Introduction: The fetal face provides essential information in the evaluation of congenital
malformations and the fetal brain function, as its development is driven by genetic factors at
early stages of embryogenesis. Three-dimensional ultrasound (3D US) can provide information
about the facial morphology of the fetus, but its use for prenatal diagnosis is challenging due to
imaging noise, fetal movements, maternal body mass index, lack of amniotic fluid, occlusions
and incomplete views that ...
Introduction: The fetal face provides essential information in the evaluation of congenital
malformations and the fetal brain function, as its development is driven by genetic factors at
early stages of embryogenesis. Three-dimensional ultrasound (3D US) can provide information
about the facial morphology of the fetus, but its use for prenatal diagnosis is challenging due to
imaging noise, fetal movements, maternal body mass index, lack of amniotic fluid, occlusions
and incomplete views that are often obtained. For these reasons, associations between craniofacial morphology and developmental disturbances have been established primarily in adults and,
only more recently, in children and newborns. Our aim is assessing craniofacial morphology
even earlier, by looking at images acquired before birth. Methods: We propose a fetal face
reconstruction algorithm from 3D US images based on the use of a novel baby 3D morphabel
model (3DMM). The deformable model used contains the statistical information of baby facial
geometry and it is fitted to the US mesh acting as a regularizer to reduce the noise present in
the 3D US. To assess if the estimated morphology accurately describes the actual anatomy, we
compare the prenatal reconstruction with the 3D face of the baby once he/she is born. To obtain the post-natal face we fit the same baby 3DMM to a set of 3 standard pictures taken from
specific viewpoints to allow a complete estimation of the facial geometry. Results: Promising
results are obtained as the baby 3DMM is able to remove the noise present in the 3D US meshes
and estimates realistic fetal faces. Also, the proposed technique is able to combine multiple
US scans to estimate the whole fetal face, enabling better estimations when the position of the
baby may obstruct the visibility of some facial parts. Conclusion: The proposed algorithm
has the potential to aid in-utero diagnosis, particularly syndromes or diseases in which facial
dysmorphology is considered an index of early developmental disturbance.
+