Machine Learning for Medical Image Segmentation
Contact: Mazdak Abulnaga, email@example.com
Project PIs: Polina Golland and Justin Solomon
We are developing neural network models to automatically segment the placenta embedded in magnetic resonance images (MRI) of the uterus. The placenta is a critical organ that forms during pregnancy and provides nutrients to support the growing fetus. The UROP project will focus on developing an end-to-end convolutional neural network model to automatically segment placenta volumes. We currently have a large data set with over 100 labeled images. The student will experiment with state-of-the-art network architectures, data augmentation techniques, and adversarial training. There will be opportunity for temporal analysis as our dataset contains images acquired over time. The segmentation algorithm is an essential component of a larger pipeline that seeks to analyze placental biomarkers to identify pathology that affects the development of the fetus.