A deep residual learning model for classification of electrocardiogram recordings is proposed to detect atrial fibrillation. The linear ECG recording is converted into a 2-dimensional spectrogram with 3-fold scaling, and used as input for a 25-layer deep residual neural network. The neural network is evaluated on the dataset from PhysioNet/CinC Challenge 2017 and achieves a weighted F-measure of 86%.
atifkhurshid / dl-resnet-for-af-classification Goto Github PK
View Code? Open in Web Editor NEWDeep Residual Learning Model for ECG signal classification
License: MIT License