Artificial Intelligence-Enabled Electrocardiography for Preoperatively Detecting Atrial Fibrillation and Mortality...

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Atrial fibrillation (AF), both pre-existing and new-onset (NOAF), is a significant perioperative risk factor.
• NOAF is linked to increased short-term mortality and adverse events after surgery.
• Current risk assessment methods for perioperative AF may be insufficient.

Purpose of the Study:

• To develop and validate an AI model for detecting hidden AF from sinus rhythm electrocardiograms.
• To assess the prognostic relevance of pre-existing AF and NOAF in non-cardiac surgery.
• To improve perioperative risk stratification using AI.

Main Methods:

• An AI model was trained and validated to detect hidden AF from ECGs.
• The model's predictive capability for NOAF and short-term outcomes was evaluated in patients without known AF.
• Conventional clinical risk scores were compared against the AI model's performance.

Main Results:

• The AI model achieved an AUC of 0.87 for AF prediction during development.
• Pre-existing AF and NOAF were significantly associated with increased 30-day mortality.
• High-risk patients identified by AI showed substantially higher 30-day mortality (HR 17.33) and outperformed clinical scores.

Conclusions:

• AI accurately identifies patients with elevated perioperative AF-related risk.
• This AI-based approach can facilitate targeted interventions to improve patient outcomes.
• AI enhances perioperative risk assessment for atrial fibrillation.