2. Existing tools encounter problems when applied to fetal data
No standardized method to process these data
We used deep learning methods to segment fetal brain from the
maternal compartment
Fetal functional MRI is rapidly emerging as an important tool to
study brain development
Build open source, semi-automated fetal fMRI preprocessing pipeline
3.
4. Resting-state functional MRI was obtained
for 207 fetuses
(gestational age 24-39 weeks, M=30.9, SD=4.2)
1,168 masks were manually created to train
a convolutional neural network
Train, validation, test split
129, 20, and 48 subjects
(855, 102, and 211 volumes)
5. 1. Lots of other tissues
2. Small size
3. Non standard
orientation
4. Motion!
7. Convolutional neural network to locate & segment the fetal brain
Quality check data
Iterative realignment (realign volume n and n+1)
Quality Check data
Normalize to age matched template
Quality Check Data
8. Average time to create a mask for a single
volume = several seconds
Raw
volume
Hand-
mask
Auto-
mask
Dice
Yale Test SubjectWSU Test Subject