HLM-Micro-ecg-v0 — RESEARCH ARTEFACT, NOT A MEDICAL DEVICE
4-class ECG anomaly classification (normal / brady / tachy / arrhythmia) in the HLM-Micro tier.
Not a medical device
This is a research proof-of-pipeline. It is not a medical device, not validated on real patient data, not cleared by any regulatory authority, and must not be used to make clinical decisions. Training data is synthetic, not patient-derived.
Access
Weights and inference code available via Early Access or commercial engagement for research collaboration.
Task
4-class classification on a single-lead ECG-like window.
| Class | Description |
|---|---|
normal | Regular sinus rhythm |
brady | Bradycardia — slower than normal |
tachy | Tachycardia — faster than normal |
arrhythmia | Irregular R-R intervals |
Headline
Normal / brady / tachy are learned cleanly (rate-based discrimination is easy). Arrhythmia is the hard class — overlap with natural beat-to-beat variability in the normal class is a known failure mode shared with clinical-grade classifiers. On the synthetic benchmark the model reaches ~84% overall, driven down by ~47% on the arrhythmia class.
Why it's here
To demonstrate the HLM-Micro architecture transfers to medical-adjacent 1-D time-series. Not to make clinical claims. Commercial path requires real MIT-BIH / PhysioNet data, regulatory workstreams, and clinical validation — a multi-year product effort, not a weekend release.
Out-of-scope
Explicitly:
- Clinical decision-making of any kind
- Patient-facing applications
- Any regulator-facing claim
- Any use that influences medical care
Intended use
- Reference for researchers exploring polynomial-Hopfield models on medical-adjacent time-series
- Demonstrator for the HLM-Micro architecture's cross-task applicability
- Starting point for a research collaboration with clinical partners
How to get started
For research collaboration: Contact us.
License
BSL 1.1 (Apache 2.0 in 2030) once Early Access is granted.