FetMRQC: A robust quality control system for multi-centric fetal brain MRI

Area of Science:

• Neuroimaging
• Medical Imaging Analysis
• Artificial Intelligence in Medicine

Background:

• Fetal brain MRI is vital for diagnosing developmental abnormalities.
• Image quality is often compromised by fetal motion and varied scanning protocols.
• Inconsistent data quality can negatively impact research findings.

Purpose of the Study:

• To introduce FetMRQC, an open-source machine learning framework for automated fetal brain MRI quality control.
• To develop a robust system that addresses data variability across different clinical settings and scanners.
• To improve the reliability and interpretability of fetal neuroimaging studies.

Main Methods:

• Utilized a machine learning approach, specifically random forests, for image quality prediction.
• Extracted an ensemble of quality metrics from unprocessed anatomical fetal brain MRI scans.
• Validated the framework on a large, diverse dataset of over 1600 manually rated T2-weighted fetal brain MR images from multiple clinical centers and scanners.

Main Results:

• FetMRQC demonstrated robust performance and generalization to unseen data, even with domain shifts.
• The framework's predictions were found to be interpretable.
• The system effectively predicts expert ratings for fetal brain MRI quality.

Conclusions:

• FetMRQC offers a reliable solution for automated quality control in fetal brain MRI.
• This tool enhances the robustness of neuroimaging, facilitating deeper insights into human brain development.
• The framework supports more dependable perinatal diagnosis and research through standardized image quality assessment.