Fetal ECG Signal Analysis with Bi-directional Long Short-Term Memory Networks for Neonatal Applications

Fetal electrocardiogram (FECG) signals are essential for tracking the health of the fetus's heart because they offer important information about heart rate trends, fluctuations, and possible anomalies that can call for prompt medical attention. Noise and artifacts, on the other hand, present a serious problem since they frequently degrade signal quality and result in inaccurate diagnoses. Reliable fetal monitoring requires precise FECG signal categorization and efficient noise suppression. Throughout pregnancy and labor, continuous FECG analysis along with clinical assessments is essential to guaranteeing the best possible health for both the mother and the fetus. This study employs deep learning techniques to automate the classification of FECG signals as normal or abnormal using a Bidirectional Long Short-Term Memory (BiLSTM) classifier. The model processes FECG signals directly, without pre-processing, and achieves robust performance metrics: 91.98% accuracy, 90% precision, 85% recall, 87.5% F1 score, and 97% specificity. These results highlight the classifier's reliability and its potential as a valuable tool for real-time fetal health monitoring in clinical settings.