Detecting heart failure using wearables: a pilot study.

Area of Science:

• Biomedical engineering
• Cardiology
• Artificial intelligence in medicine

Background:

• Diagnosing heart failure (HF) solely through physical examination can be challenging.
• Wearable technology offers potential for enhanced bedside diagnostic capabilities.

Purpose of the Study:

• To investigate the efficacy of wristband-based wearable technology in facilitating accurate bedside diagnosis of heart failure.
• To develop and validate a machine learning model for HF prediction using wearable sensor data.

Main Methods:

• A cohort of 97 inpatients was monitored, collecting photoplethysmography (PPG) and accelerometry data via wristband at 128 Hz for 5 minutes.
• A machine learning classification algorithm was employed, utilizing extracted features of beat-to-beat variability and signal quality.
• HF diagnosis was confirmed through chart review, and the model was trained and tested using cross-validation.

Main Results:

• A support vector machine model demonstrated the best performance for HF diagnosis.
• Waveform-based features alone yielded an area under the curve (AUC) of 0.80, with 50% specificity at 90% sensitivity.
• Incorporating demographics, medical history, and vital signs improved the AUC to 0.87 and specificity to 72% (at 90% sensitivity).

Conclusions:

• A heart failure classifier was successfully developed using data from a wristband wearable in monitored inpatients.
• This study is the first to demonstrate an algorithm leveraging wristband technology for HF patient classification.
• The findings support the utility of wristband devices as adjunct tools for bedside HF evaluation and risk stratification.