Related Experiment Video
Updated: Aug 21, 2025
Human Fetal Blood Flow Quantification with Magnetic Resonance Imaging and Motion Compensation
Published on: January 7, 2021
Semi-Supervised Learning for Fetal Brain MRI Quality Assessment with ROI consistency
Junshen Xu1, Sayeri Lala1, Borjan Gagoski2
1Department of Electrical Engineering and Computer Science, MIT, Cambridge, MA, USA.
Abstract:
Fetal brain MRI is useful for diagnosing brain abnormalities but is challenged by fetal motion. The current protocol for T2-weighted fetal brain MRI is not robust to motion so image volumes are degraded by inter- and intra-slice motion artifacts. Besides, manual annotation for fetal MR image quality assessment are usually time-consuming. Therefore, in this work, a semi-supervised deep learning method that detects slices with artifacts during the brain volume scan is proposed. Our method is based on the mean teacher model, where we not only enforce consistency between student and teacher models on the whole image, but also adopt an ROI consistency loss to guide the network to focus on the brain region. The proposed method is evaluated on a fetal brain MR dataset with 11,223 labeled images and more than 200,000 unlabeled images. Results show that compared with supervised learning, the proposed method can improve model accuracy by about 6% and outperform other state-of-the-art semi-supervised learning methods. The proposed method is also implemented and evaluated on an MR scanner, which demonstrates the feasibility of online image quality assessment and image reacquisition during fetal MR scans.
Insights
This study introduces a semi-supervised deep learning method to automatically detect motion artifacts in fetal brain MRI scans, improving diagnostic accuracy and reducing assessment time.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neuroscience
Background:
- Fetal brain MRI is crucial for diagnosing abnormalities but is significantly hampered by fetal motion, leading to image artifacts.
- Manual assessment of fetal MR image quality is time-consuming and subjective.
Purpose of the Study:
- To develop a semi-supervised deep learning method for automated detection of motion-corrupted slices in fetal brain MRI.
- To improve the efficiency and accuracy of fetal MR image quality assessment.
Main Methods:
- A semi-supervised deep learning approach utilizing a mean teacher model with ROI consistency loss.
- The model enforces consistency between student and teacher networks, focusing on the brain region to detect artifacts.
Main Results:
- The method achieved approximately 6% higher accuracy compared to supervised learning methods.
- It outperformed existing state-of-the-art semi-supervised learning techniques on a large dataset.
- Online evaluation on an MR scanner demonstrated feasibility for real-time quality assessment.
Conclusions:
- The proposed semi-supervised deep learning method effectively detects motion artifacts in fetal brain MRI.
- This approach enhances diagnostic accuracy and offers a feasible solution for online image quality control during scans.
