Navigation and service

Using structure tensor analysis to measure PLI image registration quality

Background and Motivation

The Three-dimensional Polarized Light Imaging (3D-PLI) technology is used to capture high resolution images of thinly sliced sections of post-mortem brains. These images can then be stacked to reconstruct the brain in three dimensions.

After the aforementioned 3D reconstruction has been performed, its quality needs to be checked. This cannot be done by eye because of the amount of sections on the one hand, and small differences caused by slightly different parameter-sets which are not detectable by eye on the other hand.

Slices of a brain

Figure 1: One of over 1000 cross-sections through 1182 slices of a brain; unregistered (left), registered (right)

Our approach

Since the registered stack of slices should be a much smoother image, Structure Tensor Analysis (STA) can be used to determine its quality.

STA builds a matrix for each pixel, summarizing the predominant directions of pixel intensity gradients in its neighborhood. The energy value, i.e., the root of the sum of squared eigenvalues, of these matrices can be used as a measure of registration quality because low gradients result in low energy and vice versa.

STA energy values

Figure 2: STA energy values of the cross-sections shown in figure 1. Energy values of the registered stack are much smaller compared to the unregistered image.

In addition to being involved in algorithm design, implementation, and data analysis, we are also actively developing software tools that can be used by researchers to perform STA with limited programming effort.

Our collaboration partners

This project is being conducted in collaboration with the Fiber Architecture group of the INM-1.

Related projects at Simulation Laboratory Neuroscience