Count/Size Protocol Segmentation

The segmentation step of an Count/Size Protocol protocol defines the members of object classes. During this step, you are prompted to set segmentation parameters.  The segmentation method can be either Threshold or Smart Segmentation.

Image-Pro prompts you to adjust the thresholds during a Threshold segmentation in order to separate objects of interest from the image background.

If you wish to make adjustments using the threshold tool, simply click and drag upper and low edges of highlighted area in the Threshold Tool panel until the desired limits are set. Clicking the OK button in the interactive balloon sets the threshold values for the protocol and re-enables the protocol panel controls. The number of objects found in the image is reported at the bottom left of the protocol panel.

The following segmentation and measurement controls may be found in the segmentation step (depending on the protocol):

Threshold or Smart Segmentation Radio Button

Click the radio button of the desired method of segmentation, and click the Adjust button to access the segmentation controls.

Derived

Use derived outlines to select objects from a class selected in a preceding segmentation step.  Using derived outlines eliminates the need to segment objects in the current segmentation step.

Grow/Ring

These tools provide options to either grow or shrink segmented objects, and to modify the objects as Rings. The Grow control defines the change of size that will be applied to objects.  Positive values will grow objects, negative values will shrink them. The units are pixels. The offset control is only active if the Ring mode has been selected.  This control defines the distance (in pixels) of the inner ring outline from the original outline. Negative values will move the inner line boundary to the inside of the objects, whilst positive values will move the boundary outside of the objects.

Mask

Mask can be activated to perform logical operations between current the current segmentation objects and the objects from a previously defined class. The outlines can be combined with AND, NAND, OR or Seed options. When the Seed option is selected only outlines with Seeds will be counted and if the outline has multiple seeds, it will be split using a seed and fill algorithm.

Filters/Clean borders

Theses controls allow you remove objects that touch the image border or are below an area threshold. Clicking the Edit Range button allow you set both minimum and maximum valid areas. The Edit Range dialog can be used to define filters for other measurement types, such as Intensity (Max).

Split/Fill holes

Select the Fill Options to fill holes within segmented objects. Set the Split option to either Watershed or Boundary Shape to separate touching objects that can’t be separated by segmentation alone.

Reference size

If the protocol uses Relative Size measurements to provide an estimated count of the clustered objects, the size of the reference object can be defined either in the edit box of by clicking the Select button.  After clicking the Select button pick a segmented cluster of objects on the image and define the number of objects in the cluster.

Vectorize

This option can be used to vectorize segmented objects and extract tree components of the vectorization, such as Skeleton, Branches, Nodes and End points. This can be useful of analysis of elongated objects such as neurons, and this feature is utilized in predefined Neurite outgrowth and Angiogenesis protocols. The Link parameter defines the size in pixels the skeletal gaps that will be automatically joined. The Prune parameter is the minimum length of branches in pixels (branches shorter than this value are deleted).

Classification

The classification option can be used to classify segmented objects into groups.  This classification is based on one of the measurements (e.g., Particle diameter) using Single measurement classification, or more complex, multi-parameter classification using Learning classification. The classification can be defined by clicking the Classes button.

Interactive/Ignore

The “Ignore on result image” option can be selected if the segmentation results of a step are used only as inputs for other steps and should be Ignored on the final results image.

The “interactive step” option can be selected if the protocol must always pause on this step allowing user to adjust some parameters interactively.  This option makes the processing of large datasets much slower and requires considerable input from you during execution.

There are also Custom controls that can be displayed in the panel to perform actions specific to some protocols (see specific protocol descriptions for custom controls).

Optionally, other measurement options exposed in ribbon tabs and panels can be adjusted to refine protocols. The Apply button must be clicked in the Protocol panel to re-count objects and to save the measurement options to the protocol step.

Segmentation steps will be executed for all of the object classes in a protocol.