This repositories include python code for infant deep learning segmentation framework we developed.
First, we trained our model on a big dataset (dhcp dataset) using pretrain script on three main plain views (axial, coronal, and sagittal) and tested the performance of our 3D model using pretrain_validation code.
After that, we used train script for transfering our knowledge on a smaller dataset (MCRIB), where we used Leave One Out Cross Validation (LOOCV) technique. Again, we trained our model on three different plain views. Using save_prediction code, we saved predictions of our model on our dataset and then used test script to calculate accuracy and dice scores of results.
The codes are ready to use. You just need to install needed requirements and provide the paths to dataset as instructed by comments in the code. Requirements are nibabel, scikit-image, scikit-learn and pytorch, nn-common-modules.