An AI algorithm was developed to detect atrial fibrillation (AF) from ECGs taken during normal sinus rhythm and was evaluated in patients with palpitations.
The study involved 1403 patients from emergency departments, with a focus on those who had follow-up ECGs or Holter monitors within a year, excluding those with a prior diagnosis of AF.
The algorithm predicted new AF occurrences with a sensitivity of 79.1% and specificity of 66.1%, indicating it has potential but limited clinical usefulness in populations with a low incidence of AF.