Functional brain image synthesis using the KDE or MVN distribution.
brainSimulator is a brain image synthesis procedure for data augmentation and standardization of evaluation of ML neuroimaging pipelines. It intends to generate a new image set that share characteristics given an original one. The system focuses on nuclear imaging modalities such as PET or SPECT brain images. It analyses the dataset by applying PCA to the original dataset, and then model the distribution of samples in the projected eigenbrain space using a Probability Density Function (PDF) estimator. Once the model has been built, anyone can generate new coordinates on the eigenbrain space belonging to the same class, which can be then projected back to the image space.
brainSimulator is now available via pypi and can be installed directly from:
pip install brainSimulator
Alternatively, download the package, uncompress and execute:
cd /path/to/uncompressed/brainSimulator/ python setup.py install
This allows to train the model once and then perform as many sample drawings as required.
#navigate to the folder where simulator.py is located import brainSimulator as sim simulator = sim.BrainSimulator(algorithm='PCA', method='mvnormal') simulator.fit(original_dataset, labels) images, classes = simulator.generateDataset(original_dataset, labels, N=200, classes=[0, 1, 2])
- Code: https://github.com/SiPBA/brainSimulator
- Documentation: https://brainsimulator.readthedocs.io