Fetal R-peak detection using a swin transformer network with dynamic encoding and parallel decoding

The FHR Detection employing dynamic encoding and parallel decoding swin transformer-based fetal R-peak detection network (DPSTFR-Net) was developed to address this issue. A cascaded Sparse Low-Rank and Kernel Recursive Least Squares (CSKL) filter is used to pre-process the input signals in order to eliminate high-frequency noise and baseline interference. DPSTFR-Net with adaptive position encoding is used to combine the position information of multiple receptor fields in order to perform the binary classification. Additionally, the parallel decoder's contextual information is used to reduce the impact of incorrectly categorized data points. Accuracy, precision, Mean Average Error (MAE), and other metrics are assessed using the PhysioNet/Computing in Cardiology Challenge database (PCDB) and the abdominal and direct fECG database (ADFECGDB). The proposed approach obtained 97.02 % accuracy, 98.25 % precision, 97.35 % recall, 97.28 % F1-score, 0.10 TNR, 0.55 FPR, and 0.56 MAE on the PCDB dataset. The approach obtained 97.52 % accuracy, 98.25 % precision, 97.05 % recall, 97.28 % F1-score, 0.09 TNR, 0.60 FPR, and 0.66 MAE for the ADFECGDB dataset. The outcomes of the experiment show that the proposed method is capable of effectively estimating FHR from abdominal ECG data. Commercial applications, such as long-term maternal and fetal monitoring systems, can use the proposed paradigm.