Smarter Ultrasound, Safer Births: The Science Powering Next‑Generation Pregnancy Monitoring
A leading cause of stillbirth is a failure of oxygen delivery to the baby. However, current tests cannot directly assess fetal oxygenation, which can lead to mistimed interventions because clinicians do not have access to all the information they need.
SADIE (in-Silico Assessment of pregnancy via Digital Integrated Environments) is a new computational modelling software which can use existing ultrasound technology and clinical data to predict fetal oxygen status in under 30 seconds. This data gives clinicians additional information, enabling them to make the best decision – not only reducing stillbirth but also preventing unnecessary early intervention and NICU stays.
SADIE was created by an international network which included researchers from Universities of Auckland, Manchester and Birmingham, University College London, Cardiff University, Swansea University, and the Rosalind Franklin Institute, funded by the Wellcome Leap In Utero program.
Now that the initial research project has come to an end, the team are raising additional funding to validate SADIE and move into clinical trials. Prof. Jo James, one of the project leads from the University of Auckland, recently attended the JP Morgan (JPM) conference in San Francisco to engage with early stage philanthropic and venture capital investors.
Professor Jo James said “The Wellcome Leap Emerging Breakthroughs event at JPM provided such a fantastic opportunity to showcase SADIE. It was so inspiring to feel the energy in the room, and engage with potential industry partners and investors who believe in the importance and urgency of ensuring everyone gets to take a healthy baby home.”
”In obstetrics, we are too often forced to make high-stakes decisions without truly knowing whether a baby is getting enough oxygen. SADIE is exactly the kind of breakthrough Wellcome Leap’s In Utero Program was designed to enable — integrating computational modelling, imaging science and clinical insight to generate real-time, actionable information. By extracting far greater value from routine clinical and ultrasound data, this approach has the potential to reduce stillbirth while also sparing families the consequences of unnecessary early intervention.”
The researchers at the Franklin became part of this international team through their expertise in semi-automated segmentation, which involves developing deep learning approaches to label different parts of the placental tissue in images.
Dr Michele Darrow said, “My team and I have been able to bring meaning to the large volumes of placenta imaging data we had by assigning a label to each pixel within each of the images. Through this process we are able to then quantify characteristics and describe the geometries of shapes within the image which can then be fed into the computer models.”
This labelled image data was used by colleagues to create digital placenta models that simulate physiology, and form a key part of SADIE’s ability to predict the oxygenation status of the fetus.