condition Cardiac Detection
4 ML architectures. 25+ rule-based algorithms. 6 diagnostic phases. Your ECG gets analysed for 49 cardiac conditions — from AFib to Brugada syndrome — with confidence-calibrated results. No single model decides alone.
- CNN, Transformer, XGBoost, and ONNX Ensemble working in parallel
- Isotonic regression calibration for clinical-grade confidence scores
- Automatic rule-based fallback — analysis never stops
Real-Time WebSocket Analysis
Upload an ECG. Watch it process live. Our 9-stage pipeline streams every step via WebSocket. When we detect an arrhythmia mid-analysis, you get an instant alert. No polling. No refreshing.
- 9 processing stages from file validation to AI-powered report
- Instant arrhythmia alerts during processing
- Multiple clinicians can monitor the same analysis simultaneously
Multi-Format Clinical Reports
One analysis. Five formats. Patient PDFs with colour-coded findings. Clinician PDFs with full intervals and confidence scores. CSV, JSON, and FHIR R4 for EHR integration. AI-powered summaries by Claude — not templated filler.
- Patient PDF in 3 languages with actionable recommendations
- FHIR R4 bundles with LOINC and SNOMED CT coding — EHR-ready
- Bulk generation: up to 100 reports with ZIP download
Medical-Grade Signal Processing
Raw ECG data is noisy. Our pipeline removes baseline wander, powerline interference, and motion artifacts. Then 4 R-peak algorithms vote on every heartbeat. Accepts 5 input formats from any ECG device.
- Ensemble R-peak detection validated against MIT-BIH database
- Signal Quality Index: 0–100% with automatic poor-segment rejection
- Supports CSV, EDF, WFDB/PhysioNet, Raw Binary, and MATLAB
50+ REST API Endpoints
Build cardiac analysis into your product. 50+ FastAPI endpoints. Pydantic v2 validated responses. Batch up to 50 ECGs per request. FHIR R4 export for direct EHR integration. Full OpenAPI 3.0 docs.
- Async processing with Celery workers and WebSocket progress streaming
- FHIR R4 Patient, Observation, and DiagnosticReport resources
- Webhook callbacks and PPG wearable analysis endpoints
HIPAA & GDPR-Compliant Security
Patient cardiac data demands real security. Five auth methods. Row-level database isolation. Full HIPAA audit logging. UK GDPR with right-to-erasure. Automated PHI de-identification. Every access logged. Every token expires.
- JWT, TOTP 2FA, WebAuthn Passkeys, Google OAuth, Apple Sign-In
- Row-Level Security via Supabase — isolation at the database layer
- Role-based access: Patient, Clinician, and Admin tiers