Segmentation

From TLMtracker

Jump to: navigation, search

The segmentation is an image operation process that accomplishs the recognition of cells by TLM-Tracker. In addtion, cell properties like size, position and fluorescence intensity are achieved.

Contents

General

Tlmtracker segmentation.png

The Segmentation process panel is centered at the bottom of the main window. The process is divided into two parts.

The first part is the indexing step. It consists of a series of image processing operations with the purpose to convert an original image (either brightfield or fluorescence) to an indexed segmented image. This means that a region in the final image, corresponding to the area of one cell in the original image, is filled with just one index. This indexing process is conducted either by loading a predefined process from file or by creating a customized image segmentation process. A failure of the indexing step results in a failure of the complete segmentation procedure.

The result of the indexing step should look similar to this image:


The second part of the segmentation process performs the determination of the properties of the segmented cells. These properties are amongst others:

  • dimension, position, orientation and area of the cell
  • mean, maximum and standard deviation of the fluorescence intensity of the cell

The first set of properties is determined from the segmented image. To obtain the second set of properties the segmented bright field image is combined with the fluorescence image. Because both images show the same section it is possible to determine the fluorescence values for each cell.

When the Execute button in the Segmentation process panel is pressed the segmentation process for a certain image is conducted to the last step to obtain the indexed segmented image and afterwards the properties of the segmented cells are determined. As a result the recognized cells are marked with frames on the original image in Screen 1.

Segmentation process panel (1)

A predefined segmentation process can be loaded by selecting the corresponding radio button (Watershed, Threshold, Active Contours - Chan-Vese, Active Contours - GAC or Active Contours - Hybrid). The selected process can be executed for all images by clicking the Execute all button or for arbitrary images by pushing the Execute button which opens dialogue windows described here.

Details/Edit

By clicking the Details/Edit button the Edit process panel is opened to verify or modify the image segmentation process. All changes made in this panel only affect the currently selected image. The segmentation process is transferred to the other images by pressing the Execute all button for all images or the Execute button and choosing the range of images that shall be segmented.

New

Clicking the New button opens the Edit process panel without any predefined image processing steps thereby giving the opportunity to start a new segmentation process from scratch.

Load

With this button a predefined segmentation process is loaded from file and the image processing operations of this process line are assigned to the current image. There are several predefined processes, which are described and explained here. The standard segmentation processes subjecting the radio buttons can be found in the folder processesSTD whereas some more predefined processes can be found in the folder processesCST. The latter is defined as default directory for altered standard processes and custom segmentation steps.

Save

The function of this button is to save the image processing operations of the current image as a segmentation process to file. So it is possible to develop and reuse customized segmentation procedures. When the file is saved, a thumbnail for the process is generated, showing the first and the last image of each process line.

Execute/Execute all

By pressing the Execute or Execute all button the two steps of the segmentation process described here are performed. The Execute all button allows to copy and execute all image processing operations from the current image to the complete loaded image series without any further confirmation dialogue. Contrary, pushing the Execute button opens a dialogue window which provides a selection for which images the segmentation will be executed.

Reset

The Reset button provides the possibility to clear all operation processes from the current image. Thus, the image is reset to its initial raw data status.

Execute segmentation.png

It is possible to choose to execute the segmentation just for the current image, a range of images, to execute the segmentation for the following or for all images. In addition it is possible to choose if the image processing operations are copied from the current image to the images which shall be segmented or not. Afterwards the result is shown by marking the recognized cells with blue frames in Screen 2 on the original image.

Manual correction (2)

The 'Manual correction' panel provides tools for manual changes in presegmentated images. These tools comprise 'Paint Pixels', 'Fill Area' and 'Fl Bg Correction' (Fluorescence Background Correction) which can be also found in the toolbar (Tool shape stroke.png, Tool shape fill face.png, Tool pushpin.png). To use these tools segmentation data for the current image has to be already present.

  • The 'Paint Pixel' button allows to print white (left mouse-click) and black pixels (right mouse-click). These can be used to manually devide cells which are not correctly segmentated.
  • The 'Fill Area' button provides the opportunity to fill complete areas with white (left mouse-click) or black color (right mous-click). Thereby, false identified objects can be removed.
  • The 'Fl Bg Correction' tool is recommended to mark the background at various positions which should be substracted from the flueorescence values of the cells, if this kind of data is present. When the button is selected left-clicking on screen 1 sets a marker at this position, right-clicking on a marker removes it. These markers should be set at positions where background can be found in all images of a sequence. The background marker boxes will visually disappear if the corresponding button is deselected but their positions are already saved.

The 'Reset' button removes all manual correction from the current image, even if already executed. All manual changes have to be applied to the current image by pushing the 'Execute' button in the 'Manual correction' panel. It is also possible but not recommended to make all desired changes in all images and finally pushing the 'Execute' button in the 'Segmentation process' panel with the 'All images' option but without selection of the 'copy operations from this to target images' checkbox.

Continue

By pressing the Next button in the Workflow panel or selecting the appropriate entry in the Workflow menu you will reach the next step in the workflow, the Tracking.


Common steps

Go to the first image of a sequence and load a segmentation process using the predefined segmentation operations subjected to the radio buttons or the 'Load'-button. Afterwards one should inspect the segmentation results in screen 1 and screen 2 for each image. Further investigation can be done using the 'Details/Edit' button to open the 'Edit process' window allowing to go through every image processing step and to compare the results to the images shown for the process here. If an operation doesn't have the expected results, it is necessary to edit it using the 'Edit'-button. Usually it should be possible to get the desired result by manipulating the operation parameters. Finally press the 'Preview'-button to test the process. The recognized cells are shown on screen 2.

Another possibility is to create a new segmentation process adapted for a particular sequence of images. This can be done by pressing the 'Details/Edit'-button and by altering, adding or removing image processing operations and process pipelines as needed. Alternatively, the 'New' button can be used to start a segmentation process from scratch. The last image of the last process pipeline has to be an indexed segmented image for the following steps to work. A good orientation for the creation of new processes are the existing predefined process lines which can be inspected and modified. A custom process can be saved by pressing the 'Save'-button for a later reusage with other images.

When the process works for the first image, press the 'Execute all'-button to transfer it to all images in the sequence. So the segmentation process defined for the first image is executed for the whole sequence. Afterwards it is necessary to inspect the results in the single images. Therefor move through the sequence frame by frame to check if the segmentation had the desired effect. If the quality of the images changes gradually it could be necessary to modify the process beginning at a certain image. When the process works for this image, press the 'Execute'-button and select 'Following images' with the option to copy the process to these images. Then continue inspecting the images step by step.

If there are changes which have to be made for single images, like a cell which is recognized as one cell in this frame but as two different cells in the previous and the following frame, the process can be modified for just this image. Afterwards press the 'Execute'-button and select for this image. Single images can also be corrected manually using the paint and fill tools in the 'Manual correction' panel or in the toolbar. Press the according button and then modify the image in screen 1 by using left and right mouseclicks. Use the 'Execute' button in the 'Manual correction' panel to apply the manual painting to the current image (all manual corrections can be undone by pushing the 'Reset' button). When all changes for this image are done continue inspecting the following images.

If there are fluorescence images present markers for the background of the images should bet set. This is done by using the mark background tool from the 'Manual correction' panel or from the toolbar. When the button is selected left-clicking on screen 1 sets a marker at this position, right-clicking on a marker removes it. These markers should be set at positions where background can be found in all images of a sequence. Like the cells will normally grow over the sequence, usually the best way is to set the markers in the last frame. These markers will be used to apply a background correction to the fluorescence values of the cells. The background marker boxes will visually disappear if the corresponding button is deselected but their positions are already saved. Use the 'Execute' button in the 'Manual correction' panel to apply the background makers (all manual corrections can be undone by pushing the 'Reset' button). If this is not desired, the markers simply should not be set.

After inspecting all images it could be a good idea to reexecute the segmentation for all images by pressing the 'Execute'-button in the 'Segmentation process' panel, but this time without!!! the option of copying the operations to the target images selected. This will apply the changes made to all images in the case this has been forgotten for single ones. If the copy option is set, the process of the current image is copied to all target images, overwriting all individual changes, that have been made!

As a last step it is possible to add user defined data to every single recognized cell, if this is necessary. By right-clicking on the cell in screen 2, a window appears where a string or a number can be entered which will be saved for this cell. This data will be exported in the last working step together with other cell specific properties.

Personal tools