Image processing in neuroimaging applications : Reference

Image processing operations are a central part of neuroimaging.

Images are homogenous arrays, where spatial relationships are important

Many different operations can be performed on images, and processing pipelines can be build combining them

Images can be efficiently and usefully represented as arrays

Arrays can be manipulated with numpy operations, and visualized using Matplotlib

Linear models are useful to represent the images in a concise and computationally expedient manner

Linearized models are used to fit more complex models using the same mathematical framework

There are three (at least) different ways to perform segmentation

The first uses the histogram of the pixel gray values

The second detects edges and fills between them

The SyN algorithm is a powerful algorithm for calculating diffeomorphic transformations between images

It’s particularly useful for registering images of brains