adalca/neurite

What it solves

Neurite is a neural network toolbox designed specifically for medical image analysis. It provides a set of specialized tools for handling complex spatial data, such as 3D volumes and flow fields, which are common in biomedical imaging.

How it works

Built on PyTorch (with a stable version available for TensorFlow), the library organizes its functionality into several core modules:

• Functional tools: Provides tensor operations for smoothing, resampling, interpolation, and spatial mathematics.
• Modules: Offers reusable stateful layers, loss functions, and preprocessing blocks.
• Models: Contains prebuilt network architectures that can be used across arbitrary spatial dimensions (1D, 2D, or 3D).
• Plotting: Includes specialized tools for visualizing tensor slices, volumes, and flow fields.

Who it’s for

This library is intended for researchers and developers working in medical imaging and biomedical segmentation, particularly those implementing unsupervised learning or anatomical priors in convolutional networks.

Highlights

• Support for arbitrary spatial dimensions (1D, 2D, 3D).
• Specialized tensor operations for spatial math and resampling.
• Specialized plotting tools for medical volumes and flow fields.
• Integration with major deep learning frameworks like PyTorch and TensorFlow.