A 3D U-Net based Network for Segmenting medical images of Fetal MRI in three Dimensions

This study introduces a U-Net based algorithmic framework designed to segment 3D MRI images of perinatal fetal brains from a cohort of 20 fetuses, with gestational ages ranging from 20 to 36 weeks. Furthermore, an optimization network based on U-Net is also proposed to enhance the segmentation efficiency of the proposed U-Net framework. The experimental results indicate that the resultant Dice Score metric, Hausdorff distance and Jaccard Index all showed notable enhancements when compared to the U-Net pipeline. The final results confirm that the proposed method is highly effective for segmenting fetal brains across various gestational ages, with the optimized model yielding superior segmentation quality as compared to the conventional U-Net model.