PI: Phillip LeDuc
University: Carnegie Mellon University
Industry Partner: Placenta AI LLC
Decidual vasculopathy in placental blood vessels is often found in cases of preeclampsia and has been correlated to morbidity of the mother of newborn in subsequent pregnancies. The long-term vision is to develop a pipeline by using explainable AI for automatic screening of placentas, which will help save lives through the start-up Pennsylvania-based company PlacentaAI to bring it into hospitals. Our proposal builds off our previous pilot work through examining placental images across different scanners prevalent in hospitals and building a transformer pipeline using RNN(LSTM/GRUs) and graphical neural networks to predict preeclampsia risk. This work aims to build a Self-Supervised Pipeline and streamline the screening to enable quick diagnosis. This will reduce the effort of managing huge number of placentas and provide jobs through PlacentaAI who already works with a multi-billion dollar company toward commercialization along with our local collaboration with Magee-Women’s Hospital which has ~10,000 births each year.