Brain Atrophy and Lesion Segmentation

Introduction

The Brain Atrophy tool can be used to measure brain tissues volumes, or the changes in volume that occur over time. It can separately measure the volumes of grey and white matter, and of the whole brain parenchyma. The tool works with T1-weighted images collected at one or more time-points. In addition, if FLAIR images are available, GM and WM volumes can be assessed in the presence of hyperintense lesional tissue and the lesion volumes computed. Even if hyperintense lesions are not present, use of a FLAIR image may improve tissue classification, although we have not shown this to be the case.

The program uses a template image which is segmented probabilistically into grey-matter, white-matter and CSF. This template is registered to the input T1-weighted images and used as the prior probabilities in an Expectation-Maximization (E-M) segmentation algorithm. If lesions are to be segmented, the segmentation happens after the first E-M classification. A further E-M classification is then performed after excluding the lesions from the GM, WM and CSF tissue classes.

The input images must:

  1. Have an axial orientation.
  2. Cover the whole cranium, but without excessive neck within the range of slices.

The picture below illustrates the ideal coverage, from the top of the cranium to the base of the cerebellum.

Ideal converage of the cranium for the Brain Atrophy tool

Note: in general, the use of de-faced images is not recommended, since the analysis pipeline involves registration of a template image. Using defaced images as input may cause problems for the registration.

If the images are not in an axial orientation, you can use the Image Resampler or Multi-Planar Reconstruction to re-orient them to axial. If the image slices extend into the neck or well beyond the top of the cranium, you can crop the images using either the Image Resampler or Multi-Planar Reconstruction.

The Brain Atrophy tool is the most computationally demanding of Jim's suite of sub-programs. It requires a lot of both processing power and computer memory. We recommend a system with at least 16 GB of memory, and a modern multi-core processor such as an Intel Core i7 or AMD Ryzen 7 or 9. Even then, you can expect long processing times if you have multiple time-points to analyse with white-matter lesions.

Start the Brain Atrophy tool from the Toolkits menu: Launching the Brain Atrophy tool This will bring up:

The Brain Atrophy tool

Input Images

You set the number of time points to be analysed using the Spinner to set the number of time points to be analysed spinner. For cross-sectional analysis, set this to 1. For longitudinal analysis, set this to 2 or more.

Set the number of image contrasts using the Spinner to set the number of image contrasts to be analysed spinner. If you have just T1-weighted images, set the number of contrasts to 1; if you also want to use FLAIR images, set the number of contrasts to 2.

This will alter the size of the N Time-Points × M Contrasts matrix of image selection panels below. For each image selection panel in the matrix, you must select the appropriate input image (T1-weighted image in the first row, and [optionally] FLAIR in the second row) corresponding to the time point column.

You can now go on to setup the other options for analysis.