Motivation: Fetal MRI is essential for monitoring development and clarifying inconclusive ultrasound results. Biometrics like estimated fetal weight, amniotic fluid volume, and placental volume indicate fetal and maternal health, yet assessing these currently requires time-intensive manual segmentation. Goal(s): Create an automated 3D deep-learning model for accurate fetal MRI segmentation, enabling precise volume and weight estimations for timely diagnosis and monitoring. Approach: Developed a U-Net-based neural network with attention mechanisms and multi-level feature extraction, trained on a dataset of 58 healthy pregnancies. Results: Achieved DSC scores of 95.06% for the fetal body, 95.50% for amniotic fluid, and 88.27% for the placenta, outperforming other segmentation networks. Impact: The Fetal MRI Segmentation Network (FetSegNet) enables precise fetal body, amniotic fluid, and placenta segmentation, enhancing clinical efficiency and supporting more accurate pregnancy monitoring, paving the way for improved maternal-fetal health diagnostics and a deeper understanding of fetal development.
Lim et al. (Tue,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: