Fetal electrocardiography (FECG) is crucial for monitoring fetal health during pregnancy, particularly analyzing heart rate and QRS components. From the 28th week of gestation onwards, FECG extraction becomes challenging due to the formation of vernix caseosa, a non-conductive fatty layer forming on the fetal body that degrades the signal quality. This layer dissolves soon after appearing, but the dissolution rate varies among fetuses. Despite advances in modeling, the lack of real-world datasets hampers the benchmarking and development of robust algorithms for evaluating FECG during this critical gestational phase.
To address this gap, we introduce the Longitudinal Observation of Non-invasive Fetal ElectroCardiogram (LONGFECG) dataset, soon available on PhysioNet, designed to advance research in fetal health monitoring and signal processing. The dataset includes 9 pregnant participants each contributing at least three longitudinal sessions after 30 weeks gestation for a total of 40 recordings, collected under IRB approval.
Each session, lasting 40 minutes, utilizes a high-density electrode setup: 26 on the maternal abdomen, 4 on the back, and 2 on the chest. Signals were recorded at 1024 Hz using the g.tec HiAmp amplifier. The gestational age of participants at recording ranged from 30 to 38 weeks (34.1 ± 2.2). Participants body mass index ranged from 21.7 to 35.3 kg/m2 (28.5 ± 5.7), ensuring diverse physiological representation.
The longitudinal nature and the high-density electrode setup are the key features of this dataset, which enables the analysis of temporal changes in FECG morphology and signal quality as gestation progresses. The strategic encircling electrode configuration is designed to maximize the FHR detection accuracy to counteract signal loss from vernix caseosa. It also supports inter- and intra-subject variability studies. This dataset fills a critical gap in non-invasive FECG research, fostering innovation in non-invasive monitoring and potentially improving prenatal care.