In this repository is contained the code of the paper https://www.biorxiv.org/content/10.1101/2020.08.29.272831v1 relative to the registration pipeline. The aim of this model is, given an histology image of a coronal slice of the mouse brain, to identify its "depth" into the brain by comparing it with an existing atlas (e.g. Allen CCF). The model is a Siamese Neural Network based on a pretrained DenseNet encoder of which we fine-tuned the final layers. The model works as depicted in the figure below.
In this repository you can find:
- The code used to train the model
- A pretrained version of the model itself
- A jupyter notebook containing some usage examples
To run the code you need Python >= 3.6 and the following packages installed:
tensorflow > 2.0
numpy
matplotlib
imgaug
pandas
pickle
umap-learn
scipy
You will also need to populate the brain_images folder with the mouse brain samples that can be found at: LINK
To train a new model edit the file main.py with the correct hyperparameters and run the command python main.py.
To test a trained model use the notebook Predictions.ipynb.