Can Artificial Intelligence predict pregnancy complications?

The Multidisciplinary Digital Publishing Institute (MDPI) has developed an artificial intelligence-based program capable of predicting placenta accreta spectrum (PAS), a serious pregnancy complication that can lead to severe maternal bleeding and, in some cases, death. Placenta accreta spectrum occurs when the placenta attaches too deeply into the uterine wall and fails to detach after delivery. The result can be catastrophic postpartum hemorrhage, emergency hysterectomy, and significant maternal risk.

Despite advances in prenatal care, early detection remains a major challenge. According to the research, only about 30% of PAS cases are diagnosed before delivery, largely because the condition does not always appear clearly on standard ultrasound scans.

How the AI model works

To test the model, researchers analyzed ultrasound imaging data from 113 high-risk pregnant women, individuals considered vulnerable due to previous cesarean sections or other known risk factors.

The AI system was trained using a combination of ultrasound imaging data, history of prior cesarean deliveries, and the presence of placenta previa, a condition where the placenta partially or fully covers the cervix, increasing the risk of PAS.

The results: Promising accuracy

In early testing, the program delivered striking results. It correctly identified all confirmed cases of placenta accreta spectrum and ruled out the condition in 75% of pregnancies that did not involve PAS. Overall, 82% of women who tested positive were truly affected, while all women who tested negative were confirmed not to have the condition. For a complication that is frequently underdiagnosed, these figures represent a potentially significant clinical breakthrough.

Why early detection matters

Placenta accreta spectrum is not rare, and its incidence has been rising, partly due to increasing cesarean section rates worldwide. When detected early, doctors can plan specialized deliveries in equipped hospitals, assemble surgical teams in advance, and significantly reduce maternal risk. When missed, the condition can become an emergency within minutes of childbirth.

What comes next?

All participants in the study delivered at Texas Children’s Hospital between 2018 and 2025. While the sample size remains relatively limited, researchers say the early findings justify broader clinical trials.

If validated in larger studies, AI-assisted screening could become an additional layer of protection in maternal care, not replacing doctors, but enhancing their ability to detect high-risk conditions before they escalate. In a field where minutes can mean the difference between life and death, earlier warning could save lives.