A prior knowledge-guided convolutional neural network model for predicting fetal brain age from MRI during second and...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Fetal Development

Background:

• Estimating fetal brain age is crucial for monitoring pregnancy health.
• Current methods may have limitations in accuracy and precision.
• Advanced computational models offer potential for improved fetal age assessment.

Purpose of the Study:

• To evaluate a convolutional neural network (CNN) model for fetal brain age estimation using MRI.
• To assess the accuracy and reliability of the CNN model in second and third-trimester pregnancies.
• To explore the potential of AI in optimizing prenatal care through precise fetal age prediction.

Main Methods:

• Utilized 407 T2-weighted fetal MRI scans (2310 stacks) from 22-39 weeks gestation.
• Employed a prior knowledge-guided CNN model for brain age prediction.
• Validated predictions against first-trimester ultrasonography, using Lin's concordance correlation coefficient (ρc) and R² score.

Main Results:

• The CNN model achieved a mean absolute error (MAE) of 4.62 ± 3.31 days.
• Demonstrated strong agreement with reference standards (ρc = 0.977) and high correlation (R² = 0.953).
• Outperformed existing methods in accuracy and segmentation, with significant statistical differences (P<0.001).

Conclusions:

• A knowledge-guided CNN model accurately predicts fetal brain age from MRI scans.
• The model's high accuracy and correlation suggest its utility in clinical settings.
• This technology has significant potential for enhancing pregnancy management and outcomes.