Proprietary AI Architecture
A 1D Swin Transformer foundation network built on hierarchical self-attention.
Rather than a single classifier, it uses Multi-Output Distributional Regression (Student-t) to simultaneously output continuous estimates and 95% confidence intervals for 35 biomarkers
Proven Data Moat
20 years of data from Seoul National University Hospital (SNUH), Seoul National University Bundang Hospital (SNUBH), and the Healthcare System Gangnam Center (SNUH-HSGC), combined with the US multi-ethnic critical care cohort (MIMIC-IV) — an exclusive dataset of 3,784,948 precision-matched ECG-Lab pairs
Clinical Performance
-
Creatinine
0.000 -
NT-proBNP
0.000 -
Hyperkalaemia
0.000 -
Anemia
0.000 -
CRP
0.000
aUROC
Patent Moat
2 domestic patent applications filed (No. 10-2026-0013451, No. 10-2026-0012502)
1 international PCT patent application filed (PCT/KR2026/007494)
Building an exclusive rights barrier for "Simultaneous Multi-Biomarker Inference from a Single ECG Waveform and Two-Track Central Control"