A newly created tool employing computer vision and artificial intelligence (AI) could assist healthcare professionals globally in swiftly analyzing placentas at the time of birth, thereby enhancing care for mothers and newborns. The prompt detection of placental infections may allow for timely antibiotic treatment for both mothers and infants. This tool would be particularly valuable in resource-limited regions where pathology labs or specialists aren’t available, enabling doctors to quickly identify potential issues. In more advanced medical facilities, it could help to discern which placentas require further investigation.
Researchers from Northwestern Medicine and Penn State have developed a groundbreaking tool that utilizes computer vision and artificial intelligence (AI) to enable doctors to quickly assess placentas at birth, potentially enhancing both neonatal and maternal healthcare.
Published on December 13 in the journal Patterns, which featured the study on its cover, the research introduces PlacentaVision—a software program that evaluates a straightforward photo of the placenta to identify abnormalities linked to infections and neonatal sepsis, a critical condition affecting many newborns worldwide.
Dr. Jeffery Goldstein, a co-author of the study and director of perinatal pathology at Northwestern University Feinberg School of Medicine, said, “The placenta is a common specimen in labs. For a child in the neonatal intensive care unit, even seconds can influence medical decisions. With a diagnosis derived from these photos, we can achieve results days earlier than our traditional methods allow.”
Northwestern provided the most extensive collection of images for the study, with Goldstein overseeing the development and optimization of the algorithms.
Alison D. Gernand, the primary investigator on the project, came up with the concept for this tool through her work in global health, especially regarding cases where women give birth at home due to inadequate healthcare resources.
Gernand, who is an associate professor in the Penn State College of Health and Human Development, noted, “Failing to examine the placenta is a common issue that is often underestimated. This oversight means missing an opportunity to spot potential problems and intervene early, which can optimize outcomes for both mothers and their babies.”
The importance of timely placental examination
The placenta is crucial for the health of both the pregnant individual and the baby, yet it frequently goes unexamined right after birth, particularly in low-resource settings.
“This research has the potential to save lives and enhance health outcomes,” stated Yimu Pan, a doctoral candidate in the informatics program at the College of Information Sciences and Technology (IST) and the lead author. “It could also make placental examinations more accessible, benefiting future pregnancies, especially for higher-risk mothers and babies.”
Tools like PlacentaVision could allow doctors to promptly identify placental infections, enabling them to act quickly—such as administering antibiotics to the mother or newborn and closely monitoring the baby for infection symptoms.
The developers of PlacentaVision aim for it to be widely applicable in various medical contexts, according to the researchers.
“In areas with limited resources, such as hospitals without pathology labs or specialists, this tool could assist doctors in quickly identifying problems like infections from a placenta,” Pan mentioned. “In well-equipped facilities, it could eventually help prioritize which placentas need more detailed examinations, streamlining the process and ensuring critical cases receive the focus they require.”
James Z. Wang, a distinguished professor at Penn State and a principal investigator on the study, explained, “Before we can implement this tool on a global scale, we needed to address key technical challenges to ensure the model can adapt to various placental diagnoses and withstand different delivery settings, including diverse lighting conditions and imaging quality. It was essential for PlacentaVision to maintain accuracy even when many training images are sourced from an advanced urban hospital.”
Training the tool to analyze placental images
To train the software to analyze placental images, researchers utilized cross-modal contrastive learning—an AI technique that aligns and interprets the relationship between different data types, in this case, visual (images) and textual (pathological reports). They compiled a substantial and varied dataset of placental images and reports over 12 years, investigating how these images correlate with health outcomes and developing a predictive model for new images. Additionally, they created various image manipulation techniques to simulate different photographic conditions, ensuring the model’s robustness could be assessed effectively.
The outcome was PlacentaCLIP+, a resilient machine-learning model capable of accurately assessing placental photos for health risks. It underwent cross-national validation to verify consistent performance across different populations.
Researchers designed PlacentaVision to be user-friendly, enabling functionality through a smartphone app or integration into medical record systems so healthcare professionals can obtain quick assessments post-delivery.
The next step: A mobile app for healthcare providers
“Our future endeavors involve creating a user-friendly mobile application for healthcare providers that requires minimal training for use in clinics or hospitals with limited resources,” Pan stated. “This app would allow doctors and nurses to capture photos of placentas and receive immediate feedback, ultimately enhancing patient care.”
The team aims to increase the tool’s effectiveness by incorporating more placental features and relevant clinical data to refine predictions and contribute to research on long-term health. They also plan to test the tool in various hospitals to ensure its effectiveness across different settings.
Gernand remarked, “This innovation has the potential to revolutionize postpartum placental examinations, particularly in regions where such assessments are infrequently conducted. Enhancing accessibility in both low- and high-resource environments could transform neonatal and maternal healthcare by enabling early, tailored interventions that prevent serious health issues and improve outcomes for mothers and infants worldwide.”
This research received financial support from the National Institutes of Health National Institute of Biomedical Imaging and Bioengineering (grant R01EB030130). The project utilized advanced computing resources from the National Science Foundation-funded Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) program.
