Deep Learning for Detection of Fetal ECG from Multi-Channel Abdominal Leads

In this paper, we propose to use a CNN-based approach for fetal ECG detection from the abdominal ECG recording. Our work flow contains a pre-processing phase and a classification phase. In the pre-processing phase, abdominal ECG waveform is normalized and segmented. Then, short-time Fourier transform is applied to obtain time-frequency representation. The 2D representation is sent to 2D convolutional neural network for classification. Two convolutional layers, two pooling layers, one fully-connected layer are used. The softmax activation function is used at the output layer to compute the probabilities of four events. The classified results from multiple channels are fused to derive the final detection according to the respective detection accuracies. Compared to the K-nearest neighbor algorithm, the CNN-based classifier has better detection accuracy.