Individualization of Fetal Growth Assessment using Maternal Genetics and Explainable AI

ABSTRACT
This is a K08 award application for Dr. Nathan Blue, a Maternal-Fetal Medicine physician and young inves-
tigator pursuing translational and clinical research to improve risk stratification approaches to fetal growth re-
striction (FGR). A K08 award will provide him with the means to acquire critical skills in three key career devel-
opment areas: 1) programming skills to carry out analyses and visualizations (Unix, R, Python), 2) novel bio-
medical informatics approaches to quantify risk of adverse outcomes, and 3) interdisciplinary research leader-
ship and management. By acquiring these skills, Dr. Blue will fulfill his career goal of becoming an independent
investigator who can improve prenatal recognition of fetuses at risk of adverse outcomes. To pursue this goal,
Dr. Blue has assembled the mentoring team of Dr. Robert Silver (primary mentor), a Maternal-Fetal Medicine
physician and international authority on obstetric complications, Dr. Mark Yandell (co-mentor), a human genet-
ics scientist, expert in computational biology, and developer of medical risk assessment software tools, and Dr.
Martin Tristani-Firouzi (co-mentor), a Pediatric Cardiology physician and widely recognized leader in applica-
tion of new informatics tools to complex clinical problems such as congenital heart disease.
Fetal growth restriction (FGR) is a leading cause of preventable stillbirths, postnatal complications, and re-
sults in a lifelong increased risk of cardiovascular disease. Based on his own published data, Dr. Blue’s central
hypothesis is that current fetal assessment tools function poorly because they assume all fetuses should be
the same size and fetal growth ultrasounds are interpreted in isolation of other factors that could be useful to
inform risk. He will test this hypothesis by analyzing maternal genetic variants and using a novel explainable
artificial intelligence (AI) method to develop individualized prediction models for expected fetal growth and risk
of perinatal morbidity. This will uncover insights into normal fetal growth as well as produce a new neonatal
morbidity risk calculator. By pursuing the following aims, Dr. Blue will test his hypothesis and lay the ground-
work for refining his new tools prior to application to fetal growth in a prospective cohort (to be proposed in an
R01 application during the K08 award period). Specific Aim 1 will test the hypothesis that maternal genetic
information can be used to individualize birth weight prediction in uncomplicated pregnancies. Specific Aim 2
will test the hypothesis that genetics, specific clinical variables, and social determinants of health interact syn-
ergistically to increase the risk of poor outcomes in FGR, which can be captured by new explainable AI.
The proposed research is significant because despite FGR’s enormous global burden, current approaches
to fetal growth assessment continue to perform poorly, forcing clinicians and families to make plans without
appropriately individualized information. The proposed research is innovative because of its use of 1) maternal
genetic rather than clinical data such as height, weight, and race to predict healthy birth weight, and 2) explain-
able AI for risk stratification rather than black-box AI techniques that are too opaque for trustworthy application.

Grant Number: 5K08HD113835-02
NIH Institute/Center: NIH
Principal Investigator: Nathan Blue