printlogo
http://www.ethz.ch/index_EN
Institute of Microbiology
 
print
  

E - Define Settings

The settings allow defining the behavior of Enhanced CellClassifier. Some settings can later also be changed in the Main Window (indicated below).

Note: "Save object data" and "Load object data" in the CC main window also loads and saves almost all settings.

For some fields CC can present default values. The choices the user has correspond to the CellProfiler pipeline used for calculating the data. After all necessary changes in the "Define Settings" window are made one can press the return button or close the window to return to the Enhanced CellClassifier main program.

Note: If during training one navigates back from the "Main Window" of Enhanced CellClassifier to the "Define Settings" window some changes are prevented in order to avoid data confusion. However, if one moves back to "Define Settings" from the Main-Window in the "Save-Mode" (see starting Enhanced CellClassifier with limited functionality) all changes are allowed.

Analysis type: will be saved with your data but never displayed.

CLASSES

Number Classes: (2…5). The value indicates the number of classes that can be trained in CC. In many cases these 2 classes will be sufficient; CC can handles up to five classes.

Class names: The names that appear on the panels in the upper right corner of the main window. These names are also used for displaying classes, for instance in the SVM-window after training cells. It makes most sense to define the first class as something equivalent to "positive", (i.e. having a certain phenotype, for instance "mitotic"), the second as "negative" (for instance non-mitotic). The remaining classes could be special cases, for instance "mixture" or "special".

Class colors: Defines the colors used for the display of the different classes. These colors define the appearance of the Uipanels (upper right in the "Main Window") and the colors of the object outlines for display. Colors are in RGB format, each color consisting of 3 values between 0 (no intensity) and 1 (brightest intensity) for red, green and blue, respectively. Type your color of choice (or even 'red', 'green') or use the "Select" button for your choice of color.

Internal value: (CellValues, one integer number for each class). Variable is intended for internal reference within Enhanced CellClassifier. It is easiest to select number 1 for class 1, 2 for 2 and so on. For some specialized applications it might be useful to have other numbers available. If you do not know exactly what you are doing, do not change the default values.

TRAINING AND DATA HANDLING

Big window: displays, which data will be memorized by CC while training.

PLUS button (+): adding new data. You will be prompted to select a CellProfiler object and a CellProfiler measurement for this object.

MINUS button (-): remove data.

Use Up and Down to change the order of the data.

Note: Very important variables. The objects to be named here are the objects whose measurements can be used for classification. Objects might be for instance nuclei or cells. You can mix measurements for different object, but only if both exactly correspond to each other. As an example, in a typical CellProfiler pipeline each nucleus will have a cell surrounding it and each cell will have a cytoplasm within. Therefore you can combine measurements for cells with those for nuclei and cytoplasm but not with measurements of an independent object (like a spot).

If for a later analysis no matching measurements are found within this list, this line will be ignored.

Number objects for training in random mode:

These parameters are explained in the Training section of this manual. Values can be also be changed within the main window of Enhanced CellClassifier.

Behavior if insufficient objects in one image ("continue", "repeat"). In the Training-modes "Random" and "Decision boundaries" a predefined number of objects are selected. Objects already trained in this image will not be presented again. If not enough objects can be found in this image, CC will either "continue" and allow training with a lower number of objects or "repeat" the random search of a new image. If no object is found in an image, CC will select a new image in both cases.

Default Parameter C, Default Parameter gamma: Default values for the training of the SVM-model. Please check the part of the manual dealing with SVM for explanations.

Paths, Files and Well names

Load data path: path to data directory with split data files (directory containing the _handles.mat file)

Note: when browsing to the path, you can also select any file in the vicinity of the _handles.mat file; CC will try to locate the correct files and the correct folder.

Load original images path: path to images used by CellProfiler.

LUCKY facilitates this process. CC loads the output-file and checks, where CellProfiler has stored the images. Images can be found if they have not been moved by the user after the analysis

Load segmented images path: path to LabelMatrixImages (see basic concepts)

As above, LUCKY can accelerate this process

LibSVMPath: path, where the program LibSVM is found

If any of the paths mentioned are incorrect, CC will start with limited functionality ("Save mode"); the paths can subsequently be corrected.

OrigImagePathAddition

The OrigImagePathAddition is only important if more than one experiment (for instance 96-well plate) has to be analyzed. For a simple data-structure of the original images (MasterDirectory\ImageDirectory1\image files) simply type ‘none’. However, some microscopes generate extra folders (for instance MasterDir\ImageDir1\TIFF\image.tif or MasterDir\ImageDir1\VariableDir1\VariableDir2\image.tif

If working with many image directories (for instance in the "directory" mode of the SaveExcelFile window), CC needs to know these additions to the image path.

Type ‘\TIFF’ for the first example and ‘\*\*’ for the second.

Identification of segmented images

CC uses the filename of image 1 to calculate the filename of the LabelMatrixImage. Thereby, STRING1 in the filename of image 1 is displaced by STRING2. In the simplest case, ".tif" is replaced by ".mat". In some cases it might be necessary to make longer replacements.

Identification of Wells

In several instances CC needs to know in which well the current images are situated. In all microscopes we have been using the well name is encoded in the file name (for instance 'A01') preceded by a underscore ('_') and followed by an underscore ('_'), for instance EXPNAME_A01_s1. Change these strings ('_') if your microscope handles filenames differently.

You can also specify the default paths for saving the outputs. All these pathnames can also be changed later.

Default path for saving Excel files

Default path for saving Outlined images

Default path for saving Heat maps

Default path for saving Matlab summaries

WELLS

Here one can define the setup of the plate (96 and 384 well-plates are supported). Enhanced CellClassifier allows defining 4 groups of wells. The default names for the well groups are "positive", "negative", "mixture" and "special"; these names can be changed. During training, one well group is active (for instance "positive"), only images of this well group will be presented for training (usually randomly).

Name: name of the well group. These names will be displayed on the radio buttons in the lower right of the Enhanced CellClassifier "Main Window".

The wells belonging to this group can now be defined, either by typing well names separated by commas, or by a graphical user interface after using the "select" button.

Warning: for a successful start, each group must contain at least one valid well.

Further information: For training it might be useful to predefine several well-populations. Some controls might contain only (or mostly) positive objects =class 1, others only (or overwhelmingly) negative =class 2. Mixture should contain representative wells for the whole experiment and special might contain a difficult to train or an especially important subset of wells.

Note: One well can be in more than one well group.

DISPLAY

Channel 1 will be displayed in grey in the Enhanced CellClassifier "Main window"

Channel 2 will be displayed in green

Channel 3 will be displayed in red

CellProfiler image names: CC can only display the channels that have been loaded into CellProfiler. Choose the channel most important for training to be channel 1, since one is better able to appreciate subtle details in the grey channel.

Note: While Channels 2 and 3 can be empty ("none), channel 1 will always be displayed (cannot be "none").

Max channel 1, 2, 3: Default value: 0.79 for channel 1; 1 for channel 2 and 3. Range: 0-0.79 for channel 1, 0-1 for channels 2 and 3. Choosing a low number dims the respective channel.

ShrinkImage: "on" or "off". If no gap in the LabelMatrixImages exists between objects, CC will be unable to calculate and display object outlines if neighboring objects are marked. In this case, the "on" option should be selected; this shrinks objects by one pixel. This problem exists for instance the case, if cells in a confluent layer are defined by CellProfiler with the IdentifySecondaryObjects module.

MarkCells: "outlines" or "rectangles". One can mark trained objects either by outlines or by rectangles (can also be modified in the main window).

RectangleDistance: Range: 0 to any number of pixels. If objects are marked by a rectangle and this distance is 0, the smallest rectangle surrounding the object will be calculated. Otherwise, the rectangle will be enlarged by the chosen number of pixels.

MouseDistance: Range: 0 to any number of pixels. If this number is 0, you can select an object only by clicking directly on its shape. This might be inconvenient for small or very irregular objects. If the number is larger then 0, all pixels in the respective distance from the mouse click will be considered. The object containing most pixels in that area will be selected.

Enlarge outlined image while generating output: While making "control images" in the "Generate Excel Files" window, images can be enlarged for a better display. Choose a value above 1 for enlarging images, a value below 1 for shrinking images. Larger images clearly look better but take more time to process. Values above 4 might lead to a Matlab out of memory error.

FOCUS:

For the general concept please read the "FOCUS" section later in the manual

Focus Control: if set to "on", out of focus images will be excluded (i.e. a new image is loaded in the Main Window and all measurements are ignored in the "Generate Excel Files" window. This value can be modified in the Main Window.

Focus Objects and Granularity Measurements

We recommend selecting nuclei as the focus objects. In the Focus-Window, the FocusObjects (usually the nuclei) are plotted against the results calculated by the granularity measurements (usually the image granularity of the DAPI-image). Select the desired measurements.

Granularity Method: "Maximum" or "First_Value")

The measured granularity spectrum has 12 values. Depending on the application, either the first value or the maximum of the 12 values are most useful. This choice can also be modified within the Focus-Window.

Line Offset Range: any rational number

Line Slope Range: any rational number

Offset and slope of the line separating images in focus from images out of focus in the plot of "Focus Objects" vs. "Granularity Measurements". These values can also be modified within the Focus-Window.

Focus Direction "below", "above". Default: "below"

Indicates, whether the images in the "Focus Objects" vs. "Granularity Measurements" plots "below" or "above" the separation line will be considered out of focus.

Show Out of Focus Images: If value is set to ‘on’ in the main window, each out of focus image will quickly be displayed in a separate window. This is useful for assessing the performance of the Focus-settings. Simply close the windows CC has opened.

Rescale Images: Use 'on' if you are working with 12-bit images. Explanation: These images probably had been rescaled within CellProfiler for optimal measurements. Therefore, if the FOCUS window works in the calculate mode these images should also be rescaled, otherwise the calculation within CP and the "Focus Window" will not be compatible.

Starting Enhanced CellClassifier with limited functionality

Save-Mode

When Enhanced CellClassifier starts, the settings file is loaded. If this fails, CC gives an error message and switches to the Save-Mode where most functions are disabled and the main window appears grey. Choose "Modify settings" or "New experiment" to correct the underlying problem.

Without original images

The main window is black. You cannot classify cells; however, you can train a model (using formerly loaded cells) and generate output files.

Without segmented images

Similar to "without original images" but you can view the pictures.

Without data selected

You cannot classify cells and cannot train a model.
Note: you might need to restart Enhanced CellClassifier to correct this  

 

Wichtiger Hinweis:
Diese Website wird in älteren Versionen von Netscape ohne graphische Elemente dargestellt. Die Funktionalität der Website ist aber trotzdem gewährleistet. Wenn Sie diese Website regelmässig benutzen, empfehlen wir Ihnen, auf Ihrem Computer einen aktuellen Browser zu installieren. Weitere Informationen finden Sie auf
folgender Seite.

Important Note:
The content in this site is accessible to any browser or Internet device, however, some graphics will display correctly only in the newer versions of Netscape. To get the most out of our site we suggest you upgrade to a newer browser.
More information

© 2014 Microbiology ETH Zürich | Imprint | Disclaimer | 23 December 2009
top