Hybrid attention for automatic segmentation of whole fetal head in prenatal ultrasound volumes

Toward deep observation: A systematic survey on artificial intelligence techniques to monitor fetus via ultrasound images

“… CNN+ Transfer learning DCNN N/A 18 - 22 weeks 40 th week Classification ( Cuingnet et al., 2013 ) To help the clinician or sonographer obtain these planes of interest by finding the fetal head alignment in 3D US Random forest classifier Shape model and template deformation algorithm Hough transform 19 - 24 weeks. Classification segmentation ( Singh et al., 2021b ) To segment the fetal cerebellum from 2D US images U-NET +ResNet (ResU-NET-C) N/A 18 - 20 weeks Segmentation ( Yang et al., 2021b ) To detect multiple planes simultaneously in challenging 3D US datasets Multi-Agent Reinforcement Learning (MARL) RNN Neural Architecture Search (NAS) Gradient-based Differentiable Architecture Sampler (GDAS) 19 - 31 weeks Miscellaneous ( Lin et al., 2019b ) To detect standard plane and quality assessment Multi-task learning Framework Faster Regional CNN (MF R-CNN) N/A 14 - 28 weeks Miscellaneous ( Kim et al., 2019a ) To tackle the automated problem of fetal biometry measurement with a high degree of accuracy and reliability U-Net, CNN Bounding-box regression (object-detection) N/A Miscellaneous ( Lin et al., 2019a ) To determine the standard plane in US images Faster R-CNN Region Proposal Network (RPN) 14 - 28 weeks Miscellaneous ( Namburete et al., 2018 ) To address the problem of 3D fetal brain localization, structural segmentation, and alignment to a referential coordinate system Multi-Task FCN Slice-Wise Classification 18 - 34 weeks Classification segmentation ( Huang et al., 2018 ) To simultaneously localize multiple brain structures in 3D fetal US View-based Projection Networks (VP-Nets) ...…”

A Review on Deep-Learning Algorithms for Fetal Ultrasound-Image Analysis

“…A total of 285 3D US volumes is partitioned following five-fold crossvalidation, achieving an IoU of 0.63. A 3D U-Net is also used in [79], [80], [81] and [82] . In [79], it is used to segment the fetal cortical plate and measure the depth of the Sylvian fissure on a dataset annotated by expert clinicians using atlas.…”

“…Results obtained on a test set of 15 volumes are evaluated with two custom metrics. In [82], a 3D U-Net is trained to segment the whole fetal head. The 3D U-Net is combined with a hybrid attention scheme to enhance the feature maps.…”

Artificial Intelligence in Obstetric Ultrasound: An Update and Future Applications

“…The research included data from all trimesters, established the growth curve, and confirmed that evaluating results for each trimester separately was necessary, which was cost-effective and suitable for clinical settings which lacked experienced sonographers (16). On the other hand, to provide a basis of extraction of representative biometrics in the fetal head, Yang et al proposed a fully automated solution to segment the whole fetal head based on 3D ultrasound, which achieved a Dice Similarity Coefficient of 96% (17). The research team further investigated a general framework for automatically segmenting fetal anatomical structures in 2D ultrasound images and thus made objective biometric measurements available.…”