By harnessing artificial intelligence to decode the 12-lead electrocardiogram, the project will enable precise ECG-based phenotyping of hypertrophic cardiomyopathy-accurately classifying septal, apical, and other morphologic subtypes-while simultaneously differentiating HCM from hypertensive heart disease, aortic stenosis, and other phenocopy disorders.
To overcome the twin bottlenecks of late detection and poor inter-centre reproducibility, the project leverages a large, multicentre historical cohort and anchors its pipeline on the 12-lead ECG-an inexpensive, ubiquitously available signal that can be captured in any department. Using deep-learning architectures augmented with attention mechanisms, we will develop (1) a discriminative model that separates HCM from phenocopies and normal hearts, and (2) an algorithmic framework that remains stable across devices and populations. Model governance will be embedded through version-controlled releases, cloud-edge deployment, and an "offline replay" evaluation loop, producing an end-to-end evidence chain that mirrors real-world clinical workflows.
Study Type
OBSERVATIONAL
Enrollment
15,000
Second Affiliated Hospital, Zhejiang University School of Medicine
Hangzhou, Zhejiang, China
RECRUITINGmodel diagnostic performance
Model performance was evaluated using calculated metrics including accuracy, sensitivity, specificity, and the area under the ROC curve (AUC).
Time frame: year 2
model diagnostic performance
The accuracy rate of the model's phenotype-specific classification for patients with different patterns of myocardial hypertrophy
Time frame: year 2
the model's generalizability
The model's diagnostic performance on the external, multicentre validation cohort, including overall accuracy, sensitivity, specificity, and area under the ROC curve (AUC).
Time frame: year 2
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.