Learn more about neurosphere formation assays:
- Belenguer G, Domingo-Muelas A, Ferrón SR, Morante-Redolat JM & Fariñas I. (2016). Isolation, culture and analysis of adult subependymal neural stem cells. Differentiation. 91(4-5):28-41. doi:10.1016/j.diff.2016.01.005
- Fiji
- ilastik 1.3.3
- BaSiC update site (Fiji)
- ilastik update site (Fiji). Once added, it is important to configure the ilastik executable location within the Fiji's plugin.
- Morphology update site (Fiji)
- Image dataset following an IN Cell Analyzer file naming convention (note that the NeuroMol update site includes a macroinstruction to turn data acquired with diferent high content microscopes into an IN Cell Analyzer file naming convention dataset)
How to follow an update site in Fiji
- Start Fiji
- Start the ImageJ Updater (
Help > Update...) - Click on
Manage update sites - Click on
Add update site - A new blank row is to be created at the bottom of the update sites list
- Type NeuroMol Lab in the Name column
- Type https://sites.imagej.net/NeuroMol-Lab/ in the URL column
Closethe update sites windowApply changes- Restart Fiji
- Check if
NeuroMol Labappears now in thePluginsdropdown menu (note that it will be placed at the bottom of the dropdown menu)
Download an example image dataset and some ilastik classifiers which have been trained using the same subset of images and user annotations but different sets of features. Please note that the image dataset includes a TXT file containing some technical specifications, which are crucial to set the macro parameters. Additionally, along with the ilastik classifiers, another TXT file provides major information regarding their technical specifications, as well as their performance, in order to guide the user when picking up one for the analysis.
- Run the Illumination Correction macro (
Plugins > NeuroMol-Lab > Neurosphere Assay > Illumination Correction) - Select the directory containing the images (TIF files)
- Select the illumination correction mode. The retrospective multi-image method is the easiest to apply, since it does not require reference images. However, since this method calculates the illumination correction from the image datset, it requires a medium-size dataset in order to perform a proper estimation (>25-30 wells). Alternatively, the prospective method uses reference images, e.g., images acquired on plates without any specimen, just the background. Note that this method will ask for an additional folder containing the reference images to apply the correction function.
- Run
- Corrected images will be saved as H5 files in a new subfolder (named adding the tag corrected) within the original directory, whereas illumination functions will be saved as TIF files in a different subfolder (named adding the tag flat-field).
- Start a pixel classification project in ilastik
- Load some corrected images (H5 files). Try to use images with different appearence, from several field-of-view positions and wells
- Select some features (you can start by selecting all of them)
- Annotate the images. Note that label 1 must correspond to background, label 2 to neurospheres and label 3 to well edges
- If it takes too much time to perform the classification, ilastik includes implemented algorithms to keep the major features (look at the features applet)
- Once the classifier performs fine, set up the prediction export applet: i) set Probabilities at the souce drop-down menu; ii) check Renormalize (from 0 to 1)
- Save the project
Learn more about the ilastik pixel classification workflow
- Run the Pixel Classification & Stitching macro (
Plugins > NeuroMol-Lab > Neurosphere Assay > Pixel Classification & Stitching) - Select the directory containing the corrected images (H5 files)
- Load the ilastik pixel classifier (ILP file)
- Select the stitching mode. Note that it depends on the image acquisition parameters. Also note that the macro assumes an overlapping of 15 %
- Adjust some workflow parameters (mean, threshold, open iterations). Know more about the parameters of the workflow on the wiki page (not yet)
- Run. Note that the stitching workflow may take several hours, depending on the dataset size
- A series of new files will be saved within different directories: stitched brightfield and probability maps files (TIF), a results table file (CSV) and the ROI files (ZIP) of each well
Since the Pixel Classification & Stitching macro requiures a significative amount of computational time (which is mainly due to the pixel classification and, to a lesser extent, the stitching step), the Segmentation macro enables to reset the segmentation parameters and quickly test them on the already stitched probability maps.
- Run the Segmentation macro (
Plugins > NeuroMol-Lab > Neurosphere Assay > Segmentation) - Select the directory containing the stitching output (a folder tagged with the postfix corrected_stitched). The folder must contain two subfolders: Pixel_Prediction_Map and Raw_Data. Please note that the TIF files stored on both subfolders should never be reallocated or modified in any way in order to be reused with the Segmentation macro.
- Adjust some workflow parameters (mean, threshold, open iterations). Know more about the parameters of the workflow on the wiki page (not yet)
- Run
- Note that the macro will generate a new results table (CSV), but the ROI files (ZIP) will be aur¡tomatically replaced. The stitched images, both brightfield and probability maps, will not be modified.
Neurosphere Assay in licensed under MIT