Fetal Ultrasound Brain Biometry: An Integrated Deep Learning Framework

The assessment of fetal scans during obstetric scanning in the second trimester plays a crucial role in the assessment of the fetus to ascertain the growth of the baby and the presence of anomalies. Though there are various parameters to consider in this period, the ones used to find gestational age using the Hadlock formula include Head Circumference (HC), Biparietal Diameter (BPD), Occipital Frontal Diameter (OFD), Femur Length (FL) and Abdomen Circumference (AC). However, this value is prone to error due to manual bias from the operator and the unavailability of additional data due to late follow-up or negligence. To address this issue, this paper deals with the automation of fetal brain biometry using Deep Learning (DL) models U-Net and CNN to measure the values. This model also utilizes Cephalic Index (CI) to verify the Gestational Age (GA) helping radiologists improve the performance of their scanning. The results demonstrated a strong correlation between the manual and AI values visually, having a Mean Squared Error (MSE) of 1.769 and 0.853 for BPD and OFD respectively.