What the study found
The review found that machine learning (ML) methods performed better than traditional approaches for early diagnosis of placenta accreta spectrum (PAS), a condition where the placenta attaches and invades the uterine wall. The strongest results were reported for ultrasound- and magnetic resonance imaging (MRI)-based models.
Why the authors say this matters
The authors say early and accurate prenatal diagnosis of PAS is important because it is linked to lower morbidity, mortality, severe hemorrhage, and cesarean hysterectomy. The study suggests ML could improve diagnostic accuracy, consistency, and reduce human error.
What the researchers tested
The authors reviewed 14 studies on ML for early PAS diagnosis using ultrasound and MRI. They examined several model types, including linear, ensemble, deep learning, and hybrid approaches.
What worked and what didn't
Ultrasound-based models reached reported accuracy rates of 84.6% to 92.3%, with particularly strong performance from ensemble methods and deep dictionary learning. MRI-based approaches showed even higher performance, with texture analysis using k-nearest neighbors reaching up to 98.1% accuracy. The main challenges described were limited generalizability across populations and variation in imaging quality due to differences in equipment and patient demographics.
What to keep in mind
This is a review of 14 studies, not a single clinical trial. The abstract notes that further research is needed to address generalizability and standardization before widespread clinical use.
Key points
- The review found ML methods outperformed traditional approaches for early PAS diagnosis.
- Ultrasound-based models reported accuracy from 84.6% to 92.3%.
- MRI-based models reported accuracy up to 98.1% in one texture-analysis approach using k-nearest neighbors.
- The review included 14 studies and several ML model types, including ensemble and deep learning methods.
- Generalizability and imaging-quality differences were identified as major challenges.
Disclosure
- Research title:
- Machine learning improved early diagnosis of placenta accreta spectrum
- Authors:
- Daniel Waszczuk, Varsha Manikandan, Brendan L Wong, Frank Martin, Nitish Bhargava, Annika Mondal, Morgan Loy, Drake Strnad, Manikandan Panchatcharam, Gayathri Sadanala, Sumitra Miriyala
- Institutions:
- A.T. Still University, A.T. Still University, A.T. Still University, A.T. Still University, A.T. Still University, A.T. Still University, A.T. Still University, A.T. Still University, Saint Louis University, Truman State University, Truman State University, University High School
- Publication date:
- 2026-04-07
- OpenAlex record:
- View
Get the weekly research newsletter
Stay current with scholarly research without reading academic papers — one filtered digest, every Friday.