Deep Learning for Algorithmic Detection of Pulmonary Hypertension Using a Combined Digital Stethoscope and Three-lead Electrocardiogram
This prospective, observational study will evaluate whether synchronized heart sound (phonocardiogram, PCG) and three-lead electrocardiogram (ECG) recordings (entered as separate interventions in PRS, though collected together in practice) collected with the Eko CORE 500 can help screen for pulmonary hypertension (PH). Adults (≥18 years) undergoing clinically indicated transthoracic echocardiography (TTE) and/or right heart catheterization (RHC) will complete one study visit (\ 20 minutes). During the visit, study staff will obtain at least four 15-second CORE 500 recordings (aortic, pulmonic, tricuspid, and mitral areas). The clinical echocardiogram (and RHC, if performed) within ±7 days of the recordings will provide reference labels for the presence and severity of PH; de-identified demographic and clinical data may also be abstracted from the medical record. The primary objective is to develop and validate a software algorithm to detect PH and, where possible, stratify severity using noninvasive PCG+ECG signals. These recordings are investigational data acquisitions for algorithm development only; they are not diagnostic procedures and will not be used for clinical decision-making. Primary performance measures are sensitivity and specificity versus echocardiogram and RHC references. No clinical decisions will be based on the investigational algorithm, and no changes to standard care are required. The study plans to enroll up to \ 1,513 participants to obtain approximately 1,375 evaluable datasets across multiple outpatient sites.
• Age ≥ 18 years.
• Able and willing to provide informed consent.
• Clinically indicated transthoracic echocardiogram (TTE) or right heart catheterization (RHC) scheduled/performed within ±7 days of the study recording visit.