Research Project Title:
Manifold Learning for Fetal Heart Rate Detection
abstract:The fetal heart rate variability is an important indicator of the fetus’s health. Currently, the most accurate way to measure it is to put electrodes on the fetus’s head, but this is an invasive procedure. One noninvasive method is to detect the differences between two abdomen ECG signals, or ta-mECGs: due to the relative locations of the mother’s heart, the fetus’s heart and the recording sites, the maternal ECG signal is approximately the same for both recordings, but the perspectives of the fetal ECG signal are different. In this project, we aim to improve the identification of the fetal heart rate from ta-mECGs by using diffusion operators along with a neural network. We will test our approach first on toy examples that simulate an informative shared component and then ta-mECG signals.
I am participating in SuperUROP because I want to gain more experience with manifold learning. After becoming interested in manifold learning through the computational neuroscience classes I took, I started a UROP with this group last spring. I’m excited to extend my project into a SuperUROP so that I can hopefully learn even more about both manifold learning and the research process.