User Tools

Site Tools


Tapas Tutorial : Layers analysis


In this tutorial we will learn some basic analysis functions of TAPAS. We will learn how to use the EDT and EVF modules to create continuous layers inside a structure. Then we will use the EVF layers analysis module to quantify the distribution of spots within the structure. Please check you understand the basics of TAPAS. You can also check this tutorial on segmentation with TAPAS.

Computing the layers

The EVF (Eroded Volume Fraction) can be regarded as a normalized EDT (Euclidean Distance Map), it was first described in this article. The EDT will compute for each pixel inside the structure of interest the distance to the border of the structure and output a map where the pixel value is actually the distance.

Here we assume we already have a binarised structure image with value 0 for the background. First we need to input the binary image of the structure, compute the EVF and save it locally. The input data can either be a file on disk or an image in a OMERO database as explained in the tutorial on Input/Output.

// We use here the raw data as the reference image and
// we assume the binary image has the same name 
// as the reference image with -structure

// compute the evf 
// this module computes the edt map by default
// we put yes to ask the module to compute the evf instead of the edt

// and we save the evf image locally
// here in the ImageJ/Fiji folder

Binary image EVF image

Layers analysis

The values in the EVF are normalized between 0 and 1 and, then can be used to compute layer distribution inside many structures regardless of their size. Furthermore, in case of a random distribution the distribution within EVF layers should be flat since, in EVF, all layers have equal volume. Note that using EDT to compute distribution within layers may lead to biased results, since volumes based on distance to border will not have equal volumes. Compare the EVF on the left and the EDT on the right.

EVF image EDT image

We will use the evfLayers module to compute distribution of spots within a fixed number of layers (of equal volume) within the EVF image. We need to define the number of layers and the path to the corresponding EVF image. The input for the evfLayers analysis should be a binary or labelled image of the spots, the module will compute in each layer the volume occupied by the spots.

// We use here the raw data as the reference image and
// we assume the binary spots has the same name as the reference image with -spots
// input corresponding spots 

// layer analysis
// path to the previously saved evf image
// and where to save the results
// and number of layers (100 by default)

The output of the evfLayers module will be first the plot and values for the input data, respectively as PNG and csv file. In this example with 100 layers, a perfect random distribution should be flat with value 0.01. We can then see in this example a trend toward low evf values, indicating that the spots are mostly localized near the border of the structure.

EVF layer plot

The module will also output with the suffix -all the plot for volumes of the different layers, due to numerical approximation, the volumes of the layers are not exactly equal, this output can be hence used for further normalisation.

EVF layer plot all

Saving the results and cleaning temporary files

The results are saved as temporary files, we must attach them to the original raw data. The module attach will attach results file to the image within Omero or will copy the results in the same folder as the original image if data are on files.

// attach files to data in omero or on files
// first attach png and csv of the spots layer analysis

// then attach png and csv of the volume layer analysis

Finally, we can delete the temporary files.

// delete evf

// delete plots
plugin/utilities/layers_analysis/start.txt ยท Last modified: 2019/09/12 04:24 by tboudier

Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki