Reconstruction of the Human Airways - Computational Fluid and Particle Dynamics in the Human Respiratory System - page 52

The digital representation of an image can therefore be represented as matrices
which are simply in the form of a 2D or 3D array. This forms the basis for segmen-
tation which is performed by matrix calculation and manipulations. Segmentation
is the process of extracting a region of interest from an image. More precisely, it is
the process of assigning a label to every pixel in an image so that pixels with the
same label can be extracted, identified or categorised. From a respiratory simulation
perspective this involves extracting a 3D region from a series of 2D cross-sectional
monochromatic DICOM images. This process can be manually performed by select-
ing individual pixels on the cross-sectional slices or automated through segmentation
algorithms.
In manual segmentation the user selects the region of interest in every cross-
sectional image in a set of scanned images which can have up to 1,000 images.
Therefore this procedure is time-consuming and lends itself to inter-observer and
intra-observer variability. Fully automated or semi-automated segmentation algo-
rithms for monochrome images are generally based on discontinuities or similarities
within the image based on greyscale values (referred to as intensity). For example,
discontinuities at edges in an image can be identified based on abrupt changes in
the intensity (greyscale level). Similarity in the intensity can be used to extract a re-
gion which exhibits similar properties according to a set of predefined criteria. This
section provides the reader with an introduction to the field of image segmentation
and provides a description of some of algorithms among the many that exist in the
literature.

3.3.2 Edge Detection

The basic idea behind edge detection is to locate an area in an image where the in-
tensity changes rapidly. Edge detection of an image significantly reduces the amount
of data and filters out useless information, while preserving the important structural
properties in an image—an important step in the segmentation of the respiratory
airways. There are a number of algorithms for this, but these may be classified as
either:

• derivative based —where the first derivative of the intensity is greater in mag-

nitude than a specified threshold case of first derivative at the edge of the image
there is a rapid change of intensity or

• gradient based —to find regions where the second derivative of the intensity has

a zero crossing (i.e. a point where the sign of a function changes from positive to
negative or vice-versa and represented by a crossing of an axis with zero value).
In gradient based edge detection a gradient of consecutive pixels is taken in x and
y direction.