Learning to map 2D ultrasound images into 3D space with minimal human annotation

Sonographers performing neurosonography require accurate understanding of fetal neuroanatomy, skill to interpret 2D planes in a complex 3D structure, and substantial clinical experience and training.

“Sonographers performing neurosonography require an accurate understanding of fetal neuroanatomy, the skill to interpret 2D planes in a complex 3D structure, and, consequently, substantial clinical experience and training [78].”

AI algorithms in fetal neurosonography enabled the development of guiding methods for skill training.

“the development of guiding methods for skill training [78,83]”

Yeung et al., Namburete et al., Yu et al., and Di Vece et al. proposed a plane localization system as a 3D reference space for locating 2D planes to improve acquisition of standard planes and facilitate anatomical orientation.

“The establishment of a plane localization system as a 3D reference space for locating 2D planes was proposed by Yeung et al., Namburete et al., Yu et al., and Di Vece et al. for improving the acquisition of standard planes and facilitating anatomical orientation for sonographers [78,80,82,83].”

The primary limitation of AI US imaging in fetal neurosonography was rapid anatomical development of fetal brain structures due to brain maturation, increasing head size and degree of ossification with rising GA.

“The primary limitation of AI US imaging in this topic was described to be the rapid anatomical development of fetal brain structures due to brain maturation, increasing head size and degree of ossification with rising GA [78,80,84].”

AI models designed for automatic biometric measurements are usually restricted to a specific range of GA and can fail in images of different GA.

“AI models designed for automatic biometric measurements are usually restraint to a specific range of GA and can fail in images of different GA [34,40,78].”