-
Notifications
You must be signed in to change notification settings - Fork 0
Resulting Data
Once you have set up and ran the pipeline, you will get a sind result_data.csv file.
It is a text file that features the results of the statistical analysis from every input file pooled together.
Every line in the file represents an input image each, with the statistics separated by ;.
You can use the corresponding import function of Microsoft Excel or its equivalents to read this file and display it as a table.
This section features every header within this file and how you can interpret them.
Note: For skeletonization purposes, the input image is scaled up before the statistics are calculated. As such, area specific endpoints are calculated for the upscaled and downscaled (adjusted for the original image size) variants. For biological measurement purposes, you should read only the downscaled entries.
Note: All of the statistics presented in this work are calculated using pixels as a measurement unit. Make sure you know the ratio from pixels to micrometers in your input image, and convert accordingly.
The combined filename (nucleus staining and neurite staining), the following data is derived from.
The number of nodes contained in the skeleton. This includes start / stop nodes as well as branching nodes.
The number of nodes within the skeleton that have at least 3 other nodes connected to them. This number can be interpreted as the total number of branching points within your nuclei.
Because a pruned skeleton is computed, the shape that can be reconstructed with the skeleton only approximates the original neurite shape. The Hausdorff Distance measures the difference between the orginal neurite shape and the shape represented by its corresponding skeleton. If more than 1 shape (neurite) is given, the output is the average over the Hausdorff distances for each shape. This value can change depending on the given pruning parameter (epsilon). Here, lower values imply that the skeletons present a (mathematically) good representation of the original input.
If you use the same input parameters and data but exclusively different sensible values for epsion, the lowest HD value indicates the most accurate results overall. So, if you want to find a suitable epsilon parameter you can use this metric with a single neurite per input image to benchmark your results or approach sensible epsilon values iteratively for your in vitro image quality environment.
Number of milliseconds this file needed to be processed.
Number of pixels within every skeleton.
See entry (5), but scaled down to the original image dimensions.
Converted to micrometers, this value can be interpreted as the combined lengths of all neurites within the image (including soma).
Number of pixels within every skeleton that does not intercept with the soma masks.
See entry (7), but scaled down to the original image dimensions.
Converted to micrometers, this value can be interpreted as the combined lengths of all neurites, within the image, but not including areas covered by the soma.
This is the number of soma masks. Since every soma originates from a nucleus, this number can be interpreted as the number of nuclei, within the input image set.
Entry (6) divided by entry (9).
Converted to micrometers, this value can be interpreted as the the average length of a neurite per nucleus, within the image pair (including soma).
Entry (8) divided by entry (9).
Converted to micrometers, this value can be interpreted as the average length of a neurite per nucleus, within the image, but not including areas covered by the soma.