Segmenting Images Using Contours and Masks

Jump to: navigation, search

In this chapter . . .

Segmentation is the process of identifying selected regions of images as members of a common group. Physicians routinely segment structures in medical images to facilitate the treatment of patients. The quantification of important attributes, such as the volume of various tissue types, enables researchers to better understand, diagnose, monitor, and treat neurobehavioral disorders.
MIPAV provides several segmentation tools that allow you to delineate a volume of interest (VOI). You can indicate a VOI by drawing a contour or by creating a mask. You can also calculate predefined statistics on the indicated VOI.
For more information refer to the MIPAV website.
Once a VOI is segmented, you can also apply an algorithm to the volume bound by the contours. MIPAV provides a number of algorithms, such as Gaussian blur, Laplacian, threshold, and watershed. These algorithms and others are addressed in detail in volume 2 of the User's Guide.

Understanding contours and masks

Before you indicate a volume of interest, it is helpful to determine whether to delineate the VOI by drawing a contour around a structure or to generate a mask.

A contour is an array of points. If the contour is a closed polygon, the area inside defines the VOI. Contours are scalar and can be grouped to form a single multicontoured VOI. Contours are particularly useful if you want to save a VOI that covers a large area because the resulting file is generally smaller than a comparable mask file.

A mask is a filter that selectively includes or excludes certain values. You can create a mask that includes the VOI areas and excludes all other areas. Masks are particularly helpful if you use other image analysis programs that require you to supply a mask. Masks can also be used to indicate a complex area of interest. MIPAV allows you to generate binary and short image type masks. Also, you can manually create masks using paint tools.

Segmenting Images Using Contours and Masks: Using contours to segment a VOI