128

Bioinformatics of the Brain

4.3.3.2

Region Growing Technique

The region-growing segmentation technique is a pixel-based method and be-

gins with the selection of a set of initial points called as seed points manually

or automatically, and then regions are enlarged by looking at the similarities

of these selected seed points and neighboring pixels according to a criterion

(like intensity value, texture, shape). The process of implementing the Region

Growing algorithm involves the following steps [42]:

Algorithm 3 : Region Growing Algorithm

1: The seed pixels, denoted as s1, s2, ..., sn, are chosen as the n number of

initial points. Additionally, the regions corresponding to these seed pixels

are identified as C1, C2, ..., Cn

2: Determine the similarity between the seed point si and the pixel value

of neighbouring points. If the similarity measure is less than the specified

threshold value, neighboring point can be considered as element of Ci

region

3: Recompute the border of Ci and the mean values of all pixels in Ci region

are recalculated as new si(s) respectively

4: Continue to perform Steps 2 and 3 until all pixels in the image are allocated

to a specific region

The main difficulties associated with region growing include selecting suit-

able seeds, determining the similarity criterion, and managing the size and

shape of the region.

4.3.3.3

Edge Based Techniques

An edge in an image is a notable local variation in image intensity that is

typically connected to a discontinuity in the image intensity or the image

intensity’s first derivative [35]. Edge detection is a fundamental technique

employed in image analysis and holds significant importance in identifying

the contours of brain tumors.

Various techniques are available for detecting edges, with most falling un-

der the categories of Gradient and Laplacian. The Gradient method identi-

fies edges by locating the maximum and minimum values in the image’s first

derivative, while the Laplacian method detects zero crossings in the second

derivative to identify edges [43]. This section offers an overview of commonly

used edge detection methods, such as Gradient-based, Canny edge detection,

and Laplacian-based techniques.

Gradient Based Operator

The gradient operator, represented by the symbol ▽and specified as a vector,

is the standard tool for determining the magnitude and direction of intensity

changes of an image f. The gradient for a two-dimensional image f(x, y) can