Assessing the Generalizability of a Deep Learning-based Automated Atrial Fibrillation Algorithm

An Arrhythmia classification approach via deep learning using single-lead ECG without QRS wave detection

“…The results showed a sensitivity of 98.16% and specificity of 99.07% for VF, sensitivity of 90.45% and specificity of 99.73% for VT, sensitivity of 99.34% and specificity of 98.35% for NSR, and sensitivity of 96.98% and specificity of 99.68% for other rhythms, with corresponding accuracies. A. Argha et al utilized the PhysioNet/CinC Challenge 2017 dataset and developed a hybrid deep learning model capable of classifying ECG recordings into four classes by detected QRS wave: NSR, AFIB, other rhythms, and too noisy recordings [ 31 ]. Their proposed model achieved an average test F1-score of 0.892.…”