Related Experiment Video
Updated: Jun 28, 2026
10:44
Three-Dimensional Phase Resolved Functional Lung Magnetic Resonance Imaging
Published on: June 21, 2024
A two-stage deep learning model with segmentation-guided top-K slice selection for patient-level PAS prediction on
Chengzhi Song1, Shuiyan Chen2, Wei Ye3
1Department of Gynecology, Affiliated Hospital of Guangdong Medical University, Zhanjiang, 524001, China.
BMC Medical Imaging
|June 26, 2026
Summary
This study presents a novel deep learning framework for placenta accreta spectrum (PAS) prediction using MRI. The model shows promise as an auxiliary tool for screening and risk stratification in obstetric care.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence in Medicine
Background:
- Placenta accreta spectrum (PAS) poses significant obstetric risks, necessitating accurate preoperative assessment via MRI.
- Challenges in deep learning for PAS include varied MRI appearances and the need for detailed placental annotations.
Purpose of the Study:
- To develop and validate a two-stage, segmentation-guided MRI framework for automated placental localization, region of interest (ROI) selection, and patient-level PAS prediction.
- To overcome limitations of heterogeneous MRI manifestations and high annotation costs in deep learning for PAS.
Main Methods:
- A 2D U-Net model performed placental segmentation for ROI extraction.
- A ResNet-18 model classified ROIs, aggregating slice-level predictions for patient-level PAS diagnosis.
- Internal (170 patients) and external (54 patients) cohorts were used for validation.
Main Results:
- The framework demonstrated reliable placental segmentation.
- Patient-level PAS prediction achieved an AUC of 0.75 in the internal test set and 0.72 in the external validation cohort.
- External validation showed 0.78 sensitivity and 0.56 specificity for PAS detection.
Conclusions:
- The developed segmentation-guided MRI framework offers a reliable approach for placental segmentation and moderate external performance in PAS prediction.
- The model shows potential as an auxiliary screening and risk stratification tool for PAS.
- Further prospective, multicenter validation is recommended prior to clinical implementation.