Journal: European Heart Journal. Digital Health
Article Title: Dynamic risk stratification of worsening heart failure using a deep learning-enabled implanted ambulatory single-lead electrocardiogram
doi: 10.1093/ehjdh/ztae035
Figure Lengend Snippet: Dataset creation for the ambulatory electrocardiogram-convolutional neural network model development. Schematic indicates the strategy to obtain robust and reliable dataset for model development. To avoid cross-contamination, no patient data are repeated among training, validation, and testing datasets. *Patient count ( n = 5829) selected based on several inclusion, exclusion criteria to ensure the quality of LVEF data as they are obtained from Optum ® EHR and the dataset was captured in Optum ® EHR via natural language processing of procedure/diagnostic notes and prone to natural language processing errors.
Article Snippet: To ensure robust and reliable LVEF information and minimize errors introduced through automated data extraction from Optum ® EHR , we only included patients with prior HF diagnosis to create the low ejection fraction (EF) data (LVEF ≤ 40%).
Techniques: Biomarker Discovery, Diagnostic Assay