Fetal Heart Rate Classification Algorithm Based on LSTM-FCN Model

Fetal heart rate (FHR) monitoring during labor and delivery plays a crucial role in assessing fetal status. To enhance the diagnostic efficiency of fetal abnormalities, a hybrid model combining long short-term memory (LSTM) and one-dimensional convolutional neural network (CNN) is proposed for automatic FHR classification. The LSTM network is employed to extract temporal features of FHR, while the CNN operates concurrently to capture spatial features, thereby improving the efficacy of hidden feature extraction. Additionally, to mitigate the adverse effects of data imbalance, a sliding-window slice data augmentation strategy is utilized to prevent extreme classification scenarios. Comparative experiments are conducted between conventional models and the proposed hybrid model. The results demonstrate the superior performance of our model in terms of accuracy and confidence level, thus potentially aiding obstetricians in clinical diagnosis.