Related Experiment Video
Updated: Apr 10, 2026
06:56
Human Fetal Blood Flow Quantification with Magnetic Resonance Imaging and Motion Compensation
Published on: January 7, 2021
3.0K
Consistency Regularization Improves Placenta Segmentation in Fetal EPI MRI Time Series.
Yingcheng Liu1, Neerav Karani1, Neel Dey1
1Computer Science and Artificial Intelligence Lab, Massachusetts Institute of Technology, Cambridge, MA, USA.
Perinatal, Preterm and Paediatric Image Analysis (2025)
|April 9, 2026
Summary
This study introduces a semi-supervised learning method to improve 3D placenta segmentation in fetal MRI. The technique enhances accuracy and temporal consistency, aiding prenatal care and biomarker analysis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Fetal Development
Background:
- The placenta is vital for fetal development.
- Accurate placenta segmentation from fetal MRI is crucial for prenatal care.
- Existing methods may struggle with segmentation accuracy and temporal consistency.
Purpose of the Study:
- To develop an effective semi-supervised learning method for 3D placenta segmentation in fetal EPI MRI time series.
- To improve segmentation accuracy, especially for challenging cases.
- To enhance the temporal coherence of segmentation predictions.
Main Methods:
- Implemented a semi-supervised learning approach utilizing consistency regularization.
- Applied spatial transformation consistency within individual images.
- Ensured temporal consistency across adjacent images in the MRI time series.
Main Results:
- Achieved improved overall segmentation accuracy for the placenta.
- Demonstrated enhanced performance on outliers and difficult-to-segment samples.
- Showcased improved temporal coherency in segmentation predictions.
Conclusions:
- The proposed method effectively improves 3D placenta segmentation in fetal EPI MRI.
- Enhanced temporal consistency facilitates more accurate placental biomarker computation.
- This work advances placenta study and prenatal clinical decision-making.
