Dynamic risk stratification of worsening heart failure using a deep learning-enabled implanted ambulatory single-lead...

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Devices

Background:

• Implantable loop recorders (ILRs) continuously monitor electrocardiograms (aECGs).
• The utility of aECGs from ILRs for identifying worsening heart failure (HF) remains unexplored.
• Early detection of HF exacerbations is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To develop and validate an AI algorithm using ILR aECGs to detect reduced left ventricular ejection fraction (LVEF ≤ 40%).
• To assess the capability of this AI algorithm to predict HF hospitalizations in a real-world cohort.
• To evaluate the performance of AI-driven aECG analysis in identifying patients at increased risk for HF exacerbation.

Main Methods:

• An AI algorithm (aECG-CNN) was trained on 35,741 aECGs from 2,247 patients to identify LVEF ≤ 40%.
• The algorithm's performance was validated using the area under the receiver operating characteristic curve.
• The aECG-CNN was applied to a cohort of 909 HF patients to predict HF hospitalizations over 30-day intervals.

Main Results:

• The AI algorithm successfully identified patients with LVEF ≤ 40%.
• AI-identified high-risk patients demonstrated a significantly higher risk of HF hospitalization within 30 days (HR 1.89, P=0.001).
• This predictive capability remained significant after adjusting for patient demographics (HR 1.88, P=0.002).

Conclusions:

• An AI algorithm trained on ILR aECGs can effectively detect reduced LVEF.
• This AI tool can identify patients at increased risk of HF hospitalization by monitoring changes in HF probability.
• AI-powered analysis of aECGs offers a promising approach for proactive HF management.