Noninvasive Pulmonary Capillary Wedge Pressure Estimation in Heart Failure Patients With the Use of Wearable Sensing and AI.

BACKGROUND

Remote hemodynamics-guided management of heart failure (HF) with implantable pulmonary artery pressure sensors has been shown to reduce HF hospitalizations. The widespread clinical adoption of this procedure is constrained by its invasive nature and high cost. We present a noninvasive technology based on a wearable sensor (CardioTag; Cardiosense) and machine learning (ML) for estimating pulmonary capillary wedge pressure (PCWP) in patients with heart failure with reduced ejection fraction (HFrEF).

OBJECTIVES

The authors developed and evaluated (against right heart catheterization [RHC]) an ML model to estimate PCWP with the use of electrocardiography, seismocardiography, and photoplethysmography signals from CardioTag.

METHODS

A multicenter prospective study was performed, and 310 patients with HFrEF (EF ≤40%) were recruited in both inpatient and outpatient settings. A blinded core laboratory adjudicated the RHC PCWP tracings to yield criterion-standard PCWP labels against which the model was trained and tested. The data were separated into 2 sets: a training set for model training and fine-tuning, and a held-out testing set unseen until final evaluation.

RESULTS

The patients were 61± 13 years of age, 38% female, 44% White, and 39% African American, and had a PCWP of 18.1 ± 9.45 mm Hg. The model estimated PCWP values in the held-out test set with error of 1.04 ± 5.57 mm Hg (limits of agreement of -9.9 to 11.9 mm Hg), with consistent performance across sex, race, ethnicity, and body mass index.

CONCLUSIONS

The CardioTag and its ML algorithm estimate PCWP with accuracy approaching implantable hemodynamic sensors, potentially offering a more accessible and cost-effective option for hemodynamics-guided management in HFrEF patients.