Registration of 3D multiplex images with one common chanel
Project description
multireg
Registration of 3D multiplex images with one common chanel, based on itk-elastix.
Napari plugin to align 3D stacks that have one common field of view in one chanel used to calculate the alignement. The plugin will apply the registration to all other chanels and output one final stack with all the aligned chanels.
The stacks must have one common chanel (typically cell junctions and nuclei) that is used to calculate the registration transformation. It can be rotated, translated, deformed, and with a wider field of view. Then the calculated transformation is applied to all the other chanels for each stack.
The final result is one multi-chanel 3D stack, with the first chanel being an average (or not) of the common chanel and each other chanel the registered chanels from the multiple stacks. The common chanel can be averaged between the different chanels, which improves its quality.
The plugin save and load files to a folder named aligned
and created in the same directory as the source images.
Example of usage of this module is in the case of imaging the same cells with washing out or moving the sample in between. The corresponding cells will not be at the same position in the new stacks, and can even be deformed by the procedure. This plugin realign the images based on one common chanel on which the transformation is calculated.
Installation
-
You can install the plugin directly in
Napari
by going toPlugins>Install/Uninstall plugins
and search formultireg
-
Or you can install
multireg
via pip:pip install multireg
Usage
You can launch multireg
in napari by going to Plugins>multireg: do multiplex registration
.
Fixed image
It will open a prompt to ask you to select the reference (fixed) image, compared to which all other images will be aligned.
Then you have to choose the reference chanel
that will be used in all the stacks to calculate the alignement. So this chanel should be common to all stacks.
Reference points
The first part of the registration relies on reference points manually selected, because the common field of view can be quite far from each other in the acquisition. So first a affine registration is applied to bring close the region of interest between the two stacks to match.
Note that if your stacks did not move a lot then you could calculate the transformation without using the reference points. There's an option in the alignement calculation panel for this.
You have to manually placed a few reference points (4-5 should be enough). Try to spread them in the image (in x,y and z) on landmarks to recognize them in other images.
To add a new reference point, click on the "plus" sign in the left panel. To select one, click on the arrow icon (or press 3), then on the point. You can move the point in x and y. To move it in z, press u
for up and d
for down.
When all points are placed, save them. The points have to be saved to be correctly loaded by the alignement calculation step.
Then click on Fixed points done
to continue to the next step.
Moving images
Then you can choose one of the images you want to align with the reference image. Its chanel that is common to the fixed image should be the same chanel, selected in the first step (the reference chanel
).
Moving points
You now have to locate where the region of interest (the fixed image) is in your new image and find the landmarks referenced in the fixed image are in this new image. This allows the plugin to put together the region of interest in the two images in a first step, before to fine-tune the registration.
For each point placed in the fixed image, place the corresponding point in the moving image. By default, the moving points are placed close to the fixed points.
-
Each point must have the same label (number) as its corresponding fixed points to associate them correctly. You can change a point label by selecting it and putting the new value in
param
and clicking onupdate
. -
When a point is selected, you can drag it to its desired location. To move it in the Z direction, you can press
u
to move it to the next Z (up direction) andd
(down) to the previous Z. The viewed slice will also move, following the point new position, when you do so. -
You can click on
side_by_side
view to see the two images (fixed and moving) with their placed points at the same time.
When all the moving points have been correctly placed, click on Save points
to save this positions and let it be usable by the alignement step. The points have to be saved in the point file to be correctly loaded in the alignement step.
Alignement calculation
This step is the core of the plugin. The transformation necessary to change the moving image to match with the fixed image on the reference chanel
is calculated based on itk-elastix python module. It is decomposed in two steps. (1) First a global affine registration is performed, based on the correspondance between the reference and moving points. This allows to locate the fixed image postion within the moving image and apply a first shearing, scaling, rotation and translation to super-impose the region of interest. (2) The second step fine-tunes the registration. It doesn't use the reference point anymore but calculate the matching based on the images local intensities. Non-rigid transformation based on B-spline is performed at this step, thus allowing to compensate for local deformations in the moving image.
Apply alignement
Create resulting image
This step allows to save a single 3D multi-chanels stack with all the aligned chanels.
The common chanel present in all the images can be averaged together after alignement to obtain a much less noisy image. By default, the aligned reference chanel
of all the images are averaged together to create the final image first chanel. However, it is possible to unselect some images in the first panel if you do not wish to use all the images or do an average.
Then each aligned chanel of all the images that were not the reference chanel are stacked together in the final resulting image. Here also, if you don't want to keep all the other chanels in the resulting image, you can unselect the one that you don't want stacked. All the aligned chanels have been previously saved in the aligned
folder. If delete_files
is checked (default) all these interemediate files will be deleted and only the final resulting stack will be saved in that folder.
License
Distributed under the terms of the BSD-3 license, "multireg" is free and open source software
Plugin initialization
This napari plugin was generated with Cookiecutter using @napari's cookiecutter-napari-plugin template.
Issues
If you encounter any problems, please [file an issue] along with a detailed description.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file multireg-0.0.8.tar.gz
.
File metadata
- Download URL: multireg-0.0.8.tar.gz
- Upload date:
- Size: 17.8 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.8.10
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 1f52a8ad9296dc896d5060d807d59bb940acb1146e1c96bda74294ccbe3cee04 |
|
MD5 | 5131cb2e877494ee91418e3794d32199 |
|
BLAKE2b-256 | d111cd7ccb6cb83f7f63c18937a27a8f85832130dc502173eb6eef8007621d19 |
File details
Details for the file multireg-0.0.8-py3-none-any.whl
.
File metadata
- Download URL: multireg-0.0.8-py3-none-any.whl
- Upload date:
- Size: 14.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.8.10
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 24887908d7f3f8fda0e6865f4d5a3a97639523ee040504e6351fdec5b3a0b7cc |
|
MD5 | 12c714cd196200d06234164ff23b73f9 |
|
BLAKE2b-256 | fb99f265701889f79955cb0fccc44eadf45f95c28897541ab4336e22637512a0 |