Non-Invasive Detection of Fetal Ischemia Through Electrocardiography

During pregnancy, fetal distress requiring clinical intervention can be difficult to accurately monitor and diagnose, necessitating technological improvements to bring clear information to patients and their care teams. Non-invasive fetal Electrocardiography (NI-fECG) monitoring may allow for earlier and more reliable detection of global cardiac ischemia due to hypoxia. However, the lowsignal-to-noise ratio of the fetal heartbeat relative to the maternal heartbeat remains a challenge. To enable reliable recognition of ischemia in NI-fECG, we propose an approach that combines simulating a pregnant torso and an unsupervised machine-learning method. Three stages of fetal cardiac ischemia: none (healthy), moderate, and severe, were introduced to the model. For each case, Electrocardiograms (ECGs) were simulated with the standard 12 leads, plus 3 additional abdominal leads. Unsupervised Multiple-Kernel Learning (MKL) with k-means clustering identified changes consistent with fetal cardiac ischemia despite noise from the parental heart. Thus, in this early proof-of-concept investigation, our results suggest that NI-fECG may offer a means for detecting global cardiac ischemia.