An Approach for Fetal QRS Complex Detection from Abdominal ECG Recordings

Abstract

Non-invasive fetal electrocardiographic (NI-FECG) monitoring can help to diagnose fetal distress and morbidity during pregnancy. In this study, we develop a method with signal quality assessment (SQA) and signal noise cancelling (SNC) for fetal QRS complex extraction from abdominal ECG recordings. The proposed approach consists of five steps: abdominal ECG (AECG) pre-processing; SQA for AECG based on sample entropy (SampEn); SNC for AECG based on multiple filtering algorithms; maternal ECG (MECG) subtraction; fetal QRS complexes detection. Sensitivity, positive predictive accuracy (PPV), and F1 score as the metrics are utilized to evaluate the performance of fetal QRS complexes extraction. Our exploratory results indicate that diagnostic sensitivity of 0.9422, PPV of 0.9284, and F1 of 0.9352 for all 75 AECG recordings of the 2013 PhysioNet/Computing in Cardiology Challenge Database (PCDB), respectively. Diagnostic sensitivity of 0.9624, PPV of 0.9631, and F1 of 0.9628 for the selected 68 AECG recordings of the PCDB, respectively. The results prove that our proposed approach will greatly improve the accuracy of detection of the fetal QRS complexes.