Atrial fibrillation detection with multiparametric RR interval feature and machine learning technique

Automatic screening of atrial fibrillation (AF) in the out-of-clinic environment is potentially an effective method for early detection of this life-threatening arrhythmia which is often paroxysmal and asymptomatic. Different technologies such as modified blood pressure monitor, single lead ECG-based finger-probe, and smartphone using plethysmogram signal have been emerging for this purpose. All these technologies use irregularity of RR interval (RRI) as a feature for AF detection. For real-time applications scalar feature is extracted from RRI signal and classified with a threshold. In this work, we have introduced multi-parametric RRI feature yielding a multidimensional feature vector. We used machine learning technique to learn the optimal decision boundary. The proposed method was tested with a publicly available landmark database. Initial experiments show promising AF detection performances comparable to those of state-of-the-art methods. Development and implementation of such a method in existing screen devices such a smartphone could be important for prevention of AFrelated risk of stroke, dementia, and death.