2. Preparing the input files
2.1. File naming requirement
By using images that follow the naming convention in QuickNII, VisuAlign and ilastik, the output files from these software will also comply with the naming convention, and are thereby directly compatible with Quantifier.
The file name of the segmentation and the atlas map corresponding to a particular section must contain the same unique ID. They should follow the _sXXX naming convention, with XXX representing the section number (not restricted to three digits).
The section number should reflect the serial order and spacing of the sections, this is a requirement for QuickNII. For example, _s02, _s06, _s10 for every 4th section starting with section 2.
The XML or JSON file from QuickNII or VisuAlign must contain the unique IDs (check this by opening with Notepad). This happens automatically as long as the images registered in QuickNII contain the unique IDs in the file name.
Quantifier also supports user-defined IDs using regular expressions (RegExp). This means that it is possible to bypass the _sXXX naming convention. While this is not recommended, it is useful in some cases. For more information see the “Help” button in the Nutil interface or contact User Support.
Examples that comply with the naming convention:
Histological image: tg2345_MMSH_s001.png. Segmentation image: tg2345_MMSH_s001_segmentation.png. Atlas map: tg2345_MMSH_s001_nl.flat
Do not include periods (“.”) or spaces (” “) within the file names. This will not work.
As Quantifier scans and detects the “_s” part of the name, the file name should not contain additional “_s”., for example: tg2345_MMSH_ss_s001.png.
2.2. Preparing the Segmentations
Any image analysis software may be used to generate the segmentations as long as they meet the following requirements:
Format: Must be indexed 8-bit or 24-bit images in PNG format.
Quantifier is only able to quantify one RGB colour at a time. Apply one RBG colour to all the objects-of-interest, and specify this colour in the Nutil GUI. For example, the objects here are red, so extract RGB colour: 255,0,0.
Image proportions: Must have the same proportions as the images used to generate the atlas maps.
They do not have the same proportions as the actual atlas maps as QuickNII alters the proportions slightly.
While Nutil supports segmentations up to 32767 x 32767 pixels (as long as there is enough memory installed on the computer: restriction of Qt: QImage class), we do not recommend running Quantifier on images larger than approximately 10000 x 10000 pixels. Downscale the images as much as possible before segmenting them.
They do not need to be the same size as the images used to generate the atlas maps and are typically larger in size. It is recommended to downscale the images prior to segmentation. This has several advantages:
It improves the quality of the segmentation as it removes noise from the image, and makes them more compatible with the machine learning algorithms used by ilasti.
It speeds up Nutil analysis and prevent crashes due to insufficient memory.
The aim is to downscale the images as much as possible but without losing information from the images that is actually important. The optimal scaling factor is determined by trial and error and should be applied consistenty to all the images in the series.
Ilastik: We recommend the Pixel and Object Classification workflows in the ilastik software to generate the segmentations, with the Glasbey lookup table applied with FIJI to visualise the output. See here for instructions.
QuPath: is an alternative that can be used to generate the segmentations. In some cases QuPath may perform better than ilastik.
FIJI ImageJ: Fiji is also useful for generating segmentations.
2.3. Preparing the Atlas Maps
The atlas maps are generated with QuickNII, which applies linear registration of the sections to the atlas, or with VisuAlign that applies nonlinear refinement to an existing QuickNII registration. The VisuAlign atlas maps are thereby more accurate than the QuickNII atlas maps, but also take more time to generate. The atlas maps are customised to match the cutting plane and proportions of the brain sections.
File format: The atlas maps are in .FLAT format and cannot be viewed directly. The image below shows the information contained in an atlas map, but is not the .FLAT file itself.