Arrhythmia Early Warning Analysis Tool

Problem: Non-contact Early Screening

Traditional rhythm screening relies on ECG, wearables, or clinic visits. For home-based early risk observation, users often lack a low-friction, contact-free input path. This skill addresses that case: when a user provides a short facial video, it extracts heartbeat-related hemodynamic signals from the video and flags possible premature beats, atrial fibrillation, tachycardia, and bradycardia, then produces a structured risk reminder.

How It Works: Facial PPG Video Analysis

The skill processes a 10-30 second facial video captured by a standard camera. It uses subtle color variations and periodic motion to reconstruct facial photoplethysmographic signals, then analyzes heart-rate variability features for abnormality detection and risk grading. Input supports mp4, avi, and mov, up to 10MB, using either a local path via --input or a public URL via --url.

The workflow has three key steps:
- Prepare the video: keep the full face in frame, use even lighting, and avoid recording immediately after intense movement.
- Bind the user: a valid open-id is required to save and query warning reports; it must not be guessed or fabricated.
- Run the analysis: invoke scripts.arrhythmia_early_warning_analysis to output average heart rate, abnormality type, risk level, lifestyle advice, and medical follow-up guidance; historical reports are fetched through --list --open-id from the cloud API.

Boundaries: Reference Only, Not Diagnosis

This skill is useful for home screening, remote monitoring, and self-check reminders for high-risk users, but its output is an early risk warning reference only. It does not replace a professional ECG or diagnosis by a cardiology doctor. If the result indicates high risk, follow up with clinical testing. Historical report queries depend on the cloud interface and should not fall back to local memory files, which can cause missing or inconsistent records.