Import spots from labels #47
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@maarzt this PR is the result of a collaboration of @Noam-Dori and myself. It adds the functionality to Mastodon to convert a series of label images to spots at different timepoints. This may be handy if a good segmentation had already been produced outside of Mastodon. We were inspired by some code you provided: https://github.com/mastodon-sc/mastodon-ellipsoid-fitting/tree/compute-ellipsoid-for-segmentation I would be interested in your opinion before merging it. It is not urgent though. |
stefanhahmann
force-pushed
the
import_spots_from_labels
branch
4 times, most recently
from
09e6ea0 to
7891a3a
Compare
stefanhahmann
force-pushed
the
import_spots_from_labels
branch
11 times, most recently
from
e3de381 to
7930b0c
Compare
stefanhahmann
force-pushed
the
import_spots_from_labels
branch
2 times, most recently
from
9ec867d to
ff6d7cb
Compare
|
Kudos, SonarCloud Quality Gate passed! |
stefanhahmann
force-pushed
the
import_spots_from_labels
branch
from
2c35790 to
f824baa
Compare
…mmation (might need more robust fix later), anisotropic dimensions, finding minimum and max pixel value to account for unsigned conversions,and fixed a bug in the covariance math. also, added support for testing, configuration of the sigma parameter, and directly reading integer images rather than cast them to double.
…ucing BigInteger. We have no way to avoid this as we needed a 128bit integer to truly solve the problem.
…tically trivial example that can easily be tested.
…nnel to convert labels into ellipsoids (using the raw image instead of the label image) A warning is thus printed if lots of labels are detected (i.e. more than 10_000)
…porter.labelimage * These respective plugin is importing instead of exporting
…lImageView via the CommandService
stefanhahmann
force-pushed
the
import_spots_from_labels
branch
from
b1a9cde to
4a926fd
Compare
|
|
Closed in favor of #79 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds a plugin to convert label images to (spot) ellipsoids.
Notes from https://github.com/Noam-Dori who contributed to this:
To accomplish the targeted functionality, the plugin runs twice through each image read.
Once to determine maximum/minimum values for array initialization, and once to do summation for covariance & mean. It might be possible to reduce this to 1 run either by finding the min/max stored in the metadata of the image (this is a common and very useful property so it might exist in there), or by changing from an array model to a Hashmap model (at the expense of memory).
One note I saw in the comments was about the inclusion of 5 in the covariance calculation. This is not for the covariance calculation, but rather for drawing the ellipsoid. It seems mastodon draws the ellipsoid at 1σ, which only includes ~46% of the target ellipsoid. This is unsurprising but not what we are looking for. Therefore I set it to draw the ellipsoid at 2.2σ, which when squared yields the 5 in the calculation. Since this is an empirical choice I also included it as a parameter in the plugin dialog.
Finally, a short discussion on summation, since it brought many bugs to the surface. Naturally, calculating covariance requires summation over the squares of the pixel coordinates. This can get very big with dealing with large images, and with large labels. These values got so big that the longs overflew many times. This was only resolved by changing the 1D sums into long integers, and the 2D sums into variable bit depth integers (since 128 bit integers don't exist in Java). Maybe reading integers forces out some unknown factor? Regardless, this is something to look out for.