Deep learning model for predicting gestational age after the first trimester using fetal MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Fetal Medicine

Background:

• Ultrasound-based gestational age prediction is accurate in the first trimester but declines in accuracy later in pregnancy.
• Accurate gestational age estimation is crucial for optimal prenatal care and management.

Purpose of the Study:

• To evaluate a deep learning model for predicting gestational age using fetal brain MRI.
• To compare the deep learning model's accuracy against biparietal diameter (BPD) measurements.

Main Methods:

• Retrospective study of 184 fetal brain MRIs (mean gestational age 29.4 weeks).
• Deep learning model trained and tested on T2-weighted MRI data.
• Gestational age reference standard from first-trimester ultrasound and last menstrual period.
• Comparison using Lin's concordance correlation coefficient (ρc) and Bland-Altman plots.

Main Results:

• Deep learning model achieved substantial agreement (ρc = 0.964) vs. moderate for BPD (ρc = 0.920).
• Both methods showed increased prediction differences with advancing gestational age.
• The deep learning model demonstrated a significantly smaller upper limit of prediction error (2.45 weeks) compared to BPD (5.62 weeks).

Conclusions:

• Deep learning accurately predicts gestational age from fetal brain MRI acquired in the second and third trimesters.
• This AI-driven approach offers potential benefits for prenatal care in pregnancies with limited early-trimester data.