AI in Ultrasound

Expanding access to ultrasound with AI

Ultrasound is a versatile and increasingly more accessible early disease detection tool, providing real-time dynamic views of major organ systems. We are developing artificial intelligence (AI) models to make it easier to interpret important health information from ultrasound images. Our goal is to expand access to care in areas where access to trained sonographers is limited.

Ultrasound costs have decreased, but access challenges persist

Ultrasound costs have decreased, but access challenges persist

Advances in sensor technology have made ultrasound devices more affordable and portable, and they can now be integrated directly with smartphones. This could revolutionize healthcare by making ultrasound more accessible to people in under-resourced regions.

However, capturing and interpreting ultrasound is a complex medical imaging technique that requires years of training and experience, and there is a shortage of trained ultrasonography experts in these regions. This has real repercussions. For example, up to 50% of pregnant people in low-resource settings do not receive ultrasound screenings during pregnancy. This can lead to delays in diagnosis and treatment of pregnancy complications, which can have serious consequences.

some medical information with the accuracy of a skilled sonographer

Our research shows that AI can identify some medical information with the accuracy of a skilled sonographer

We are building AI models to expand access to ultrasound, allowing individuals with no background in ultrasonography to collect clinically useful ultrasound scans. In a recently published paper, we show that non-experts can use an easy-to-teach operating procedure called a blind sweep protocol to perform as well as trained ultrasonographers in determining important information like gestational age and fetal presentation. This could empower midwives, frontline health workers, or others to quickly gain important health insights based on ultrasound images.

Trained on thousands of de-identified ultrasound images

Our AI model learned by identifying key features and traits from thousands of fetal and breast ultrasound images. By understanding visual cues related to fetal age, fetal presentation, breast density, and more, this technology was able to show accuracy on par with trained ultrasonographers.

Photo of Dr Amber Watters

Validating our research with Northwestern Medicine

Beyond working on foundational, open-access research studies to show how this technology can be useful, we partnered with Northwestern Medicine to further develop and test these models. Working with diverse populations, we are exploring how to generalize our technology across various levels of experience and types of technologies.

Photo of Dr Amber Watters
Dr. Amber Watters

Assistant Professor of Obstetrics and Gynecology, Northwestern University

Obstetric ultrasound is necessary to optimize the health of birthing people and their babies, yet it remains unavailable to half of pregnancies worldwide. By developing tools using AI technology to highlight health risks, we hope to make the benefits of ultrasound available to more pregnant people, even when trained operators are not accessible.

Maternal Health Ultrasound=e30

Partnering with Jacaranda Health for Maternal Health Ultrasound

Jacaranda Health, a Kenya-based nonprofit focused on improving health outcomes for mothers and babies in government hospitals. In Sub-Saharan Africa, maternal mortality remains high, and there is a shortage of workers trained to operate traditional high-cost ultrasound machines. Through this partnership, we’re conducting research to understand the approach to ultrasound delivery in Kenya and explore how new AI tools can support point-of-care ultrasound for pregnant women.

Breast Ultrasound=e30

Partnering with Chang Gung Memorial Hospital for Breast Ultrasound

Mammograms are the gold standard for breast cancer screening, but they are not available to everyone due to high costs. Additionally, mammograms are not as effective in populations with higher breast density, which makes it difficult to detect cancer early. We are partnering with CGMH in Taiwan to explore whether our AI models can help with early detection of breast cancer using ultrasound.

Learn more about our work to expand access to maternal health ultrasound