The placenta is the first organ to develop and functions as the fetal lung, kidney, gut, skin, immune and endocrine systems. It is the cause of, and reflects changes from, most diseases in pregnancy, yet remains understudied. This career development proposal will train me in the tools and practice of digital pathology, while I apply them to the placenta with the hypothesis that there are reproducible, quantitative changes in the placenta that can be modeled and used to identify abnormalities via artificial intelligence (AI). I will create a publicly available atlas of microscopically normal placentas from throughout the 2nd and 3rd trimesters. Whole slide imaging will be performed on microscopic slides of placentas from the beginning of the 2nd trimester (13 weeks) through post-term (42 weeks). I will lead a team to annotate tissue type, structures, and cells. Algorithms will be trained to replicate the manual annotations. To study the changes in the placenta over time, automated measurements will be performed to identify changes in shape, size, and cellularity of placental structures that correlate with gestational age. This research can be used to develop a model of placental development and study prematurity. I will demonstrate detection of diseases of pregnancy, using preeclampsia (PreE) as an example. Placentas with microscopic changes classically seen in PreE will be scanned and annotated and algorithms trained and tested to identify them. Like many diseases of pregnancy, placental changes in PreE are variable and sometimes absent. Slides from PreE cases with no microscopic abnormalities will be scanned and examined using the quantitative parameters developed for normal placentas, testing the hypothesis that one or more of them will significantly differ between PreE cases and gestational age- matched controls. I am an Assistant Professor of Pathology at Northwestern University with an emerging focus in informatics and machine learning for diseases of pregnancy. The mentor for this project is Lee D.A. Cooper, PhD, an expert in digital pathology and machine learning. The co-mentor is David M. Aronoff, MD, an expert in maternal-child health. Mentor and co-mentor both have a history of NIH funding and graduating mentees to independence. The advisory committee consists of a digital pathology expert (Gutman), a pediatrician (Mestan) and a pathologist physician scientist (Yang). They have proposed an aggressive schedule of one-on-one meetings, coursework, seminars, and scientific meetings to supplement learning by doing the science. Completion of these studies will build my expertise in the application of machine learning to placental pathology while creating a new, publicly- accessible tool for the rapid assessment and understanding of organ structure and function with great potential to improve maternal-child health.
The placenta grows over the course of gestation from a single layer of cells to a complex organ that acts as the fetal skin, lung, gut, kidney, immune system, and endocrine system. This project will develop an online repository of placenta microscopic images over the course of gestation from normal placentas and one disease of pregnancy, preeclampsia. Using artificial intelligence to quantitatively describe the changes over time in normal placentas and those with disease could help understand preterm birth and diseases of pregnancy.
