The ultimate integration of AI into ultrasound imaging promises to speed up the process, increase the precision and accuracy, and uncover novel methods to evaluate a pregnancy and prevent adverse outcomes.
Ultrasound plays a critical role in obstetrical care, helping evaluate fetal health and identify maternal risks like preterm birth and preeclampsia in millions of pregnancies annually in the USA. However, conducting and interpreting ultrasounds requires highly skilled operators and clinicians. By developing AI tools, the process of acquiring ultrasound images and interpreting them accurately could be significantly improved. AI can speed up image capture, improve interpretation accuracy, and even introduce new ways to assess fetal and maternal health using novel measurements and algorithms. Integrating AI into ultrasound practices promises not only to enhance efficiency but also to increase the precision of diagnoses and potentially prevent adverse pregnancy outcomes.
A collaborative initiative by Macon & Joan Brock Virginia Health Sciences at ±¬ÁÏ¹Ï and the School of Data Science aims to harness AI for this purpose. They plan to build a training dataset from a vast collection of ultrasound images gathered over the years. The AI system will analyze these images to develop predictive models, which will then be tested on new ultrasound data. The success of AI predictions will be measured against real-world postnatal outcomes. By applying these AI tools in clinical settings, the project aims to validate their utility, while offering significant benefits with minimal risk. The ultimate goal is to refine these tools to support healthcare providers in delivering more effective obstetrical care.
In the broader field, AI’s role in healthcare imaging is growing, with the FDA recognizing various AI-based devices. Although some AI tools have yet to see widespread use in the USA, interest from organizations like the NIH and the Gates Foundation highlights a demand for improved pregnancy outcomes, particularly in underserved regions. However, challenges remain in securing comprehensive datasets for AI training and testing. Overcoming these challenges involves collaboration among academic institutions, industry players, and nonprofit organizations to share knowledge and ensure the development of robust AI models for obstetrical care.
Rebecca Horgan, MD
Assistant Professor, Obstetrics and Gynecology; Macon and Joan Brock Virginia Health Sciences at ±¬ÁϹÏ