This repository contains Jupyter Notebooks and data for the training of 2d U-net models on placental blood vessel and maternal/fetal blood volume data. In addition separate notebooks allow the application of the trained U-net models to 3d blocks of image data in order to produce segmented outputs. The notebooks can be run on the Google Colab platform if desired. The notebooks for prediction can be used separately with example data and pretrained models that are hosted on Zenodo
- A Jupyter Notebook which enables slicing of a 3d region of placental blood vessel data into image stacks in the 3 planes parallel to xy, xz and yz. This is followed by training of a 2d U-net on these images.
- A Jupyter Notebook which enables slicing of a 3d region of placental blood vessel data into image stacks in the 3 planes parallel to xy, xz and yz. This is followed by prediction of the segmentation of these images using a 2d U-net before recombining the images into 3d volumes. The image data is then rotated by 90 degrees around the 4-fold symmetry axis running perpendicular to the xy plane and the entire slicing and prediction process repeated again. This happens 4 times, before the 12 resulting output volumes are combined.
- A Jupyter Notebook which enables slicing of a 3d region of maternal/fetal blood volume data into image stacks in the 3 planes parallel to xy, xz and yz. This is followed by training of a 2d U-net on these images.
- A Jupyter Notebook which enables slicing of a 3d region of maternal/fetal blood volume data into image stacks in the 3 planes parallel to xy, xz and yz. his is followed by prediction of the segmentation of these images using a 2d U-net before recombining the images into 3d volumes. The image data is then rotated by 90 degrees around the 4-fold symmetry axis running perpendicular to the xy plane and the entire slicing and prediction process repeated again. This happens 4 times, before the 12 resulting output volumes are combined.