NeuroTorch is a framework for reconstructing neuronal morphology from optical microscopy images. It interfaces PyTorch with different automated neuron tracing algorithms for fast, accurate, scalable neuronal reconstructions. It uses deep learning to generate an initial segmentation of neurons in optical microscopy images. This segmentation is then traced using various automated neuron tracing algorithms to convert the segmentation into an SWC file—the most common neuronal morphology file format. NeuroTorch is designed with scalability in mind and can handle teravoxel-sized images easily.

NeuroTorch provides several features for reconstructing neurons from optical microscopy images:

• A platform-independent volumetric data loader that can handle teravoxel-sized datasets
• A training framework that can handle various neural network architectures and paradigms
• A segmentation framework that can scale to teravoxel-sized datasets efficiently
• An automated tracing platform for converting segmentations to neuronal morphology files

NeuroTorch can be installed in many ways and used within a few minutes. To install NeuroTorch through pip, make sure Python 3.5 is installed and type in the command-line

$ pip install neurotorch

To install NeuroTorch through Docker, make sure Docker is installed and type

$ docker run -it gornet/neurotorch:v0.1.0 /bin/bash

Otherwise, NeuroTorch can be installed manually by typing in the command-line

$ python setup.py install

To get started with NeuroTorch, create the file hello-neurotorch.py and edit its contents to contain

from neurotorch.datasets.dataset import TiffVolume
from neurotorch.core.trainer import Trainer
from neurotorch.nets.RSUNet import RSUNet
from neurotorch.core.predictor import Predictor


def main():
    net = RSUNet()  # Initialize the U-Net architecture

    inputs = TiffVolume("inputs.tif")  # Create a volume containing inputs
    labels = TiffVolume("labels.tif")  # Create a volume containing labels

    trainer = Trainer(net, inputs, labels,  # Setup a network trainer
                      max_epochs=100, gpu_device=0)

    # Set the trainer to add a checkpoint every 50 epochs
    trainer = CheckpointWriter(trainer, checkpoint_dir='.', 
                               checkpoint_period=50)

    trainer.run_training()  # Start training

    outputs = TiffVolume("outputs.tif")  # Create a volume for predictions

    # Setup a predictor for computing outputs
    predictor = Predictor(net, checkpoint='./iteration_100.ckpt',
                          batch_size=5)

    predictor.run(inputs, outputs, batch_size=5)  # Run prediction

    outputs.save()  # Save outputs

if __name__ == '__main__':
    main()

Next, let us run this script by typing in the command-line

$ python hello-neurotorch.py

If everything worked out, you should have a file “outputs.tif”—TIFF file containing the network prediction—in your directory.

The NeuroTorch project is licensed under the BSD license.

This project was created by James Gornet. Significant features were rewritten from code provided by Nicholas Turner and Kisuk Lee.