Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
3D Residual Networks (ResNet) for developmental brain age estimation in human fetal and infant MRI
The late fetal and neonatal stage is critical for brain development. In the clinical practice, it is often very difficult to quantify any lag or acceleration compared to the normal developmental trajectory. In some disorders, such as in intrauterine growth retardation or congenital heart disorders, estimating the “developmental brain age” would be of critical practical importance. 3D ResNets have previously been used to evaluate brain age in elderly subjects mainly in the context of Alzheimer’s disease. In early brain development, their use is not well documented. In this project, the thesis student will implement a 3D ResNet or similar architecture that estimates brain developmental age on 3D MRI data.
Keywords: machine learning, deep learning, ResNet, brain devleopment, MRI, fetal MRI, neonatal MRI, infants, medicine
In this project, the candidate will use super-resolution reconstructed 3D fetal MRI brain imaging data from normal and pathologically developing fetuses. Age of pregnancy will originate from ultrasound examinations. From this dataset, prenatal developmental brain age will be estimated. To estimate brain developmental age after birth, neonatal brain MRI data will be used. The developed ResNet or other architecture should be able to give an es-timate of age as well as provide a back-projection or saliency map of the features in the image that allowed the age estimation. Importantly, such a network should be well generalizable across datasets of different origin (e.g. different MRI scanner type).
In this project, the candidate will use super-resolution reconstructed 3D fetal MRI brain imaging data from normal and pathologically developing fetuses. Age of pregnancy will originate from ultrasound examinations. From this dataset, prenatal developmental brain age will be estimated. To estimate brain developmental age after birth, neonatal brain MRI data will be used. The developed ResNet or other architecture should be able to give an es-timate of age as well as provide a back-projection or saliency map of the features in the image that allowed the age estimation. Importantly, such a network should be well generalizable across datasets of different origin (e.g. different MRI scanner type).
To implement a deep residual network that estimates developmental brain age in fetal MRI and infant MRI
To implement a deep residual network that estimates developmental brain age in fetal MRI and infant MRI
PD Dr. Andrؘas Jakab, PhD
andras.jakab@kispi.uzh.ch
Center for MR-Research
University Children’s Hospital Zürich
and University of Zürich
PD Dr. Andrؘas Jakab, PhD andras.jakab@kispi.uzh.ch Center for MR-Research University Children’s Hospital Zürich and University of Zürich