paroxysmal atrial fibrillation event detection by ECG signal processing an‎d artificial intelligence

11/17/2026 12:00:00 AM

Abstract: Atrial fibrillation (AF) is one of the most common cardiac arrhythmias. With increasing age, the risk of developing this disorder rises, consequently leading to a higher likelihood of heart failure. Accurate detection of atrial fibrillation remains a major challenge due to its transient an‎d unpredictable nature, emphasizing the need for an automatic an‎d precise identification method. In this study, a deep learning model was developed for the detection of paroxysmal (momentary) atrial fibrillation. The proposed method can distinguish normal heart rhythms from atrial fibrillation rhythms an‎d classify ECG recordings into three categories: normal rhythm, persistent AF, an‎d paroxysmal AF. The designed model consists of convolutional blocks, a long short-term memory (LSTM) network, an‎d a Transformer encoder. The dataset used for training an‎d eva‎luation was obtained from the CPSC2021 Challenge, containing dynamic ECG recordings. In the record-based eva‎luation, the best model performance for atrial fibrillation detection achieved an accuracy of 96.7%, sensitivity of 94.1%, specificity of 94.7%, precision of 96.7%, an‎d an F1-score of 94.7%. In the rhythm-based detection, the results reached an accuracy of 98.4%, sensitivity of 97.3%, specificity of 98.4%, precision of 98.6%, an‎d an F1-score of 98.4%. For paroxysmal atrial fibrillation detection in record-based testing, the model achieved an accuracy of 97.6%, sensitivity of 86.9%, specificity of 98.9%, precision of 90.0%, an‎d an F1-score of 88.3%. In rhythm-based testing, the model achieved an accuracy of 98.8%, sensitivity of 95.3%, specificity of 99.1%, precision of 92.6%, an‎d an F1-score of 95.7%. The results demonstrate that the proposed model provides high accuracy an‎d stability (±0.1%) in distinguishing atrial fibrillation rhythms from normal rhythms an‎d can serve as an effective tool for the automatic detection of paroxysmal atrial fibrillation. Keywords: Heart Failure – Atrial Fibrillation – Paroxysmal Atrial Fibrillation – Deep Learning

نارسايي قلبي , فيبريلاسيون دهليزي , فيبريلاسيون دهليزي لحظه اي , يادگيري عميق

Heart Failure , Atrial Fibrillation , Paroxysmal Atrial Fibrillation , Deep Learning

seyed navid alavi

Dr. Ali Sadr