A Career Development Award in AI-Assisted Ultrasound for Early Pregnancy Localization - PROJECT SUMMARY/ABSTRACT A Career Development Award in AI-Assisted Ultrasound for Early Pregnancy Localization This research addresses the challenge of evaluating women in early pregnancy presenting with signs and symptoms concerning for ectopic pregnancy, particularly in resource-constrained settings where access to pelvic ultrasound is limited. Despite pelvic ultrasound being the diagnostic mainstay for pregnancy localization, many low- and middle-income countries and rural areas of the United States face substantial barriers to this technology, owing to the high cost of equipment and need for trained staff. Consequently, ectopic pregnancy diagnosis may be missed or delayed, increasing maternal morbidity. Importantly, diagnosis of an intrauterine pregnancy effectively eliminates the possibility of an ectopic pregnancy, and achieving this diagnosis with transabdominal pelvic ultrasound reduces the need for transvaginal pelvic ultrasound assessment. This proposal outlines an innovative strategy to evaluate women presenting with non-specific symptoms of ectopic pregnancy. Point-of- care ultrasound (POCUS) devices offer a cost-effective alternative to traditional cart-based machines. Novel ultrasound collection procedures enable providers without formal ultrasound training to conduct evaluations. And deep learning artificial intelligence (AI) models analyze sonographic data to make diagnoses. My long-term career goal is to become an independent investigator leveraging evolving AI and ultrasound technology to improve obstetric outcomes. To achieve my career goals and objectives, I need additional mentorship and training in (i) applied data science with a focus on deep learning, (ii) clinical trials design and implementation, and (iii) research communication and collaborative global health partnerships. I will leverage this training and the resources of my mentors to achieve my scientific objective: to develop a novel strategy to evaluate early pregnancy localization using advancements in deep learning and ultrasound technology. I will pursue this objective through three specific research aims: (1) develop an AI ultrasound tool using transabdominal POCUS “blind sweeps” to diagnose intrauterine pregnancy; (2) develop an AI ultrasound tool to a) guide a clinician with minimal training to perform transvaginal ultrasound assessment and b) diagnose intrauterine and extrauterine pregnancy; and (3) pilot the use of these AI ultrasound models to localize early pregnancies in women presenting with signs and symptoms of ectopic pregnancy in both Lusaka, Zambia and Chapel Hill, North Carolina. I am well-positioned to achieve these aims given the vast institutional resources available to me through the University of North Carolina at Chapel Hill, along with an internationally-renowned team of mentors and advisors, including Drs. Jeffrey Stringer and Michael Kosorok, who have a history of collaboration in the burgeoning field of AI- assisted obstetric ultrasound. The proposed research is innovative and significant, and the results have the potential to significantly improve the diagnostic evaluation of ectopic pregnancy in the most vulnerable pregnant populations.
